Search

In smart grid management, precise stability prediction is a complicated task that adds to the effective allocation of resources with grid stability. Specifically, demand-side management is considered an essential element of the overall Smart Grids system. Hence, predicting future energy demands is crucial to regulating consumption by aligning utility offerings with consumer demand. This research presents a hybrid deep learning model (Convolutional Neural Network [CNN] with Bi-LSTM) with a two-way attention method and a multi-objective particle swarm optimization method (MPSO) for short-term load prediction from a smart grid. The proposed hybrid model utilizes a two-way attention method at its encoding and decoding stages, in which an encoding attention layer helps to recognize all the essential features from an input vector, and a decoding attention layer helps to resolve the fixed context vector problem by offering better memory capacity. A CNN and Bi-LSTM are used to capture the essential features from the dataset. We also utilize a t-Nearest Neighbours algorithm to pre-process the initial dataset. An MPSO method combines the features of CNN and Bi-LSTM methods, resulting in better prediction accuracy. As far as we know, it is the first work to suggest a dynamic short-term load prediction model that considers different significant features and enables precise predicting outcomes. The performance of the proposed model and existing well-known deep learning models such as Recurrent Neural Network, Gated Recurrent Unit, Long Short-Term Memory (LSTM), Time Series Transformer, CNN-LSTM and various performance measuring parameters MAE, MSE, MAPE and RMSE are calculated on online UCI dataset (Electrical Grid Stability Simulated Dataset). The proposed hybrid model achieved a better prediction result, which proves the efficiency of the proposed model. [ABSTRACT FROM AUTHOR]

The growing demand for consumer-end electrical load is driving the need for smarter management of power sector utilities. In today's technologically advanced society, efficient energy usage is critical, leaving no room for waste. To prevent both electricity shortage and wastage, electrical load forecasting becomes the most convenient way out. However, the conventional and probabilistic methods are less adaptive to the acute, micro, and unusual changes in the demand trend. With the recent development of artificial intelligence (AI), machine learning (ML) has become the most popular choice due to its higher accuracy based on time-, demand-, and trend-based feature extractions. Thus, we propose an Extreme Gradient Boosting (XGBoost) regression-based model—PredXGBR-1, which employs short-term lag features to predict hourly load demand. The novelty of PredXGBR-1 lies in its focus on short-term lag autocorrelations to enhance adaptability to micro-trends and demand fluctuations. Validation across five datasets, representing electrical load in the eastern and western USA over a 20-year period, shows that PredXGBR-1 outperforms a long-term feature-based XGBoost model, PredXGBR-2, and state-of-the-art recurrent neural network (RNN) and long short-term memory (LSTM) models. Specifically, PredXGBR-1 achieves an mean absolute percentage error (MAPE) between 0.98 and 1.2% and an R 2 value of 0.99, significantly surpassing PredXGBR-2's R 2 of 0.61 and delivering up to 86.8% improvement in MAPE compared to LSTM models. These results confirm the superior performance of PredXGBR-1 in accurately forecasting short-term load demand. [ABSTRACT FROM AUTHOR]

The cooperative optimization and dispatch operation of the integrated energy system (IES) depends on accurate load forecasts. A multivariate load, joint prediction model, based on the combination of multi-task learning (MTL) and dynamic time warping (DTW), is proposed to address the issue of the prediction model's limited accuracy caused by the fragmentation of the multivariate load coupling relationship and the absence of future time series information. Firstly, the MTL model, based on the bidirectional long short-term memory (BiLSTM) neural network, extracts the coupling information among the multivariate loads and performs the preliminary prediction; secondly, the DTW algorithm clusters and splices the load data that are similar to the target value as the input features of the model; finally, the BiLSTM-attention model is used for secondary prediction, and the improved Bayesian optimization algorithm is applied for adaptive selection of optimal hyperparameters. Based on the game-theoretic view of Shapley's additive interpretation (SHAP), a model interpretation technique is introduced to determine the validity of the liquidity indicator and the asynchronous relationship between the significance of the indicator and its actual contribution. The prediction results show that the joint prediction model proposed in this paper has higher training speed and prediction accuracy than the traditional single-load prediction model. [ABSTRACT FROM AUTHOR]

In order to improve the efficiency of cloud computing resource utilization and avoid the problem of computing resource allocation and scheduling lagging behind load changes, the author proposes a cloud computing resource on-demand allocation and elastic scheduling method based on network load prediction. Firstly, the author takes the network load data of Wikimedia as the research object and proposes an adaptive two-stage multi network model load prediction method based on LSTM (i.e. ATSMNN-LSTM load prediction method). This method can classify the network load data into climbing and descending types based on the trend and characteristics of the input network load data, And adaptively schedule the input network load data to the LSTM load prediction model that matches its type for prediction based on the classification results. The author proposes a maximum cloud service revenue computing resource quantity search algorithm based on network load prediction (i.e. MaxCSPRNWP algorithm), which aims to improve cloud service revenue as the optimization objective. Under the premise of ensuring task service quality and system stability, the algorithm allocates cloud computing resources on demand and flexibly schedules them in advance based on the predicted network load results. The experimental results show that the ATSMNNLSTM load prediction method proposed by the author can obtain more accurate network load prediction results compared to other load prediction methods, and the MaxCSPR-NWP algorithm, which is based on network load prediction and is capable of effectively converting the network load prediction results into the required number of cloud servers, is the maximum cloud service revenue computing resource quantity search algorithm proposed by the author, not only does it achieve the early allocation and scheduling of cloud computing resources, thereby avoiding the impact of lagging behind in computing resource allocation and scheduling due to load changes on the quality of cloud computing task services and resource utilization efficiency, at the same time, it has also achieved on-demand allocation and flexible scheduling of cloud computing resources with the goal of improving cloud service revenue. [ABSTRACT FROM AUTHOR]

【Objective】With the rapid development of the satellite Internet of Things (IoT), a large number of short-burst users are aggravating collisions and interference among users of the access network. To address this issue, several organizations and individuals have put forward some dynamic access schemes. However, for most of the proposed dynamic access schemes, it is necessary to know the exact number of future time slot access applications. At present, some load estimation schemes have been proposed in the literature, but the accuracy of these schemes is not high, and they can only achieve load estimation for current time slot.【Methods】To solve this issue, we propose a load estimation method based on the leading code state and parameter estimation. A load prediction method based on machine learning is also proposed. The load estimation method based on leading code status and parameter estimation analyzes the relationship between the probability of leading code in different states within the time slot of the satellite IoT and the number of requests for access to the current time slot. It gives the maximum likelihood parameter estimation expression and uses the maximum likelihood parameter estimation method to estimate the current time slot load. The load prediction method based on machine learning takes the estimated load value as its historical data, combining the Long and Short Term Memory (LSTM) network and the Auto Regressive Moving Average (ARMA) model to predict the future time slot load.【Results】The simulation results show that the estimated error of the load estimation method based on leading code state and parameter estimation is less than 1%. The comprehensive error of the load prediction method based on load estimation results as historical machine learning data is about 6%.【Conclusion】The predicted error of the proposed load estimation and prediction method is within the acceptable range, thus offering accurate future slot access requests for dynamic access schemes.

Edge computing is a crucial technology to solve the problem of computing resources and bandwidth required for extensive edge data processing, as well as for meeting the real-time demands of applications. Container virtualization technology has become the underlying technical basis for edge computing due to its efficient performance. Because the traditional container scaling strategy has issues such as long response times, low resource utilization, and unpredictable container application loads, this article proposes a method for scheduling edge computing resources based on the elastic scaling of containers. Firstly, a container load prediction model (Trend Enhanced-Temporal Convolutional Network, TE-TCN) is designed based on the temporal convolutional neural network, which features an encoder-decoder structure. The encoder extracts potential temporal relationship features from the historical data of the container load, while the decoder identifies the trend item of the container load through the trend enhancement module. Subsequently, the information extracted by the encoder and decoder is fed into the fully connected layer to facilitate container load prediction using the dual-input ResNet method. Secondly, Markov decision process (MDP) is used to model the elastic expansion problem of containers in multi-objective optimization. Utilizing the prediction outcomes of the TE-TCN load prediction model, a time-varying action space is formulated to address the issue of excessive action space in conventional reinforcement learning. Subsequently, a predictive container scaling strategy based on reinforcement learning is devised to align with the application load patterns in the container environment, enabling adaptation to the surge in traffic generated by the container environment. Finally, the experimental results on the WorldCup98 dataset and the real dataset show that the TE-TCN model can accurately predict the container load change. Experiments in the actual environment demonstrate that the proposed strategy reduces the average response time by 16.2% when the burst load arrives, and increases the average CPU utilization by 44.6% when the jitter load occurs. [ABSTRACT FROM AUTHOR]

With the growing demand for various IoT applications, fog nodes frequently become overloaded. Fog computing requires effective load balancing to maximize resource utilization. It is essential to determine the load on host to obtain workload consolidation. Various random parameters, including CPU utilization, the number of CPU cores, RAM, memory allocated, memory available, disk I/O, and network I/O are employed to better comprehend host workloads. In the proposed work, the host's load status is detected using an ensemble approach into three categories: under-loaded, balanced and overloaded. Further, the proposed work considers three case studies and varying numbers of virtual machines (VMs) are executed with various parameter combinations. In each case study, a different number of VMs are executed in parallel on two different platforms. In the proposed study, we predicted the load on multiple hosts by employing a variety of advanced machine-learning models. To construct an ensemble model, we selected models with higher accuracy based on retrieved performance evaluation criteria. The ensemble method is applied to deal with the worst-case scenario of the model prediction. For a selected number of case studies, the Random Forest model, Ada Boost Classifier, Gradient Boost and Decision Tree models perform better than other models. These state-of-the-art predictive models are outperformed by our proposed ensemble model and achieves an improved accuracy of nearly 82% by correctly classifying hosts as overloaded, underloaded and balanced. [ABSTRACT FROM AUTHOR]

To address the environmental impact, low efficiency, and poor accuracy of existing power load prediction methods, this study innovatively proposes a power load prediction system that combines wavelet transform with digital twin technology. Compared with similar power load prediction methods, the proposed method achieved the highest power load prediction accuracy rate of 97.26%, with the lowest MAPE and RMSE being only 3.96% each. Our proposed method has good noise resistance and overcomes the disadvantage of traditional power load prediction methods that are easily affected by the environment. Moreover, the false detection rate of the load information data obtained from the power system in the Fuxin area from 2022 to 2023 was less than 5%, further verifying the reliability of the proposed method. This achievement is attributed to the powerful signal processing capabilities of the discrete wavelet transform, advanced pattern recognition and prediction capabilities of these three deep learning network algorithms, and the intelligence of digital twin technology. The combination of these three elements has brought new technological breakthroughs to the field of power load prediction. [ABSTRACT FROM AUTHOR]

Copyright of Study on Optical Communications / Guangtongxin Yanjiu is the property of Study on Optical Communications Editorial Office and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Due to rising demand and expanding economies, forecasting electricity loads is vital for electrical system management. Precise forecasts assure both economic stability and effective utilization. The basis for generating schedules and managing energy is established by prediction, which is crucial for power stations and transmitting facilities. The purpose of this research is to develop an efficient load prediction approach. Hence, this study presents a novel fine-tuned backtracking search-driven log-sigmoid recurrent network (FBS-LRN) framework for improved thermal electricity load prediction. In the proposed framework, the FBS optimization strategy is introduced for recurrent network activated dynamically in long and short term memory (LSTM) with the log-sigmoid function. In the beginning, the FBS optimization approach is employed to improve the LSTM's characteristics to tackle the issue that the LSTM's performance will be impacted by the unpredictability of its internal properties. Next, using the Python platform, the electricity load projection framework, depending on the suggested FBS-LRN will be implemented into practice and examined using several criteria. The comprehensive research reveals that the suggested approach has superior prediction accuracy and efficacy compared to the current models. Planning for power production and use in the electrical system can be aided within thermal by higher-quality load forecasts. [ABSTRACT FROM AUTHOR]

The short‐term load prediction is the critical operation in the peak demand administration and power generation scheduling of buildings that integrated the smart solar microgrid (SSM). Many research studies have proved that hybrid deep learning strategies achieve more accuracy and feasibility in practical applications than individual algorithms. Moreover, many buildings have integrated the SSM on the rooftop with the battery management system (BMS) to enhance energy efficiency management. However, the traditional methodologies only processed the weather parameters and power demand information for short‐term load prediction, ignoring the collected data from SSM and BMS by the advanced metering infrastructures (AMI), which probably improved prediction accuracy. In this research, many accumulated data of building and SSM are collected before methodology implementation. Considering the diversities of accumulated parameters from SSM and BMS, an adaptive convolution neural network long short‐term memory (CNN‐LSTM) is proposed for hourly electrical load prediction. The CNN could extract the critical large‐scale input feature, while the LSTM could achieve better accurate forecasts. The Pearson correlation matrix is calculated for the feature selection scheme from different data units. The hyperparameter tuning is utilized for obtaining the optimized hybrid CNN‐LSTM algorithm. The K‐fold cross‐validation is employed for deep learning algorithm verification, which includes LSTM, GRU, CNN, and Bi‐LSTM methodologies. The results prove that the hybrid CNN‐LSTM achieved outperformed improvements, which are 20.57%, 29.63%, 19.06% in MSE, MAE, MAPE, and 21.24%, 22.02%, 3.82% in validating MSE, MAE, MAPE, respectively. The hybrid CNN and LSTM combined with the feature selection scheme achieve superior predicting accuracies, proving the adaptability ability for integrating into the energy management system (EMS) of the building's SSM. [ABSTRACT FROM AUTHOR]

This study presents an efficient end-to-end (E2E) learning approach for the short-term load forecasting of hierarchically structured residential consumers based on the principles of a top-down (TD) approach. This technique employs a neural network for predicting load at lower hierarchical levels based on the aggregated one at the top. A simulation is carried out with 9 (from 2013 to 2021) years of energy consumption data of 50 houses located in the United States of America. Simulation results demonstrate that the E2E model, which uses a single model for different nodes and is based on the principles of a top-down approach, shows huge potential for improving forecasting accuracy, making it a valuable tool for grid planners. Model inputs are derived from the aggregated residential category and the specific cluster targeted for forecasting. The proposed model can accurately forecast any residential consumption cluster without requiring any hyperparameter adjustments. According to the experimental analysis, the E2E model outperformed a two-stage methodology and a benchmarked Seasonal Autoregressive Integrated Moving Average (SARIMA) and Support Vector Regression (SVR) model by a mean absolute percentage error (MAPE) of 2.27%. [ABSTRACT FROM AUTHOR]

Edge computing is a crucial technology to solve the problem of computing resources and bandwidth required for extensive edge data processing, as well as for meeting the real-time demands of applications. Container virtualization technology has become the underlying technical basis for edge computing due to its efficient performance. Because the traditional container scaling strategy has issues such as long response times, low resource utilization, and unpredictable container application loads, this article proposes a method for scheduling edge computing resources based on the elastic scaling of containers. Firstly, a container load prediction model (Trend Enhanced-Temporal Convolutional Network, TE-TCN) is designed based on the temporal convolutional neural network, which features an encoder-decoder structure. The encoder extracts potential temporal relationship features from the historical data of the container load, while the decoder identifies the trend item of the container load through the trend enhancement module. Subsequently, the information extracted by the encoder and decoder is fed into the fully connected layer to facilitate container load prediction using the dual-input ResNet method. Secondly, Markov decision process (MDP) is used to model the elastic expansion problem of containers in multi-objective optimization. Utilizing the prediction outcomes of the TE-TCN load prediction model, a time-varying action space is formulated to address the issue of excessive action space in conventional reinforcement learning. Subsequently, a predictive container scaling strategy based on reinforcement learning is devised to align with the application load patterns in the container environment, enabling adaptation to the surge in traffic generated by the container environment. Finally, the experimental results on the WorldCup98 dataset and the real dataset show that the TE-TCN model can accurately predict the container load change. Experiments in the actual environment demonstrate that the proposed strategy reduces the average response time by 16.2% when the burst load arrives, and increases the average CPU utilization by 44.6% when the jitter load occurs.

Focused on the problem that the stream computing platform was suffering from the high energy consumption and low efficiency due to the lack of consideration for energy efficiency in designing process, an energy-efficient optimization strategy based on elastic data migration in big data streaming platform (EEDM-BDSP) was proposed.Firstly, models of the load prediction and the resource judgment were set up, and the load prediction algorithm was designed, which predicted the load tendency and determine node resource occupancy, so as to find nodes of resource overload and redundancy.Secondly, models of the resource constraint and the optimal data migration were set up, and the optimal data migration algorithm was proposed, which data migration for the purpose of improving node resource utilization.Finally, model of the energy consumption was set up to calculate the energy consumption saved by the cluster after data migration.The experimental results show that the EEDM-BDSP changes node resources in the cluster can responded on time, the resource utilization and the energy-efficient are improved.

This article discusses a research study on predicting the charging load of electric vehicles (EVs) using spectral clustering and deep learning networks. The proposed method involves using Monte Carlo simulation to generate historical load data, clustering the data using spectral clustering to identify patterns, and constructing CNN-LSTM models for prediction. The method has shown improved accuracy compared to traditional methods without clustering. However, further research is needed to address limitations such as data dependency, uncertainty handling, and model complexity. The study presents a method for electric vehicle charging load forecasting using spectral clustering and deep learning networks. The method involves using gray correlation analysis to measure the similarity of curve shapes and constructing a curve similarity matrix based on morphological characteristics. The study then applies a CNN-LSTM network to extract spatial and temporal features from the charging load data and make predictions. The method combines the advantages of CNNs and LSTMs to accurately predict electric vehicle charging demand under different load patterns. The proposed method can improve the planning and management of charging infrastructure. This article discusses a method for predicting electric vehicle charging load based on spectral clustering and deep learning networks. The historical data of a charging station in Nanjing, China is used to extract charging load features. The data is then subjected to spectral clustering to divide it into different clusters based on charging load characteristics. For each cluster, a CNN-LSTM model is constructed to forecast the charging load. The proposed method is compared to other prediction models and shows superior performance in terms of accuracy. The article presents a method for [Extracted from the article]

With the surge in astronomical data volume, modern astronomical research faces significant challenges in data storage, processing, and access. The I/O bottleneck issue in astronomical data processing is particularly prominent, limiting the efficiency of data processing. To address this issue, this paper proposes a tiered storage algorithm based on the access characteristics of astronomical data. The C4.5 decision tree algorithm is employed as the foundation to implement an astronomical data access correlation algorithm. Additionally, a data copy migration strategy is designed based on tiered storage technology to achieve efficient data access. Preprocessing tests were conducted on 418GB NSRT (Nanshan Radio Telescope) formaldehyde spectral line data, showcasing that tiered storage can potentially reduce data processing time by up to 38.15%. Similarly, utilizing 802.2 GB data from FAST (Five- hundred-meter Aperture Spherical radio Telescope) observations for pulsar search data processing tests, the tiered storage approach demonstrated a maximum reduction of 29.00% in data processing time. In concurrent testing of data processing workflows, the proposed astronomical data heat correlation algorithm in this paper achieved an average reduction of 17.78% in data processing time compared to centralized storage. Furthermore, in comparison to traditional heat algorithms, it reduced data processing time by 5.15%. The effectiveness of the proposed algorithm is positively correlated with the associativity between the algorithm and the processed data. The tiered storage algorithm based on the characteristics of astronomical data proposed in this paper is poised to provide algorithmic references for large-scale data processing in the field of astronomy in the future. [ABSTRACT FROM AUTHOR]

Accurate heat load prediction is the key to achieve fine control, energy conservation, and carbon reduction of regional hydronics. Taking the regional hydronics of a city in the north of China as the research object, the author, respectively uses back propagation neural network (BPNN), genetic algorithm (GA) optimized BPNN (GA-BPNN), and autoregressive integrated moving average model (ARIMA) combined BPNN (ARIMA BPNN) to predict its heat load, and compares the accuracy and applicability of each prediction method. The results indicate that GA-BPNN has the smallest prediction error, followed by ARIMA-BPNN, but the latter requires less data for prediction. In practical engineering, if there is a sufficient amount of data related to heat load, it is recommended to use GA-BPNN. If there is a small amount of data, ARIMA-BP prediction method can be used. [ABSTRACT FROM AUTHOR]

Focused on the problem that the stream computing platform was suffering from the high energy consumption and low efficiency due to the lack of consideration for energy efficiency in designing process, an energy-efficient optimization strategy based on elastic data migration in big data streaming platform (EEDM-BDSP) was proposed. Firstly, models of the load prediction and the resource judgment were set up, and the load prediction algorithm was designed, which predicted the load tendency and determine node resource occupancy, so as to find nodes of resource overload and redundancy. Secondly, models of the resource constraint and the optimal data migration were set up, and the optimal data migration algorithm was proposed, which data migration for the purpose of improving node resource utilization. Finally, model of the energy consumption was set up to calculate the energy consumption saved by the cluster after data migration. The experimental results show that the EEDM-BDSP changes node resources in the cluster can responded on time, the resource utilization and the energy-efficient are improved. [ABSTRACT FROM AUTHOR]

Currently, there is an emerging requirement for applications related to the Internet of Things (IoT). Though the capability of IoT applications is huge, there are frequent limitations namely energy optimization, heterogeneity of devices, memory, security, privacy, and load balancing (LB) that should be solved. Such constraints must be optimised to enhance the network’s efficiency. Hence, the core objective of this study was to formulate the intelligent-related cluster head (CH) selection method to establish green communication in IoT. Therefore, this study develops a chaotic equilibrium optimizer-based green communication with deep learning-enabled load prediction (CEOGC-DLLP) in the IoT environment. The study recognizes the emerging need for IoT applications and acknowledges the critical challenges, such as energy optimization, device heterogeneity, memory constraints, security, privacy, and load balancing, which are essential to enhancing the efficiency of IoT networks. The presented CEOGC-DLLP technique mainly accomplishes green communication via clustering and future load prediction processes. To do so, the presented CEOGC-DLLP model derives the CEOGC technique with a fitness function encompassing multiple parameters. In addition, the presented CEOGC-DLLP technique follows the deep belief network (DBN) model for the load prediction process, which helps to balance the load among the IoT devices for effective green communication. The experimental assessment of the CEOGC-DLLP technique is performed and the outcomes are investigated under different aspects. The comparison study represents the supremacy of the CEOGC-DLLP method compared to existing techniques with a maximum throughput of 64662 packets and minimum MSE of 0.2956.

Recent advancements in Software Defined Networks (SDN), Open Radio Access Network (O-RAN), and 5G technology have significantly expanded the capabilities of wireless networks, extending beyond mere data transmission. This progression has led to the emergence of Virtual Networks (VN) and Network Slicing, enabling industries to enhance their services and applications by establishing virtual networks that utilize shared physical infrastructure. Many works in the literature have considered optimizing the allocation of on-demand slices, assuming the absolute availability of resources and their accurate load. However, accurately allocating future network slices remains challenging due to the error in load prediction, diverse Key Performance Indicators (KPIs), resource price variations, and the potential for over- or under-provisioning. This study presents a two-phase intelligent approach to address these challenges. The framework proactively predicts different slice loads while considering prediction errors in optimizing future slices with varied KPIs in a cost-efficient manner. Specifically, our method utilizes historical load data per service and employs AI-based forecasts for service load prediction. Subsequently, it employs a Deep Reinforcement Learning (DRL) agent on O-RAN’s virtual Control Unit (vCU) and virtual Distributed unit (vDU) to correct errors in prediction and optimize the cost of slice allocation based on service KPI requirements, ultimately pre-allocating future network slices at reduced costs. Through experimental validation against various baselines and state-of-the-art solutions, we demonstrate the efficacy of our proposed solution, achieving a notable reduction (37-51%) in the average cost of allocated slices while inquiring about (1.5-7%) of additional resources compared to the state-of-the-art..

Prediction in the power system network is very crucial as expansion is needed in the network. Several methods have been used to predict the load on a network, from short to long time load prediction, to ensure adequate planning for future use. Since the power system network is dynamic, other parameters, such as voltage and frequency prediction, are necessary for effective planning against contingencies. Also, most power systems are interconnected networks; using isolated variables to predict any part of the network tends to reduce prediction accuracy. This review analyzed different machine learning approaches used for load, frequency, and voltage prediction in power systems and proposed a machine learning predictive approach using network topology behavior as input variables to the model. The analysis of the proposed model was tested using a regression model, Decision tree regressor, and long short-term memory. The analysis results indicate that with network topology behavior as input to the model, the prediction will be more accurate than when isolated variables of a particular Bus in a network are used for prediction. This work suggests that network topology behavior data should be used for prediction in a power system network rather than the use of isolated data of a particular bus or exogenous data for prediction in a power system. Therefore, this research recommends that the accuracy of different predictive models be tested on power system parameters by hybridizing the network topology behavior dataset and the exogenous dataset.

Turnout is the key equipment for realizing the turnback of subway trains. Frequent movements and environmental changes often lead to abnormal increases in resistance during turnout conversion, resulting in jamming failures that directly affect traffic safety and efficiency. In order to effectively foresee the risk of failure during operation and minimize the adverse effects of failure, a real-time turnout jamming failure prewarning method is proposed. Firstly, a weighted grey prediction machine using PSO for weight optimization (PSO-WGPM) is proposed to predict the short-term action load index, and the predicted index series is used to characterize the future changes in the resistance state of the turnout; Secondly, the predicted index and the generated index are cascaded into a risk identification index series, and a multi-dimensional hybrid prewarning feature set with time continuity are constructed by time-domain characteristics and overload statistical characteristics of the risk identification index series; Then, the prewarning feature set is fed into the learning vector quantization(LVQ) network for prewarning discrimination, and an ensemble learning based on a hybrid voting strategy is designed to obtain the final prewarning result, in order to fit the learning environment of unbalanced small-scale samples; After giving an prewarning, the occurrence time range of jamming failure is inferred based on the overload rate of the predicted index series. The experimental results show that: the proposed method can improve the prewarning success rate and effectively control false alarms, with good correctness; it can infer the range time of fault occurrence, achieving prewarning accuracy; the prewarning calculation time is much shorter than the failure advance time, meeting the timeliness requirements for applications.

In the process of the vigorous development of satellite internet of things, random access protocol is regarded as the most suitable communication protocol due to its characteristics of low signaling overhead, high transmission efficiency, high flexibility and easy implementation.However, due to drastic load changes, conventional random access mode only has advantages in the specific load range.Dynamic allocation of random access resources requires the accurate number of time slot access applications.In view of the above problems, a random access adaptive scheme was proposed, which could adaptively selected the appropriate random access mode according to the future load size, so as to improved the system throughput.The load size could be obtained through the proposed load estimation and prediction scheme.When narrowband interference exists in the transmission environment, the spread-spectrum communication means was introduced to made the adaptive scheme of random access run successfully.Moreover, the proposed adaptive scheme also improved the system throughput.

In order to reasonably establish the objectives and compare the potential of development in provincial natural gas planning, innovative exploration was carried out for the natural gas load prediction model. Through data classification, correlation analysis and benchmarking analysis with the developed countries, the medium- and long-term prediction model of economy-driven provincial natural gas load in China was established by the surface fitting method. Meanwhile, the three-dimensional coordinate system of natural gas consumption potential, GDP and primary energy consumption was drawn. Ten southern provinces (municipalities directly under the Central Government) were selected as the research subjects to quantitatively analyze the natural gas consumption demand in 2030. The results indicate that there is a strong demand for natural gas in the future. Under the scenario of a compound GDP growth rate of 5%, the total demand for natural gas in the ten provinces (municipalities directly under the Central Government) will reach 3 155×108 m3 in 2030, which is 2.5 times that of 2021. Among them, Guangdong Province and Hunan Province have the largest absolute increase and the highest growth rate of consumption, respectively. Generally, the research results could provide reference for the market regulation and development of natural gas.

With the expansion and deployment of smart metering in power grid management and control, the need for security protection in the power system is continuously growing. However, the current construction of a comprehensive defense system for terminal data is inadequate. In this paper, we report a study on power loads to address the security challenges facing grid management, using the protection detection response recovery (PDRR) network security model as the basis. Firstly, we design an end-to-end security perception architecture using IoT technology and develop an optimization model for monitoring sensor information. In addition, we construct a data aggregation model that improves adversarial domain adaptation and incorporates deep convolutional neural networks to extract features. The proposed model enhances short-term load forecasting by combining linear predictions from autoregressive models with the nonlinear trend analysis capabilities of deep learning models. The performance of the proposed method is compared with those of the Adam and stochastic gradient descent (SGD) optimizers. Experimental results confirm that the proposed method ensures reliable data transmission, facilitates effective classification aggregation of heterogeneous data, and yields fast and accurate load forecasting results. Furthermore, the proposed method enhances the robustness of the model. [ABSTRACT FROM AUTHOR]

Cloud Computing has immersed researchers in accessing the resources on-demand for deploying various applications by offering infinite services. But, as the demand for cloud resources is dynamic, it significantly affects the load on the system. Thus, this research emphasizes deploying a dynamic and autonomic load prediction framework. This paper proposes an Ensemble Time-Series Approach for Load Prediction (ETSA-LP), which integrates various time-series analysis techniques for predicting CPU and memory utilization. To evaluate the efficiency of the proposed approach, a series of experiments on Google and PlanetLab traces have been conducted in a real Cloud environment. The results were compared according to different performance metrics and models, the accuracy determined and the minimal error rate selected as the best among others. The proposed ensemble approach gives the best performance over the existing models showing the remarkable accuracy improvement and reducing the error rate and execution time. [ABSTRACT FROM AUTHOR]

Heating, ventilation and air conditioning (HVAC) systems account for approximately 50% of the total energy consumption in buildings. Advanced control and optimal operation, seen as key technologies in reducing the energy consumption of HVAC systems, indispensably rely on an accurate prediction of the building's heating/cooling load. Therefore, the goal of this research is to develop a model capable of making such accurate predictions. To streamline the process, this study employs sensitivity and correlation analysis for feature selection, thereby eliminating redundant parameters, and addressing distortion problems caused by multicollinearity among input parameters. Four model identification methods including multivariate polynomial regression (MPR), support vector regression (SVR), multilayer perceptron (MLP), and extreme gradient boosting (XGBoost) are implemented in parallel to extract value from diverse building datasets. These models are trained and selected autonomously based on statistical performance criteria. The prediction models were deployed in a nearly zero-energy office building, and the impacts of feature selection, training set size, and real-world uncertainty factors were analyzed and compared. The results showed that feature selection considerably improved prediction accuracy while reducing model dimensionality. The research also recognized that prediction accuracy during model deployment can be influenced significantly by factors like personnel mobility during holidays and weather forecast uncertainties. Additionally, for nearly zero-energy buildings, the thermal inertia of the building itself can considerably impact prediction accuracy in certain scenarios. [ABSTRACT FROM AUTHOR]

In the contemporary energy landscape, power generation comprises a blend of renewable and non-renewable resources, with the major supply of electrical energy fulfilled by non-renewable sources, including coal and gas, among others. Renewable energy resources are challenged by their dependency on unpredictable weather conditions. For instance, solar energy hinges on clear skies, and wind energy relies on consistent and sufficient wind flow. However, as a consequence of the finite supply and detrimental environmental impact associated with non-renewable energy sources, it is required to reduce dependence on such non-renewable sources. This can be achieved by precisely predicting the generation of renewable energy using a data-driven approach. The prediction accuracy for electric load plays a very significant role in this system. If we have an appropriate estimate of residential and commercial load, then a strategy could be defined for the efficient supply to them by renewable and non-renewable energy sources through a smart grid, which analyzes the demand-supply and devises the supply mechanism accordingly. Predicting all such components, i.e., power generation and load forecasting, involves a data-driven approach where sensitive data (such as user electricity consumption patterns and weather data near power generation setups) is used for model training, raising the issue of data privacy and security concerns. Hence, the work proposes Federated Smart Grid (FedGrid), a secure framework that would be able to predict the generation of renewable energy and forecast electric load in a privacy-oriented approach through federated learning. The framework collectively analyzes all such predictive models for efficient electric supply. [ABSTRACT FROM AUTHOR]

Copyright of Space-Integrated-Ground Information Networks is the property of Beijing Xintong Media Co., Ltd. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

This paper investigates the effect of Support Vector Regression hyperparameters optimization on electrical load prediction. Accurate and robust load prediction helps policy makers in the energy sector to make inform decision and reduce losses. To achieve this, Bayesian optimization technique was employed for the hyperparameters optimization which are then used for the load prediction. The hyperparameters are the regularization parameters and the epsilon. In addition, the effects of sliding window during the load prediction were also evaluated. The sliding window values were varied from 1 to 5. The results showed that the sliding window of 1 had the optimized hyperparameters with the best performing evaluation metrics of 0.01912 and 0.09493 for MSE and MAE respectively. [ABSTRACT FROM AUTHOR]

This research paper presents a data-based energy management method for a vessel that predicts the upcoming load demands based on data from weather information and its automated tracking system. The vessel is powered by a hybrid propulsion system consisting of a high-temperature fuel cell system to cover the base load and a battery system to compensate for the fuel cell's limited dynamic response capability to load fluctuations. The developed energy management method predicts the load demand of the next time steps by analyzing physical relationships utilizing operational and positional data of a real vessel. This allows a steadier operation of the fuel cell and reduces stress factors leading to accelerated aging and increasing the resource efficiency of the propulsion system. Since large ships record tracking data of their cruise and no a priori training is required to adjust the energy management, the proposed method can be implemented with small additional computational effort. The functionality of the energy management method was verified using data from a real ship and records of the water currents in the North Sea. The accuracy of the load prediction is 2.7% and the attenuation of the fuel cell's power output could be increased by approximately 32%. [ABSTRACT FROM AUTHOR]

With the rapid growth of data centers, optimizing energy consumption has become a critical challenge. This paper proposes an energy management framework that integrates Long Short-Term Memory (LSTM) networks for load prediction with a dynamic resource allocation algorithm to optimize energy consumption in data centers. The LSTM model is employed to accurately predict future workloads based on historical data, enabling the proactive adjustment of resource utilization. The dynamic resource allocation algorithm, informed by LSTM-based predictions, dynamically manages server operations and cooling systems through techniques such as dynamic voltage and frequency scaling (DVFS) and server consolidation, ensuring that resources are allocated efficiently in real time. The proposed framework incorporates a multi-objective optimization approach to balance energy savings and system performance. We use genetic algorithms to fine-tune resource allocation based on the predicted load, optimizing for energy consumption and response time. Experimental results show that our method achieves significant energy savings compared to traditional static resource management approaches, with minimal impact on service quality. This research highlights the potential of machine learning-driven optimization algorithms in reducing the environmental footprint of data centers while maintaining operational efficiency.

Music resource-sharing platforms can increase students’ employment paths and enhance their employment ability. Load prediction and resource scheduling algorithms are designed in this paper to construct a music resource-sharing platform based on cloud computing technology. The mean predictor’s predictions are combined with the GRU-CNN model’s predictions, and the weighted average is calculated to enhance the accuracy and stability of the predictions. Then, the integrated migration scheduling method is designed to make the load of the cloud environment more balanced and the resource utilization more reasonable by the four processes of migration triggering, container selection, node selection, and container migration in a coarse-grained dynamic scheduling approach with nodes as objects. Following the completion of the construction, the algorithm that was designed is evaluated for load prediction and resource scheduling, and regression analysis is performed to examine the role of music resource-sharing platforms in promoting employment and entrepreneurship. It is found that after the resource scheduling by the algorithm in this paper, the node operation tension is greatly alleviated, and the CPU resource utilization and memory resource utilization of node Node1 are reduced from 80.96% and 62.69% to 58.03% and 43.36%, respectively. The resources of the low-load node Node4 are effectively utilized, the memory resource usage rate has increased from 6.86% to 49.52%, and the resource usage of each node in the cluster is more balanced and reasonable. At a statistical level of 1%, the impact of music resource-sharing platform development on employment quality is 0.194. In contrast, the industrial structure does not significantly contribute to the quality of employment, indicating that under the current economic situation, emerging economic forms such as music resource-sharing platforms are the main driving force to improve the quality of employment and the development of music resource sharing platforms generally improves the quality of labor force employment.

This paper investigates the effect of Support Vector Regression hyperparameters optimization on electrical load prediction. Accurate and robust load prediction helps policy makers in the energy sector to make inform decision and reduce losses. To achieve this, Bayesian optimization technique was employed for the hyperparameters optimization which are then used for the load prediction. The hyperparameters are the regularization parameters and the epsilon. In addition, the effects of sliding window during the load prediction were also evaluated. The sliding window values were varied from 1 to 5. The results showed that the sliding window of 1 had the optimized hyperparameters with the best performing evaluation metrics of 0.01912 and 0.09493 for MSE and MAE respectively.

Accurate prediction of the cloud workload is essential in cloud computing, especially for cloud resource demand with sudden changes. The existing research on workload prediction approaches for cloud datacentres has one disadvantage: the results of the prediction approach lack generalization and universality. Hence, to improve the accuracy and generalization of cloud workload prediction approaches, we propose a novel prediction approach, called the re-prediction for error correction model based on the change-based three-way decision model, for workload prediction problems in cloud datacentres. The core idea of our approach is to design a re-prediction system for the predicted error sequence, where we explore the re-prediction criteria of the error interval based on the trisecting-acting-outcome (TAO) model of C-3WD. First, we use a long short-term memory (LSTM) model to predict the workload, and then the error series generated by the LSTM model is represented as the error intervals by the interval set and divided into three regions by the C-3WD model. Finally, the error value of every interval region is corrected by the corresponding prediction model. The experiment results show that the RPEM-3WC model improves the prediction accuracy by up to 15.6%, 11.35%, 10.33%, 16.37%, 12.83%, 12.93%, and 8.68% over the EMA, HWES, ARIMA, SES, LSTM-RNN, NN, and TWD-RCPM models, respectively. [ABSTRACT FROM AUTHOR]

Accurately predicting power load is a pressing concern that requires immediate attention. Short-term load prediction plays a crucial role in ensuring the secure operation and analysis of power systems. However, existing research studies have limited capability in extracting the mutual relationships of multivariate features in multivariate time series data. To address these limitations, we propose a multi-dimensional time series forecasting framework called TS2ARCformer. The TS2ARCformer framework incorporates the TS2Vec layer for contextual encoding and utilizes the Transformer model for prediction. This combination effectively captures the multi-dimensional features of the data. Additionally, TS2ARCformer introduces a Cross-Dimensional-Self-Attention module, which leverages interactions across channels and temporal dimensions to enhance the extraction of multivariate features. Furthermore, TS2ARCformer leverage a traditional autoregressive component to overcome the issue of deep learning models being insensitive to input scale. This also enhances the model's ability to extract linear features. Experimental results on two publicly available power load datasets demonstrate significant improvements in prediction accuracy compared to baseline models, with reductions of 43.2% and 37.8% in the aspect of mean absolute percentage error (MAPE) for dataset area1 and area2, respectively. These findings have important implications for the accurate prediction of power load and the optimization of power system operation and analysis. [ABSTRACT FROM AUTHOR]

In the paper, an electric vehicle (EV) charging load prediction approach considering vehicle types is developed. The GM (1,1) model is initially introduced to estimate the ownership of electric vehicles (EVs), and the model is validated by examining the difference between the forecast results and the actual situation. The Monte Carlo algorithm is also used to build the probability distribution model of travel laws and charging characteristics to predict the load demand when the scale of EVs is connected to the grid. For various types of EVs, the charging behavior is analyzed by considering charging start moment, charging duration, the initial state of charge (SOC), daily mileage, and charging mode. The charging load curves for each type of EV are superimposed to obtain the total load demand curve. Then it is compared with the original load curve in a certain region to analyze the impact of EV charging load on the original load curve of the grid, and to provide theoretical guidance for grid planning and optimal dispatching.

The difficulty in balancing energy supply and demand is increasing due to the growth of diversified and flexible building energy resources, particularly the rapid development of intermittent renewable energy being added into the power grid. The accuracy of building energy consumption prediction is of top priority for the electricity market management to ensure grid safety and reduce financial risks. The accuracy and speed of load prediction are fundamental prerequisites for different objectives such as long-term planning and short-term optimization of energy systems in buildings and the power grid. The past few decades have seen the impressive development of time series load forecasting models focusing on different domains and objectives. This paper presents an in-depth review and discussion of building energy prediction models. Three widely used prediction approaches, namely, building physical energy models (i.e., white box), data-driven models (i.e., black box), and hybrid models (i.e., grey box), were classified and introduced. The principles, advantages, limitations, and practical applications of each model were investigated. Based on this review, the research priorities and future directions in the domain of building energy prediction are highlighted. The conclusions drawn in this review could guide the future development of building energy prediction, and therefore facilitate the energy management and efficiency of buildings.

Most enterprise customers now choose to divide a large monolithic service into large numbers of loosely-coupled, specialized microservices, which can be developed and deployed separately. Docker, as a light-weight virtualization technology, has been widely adopted to support diverse microservices. At the moment, Kubernetes is a portable, extensible, and open-source orchestration platform for managing these containerized microservice applications. To adapt to frequently changing user requests, it offers an automated scaling method, Horizontal Pod Autoscaler (HPA), that can scale itself based on the system's current workload. The native reactive auto-scaling method, however, is unable to foresee the system workload scenario in the future to complete proactive scaling, leading to QoS (quality of service) violations, long tail latency, and insufficient server resource usage. In this paper, we suggest a new proactive scaling scheme based on deep learning approaches to make up for HPA's inadequacies as the default autoscaler in Kubernetes. After meticulous experimental evaluation and comparative analysis, we use the Gated Recurrent Unit (GRU) model with higher prediction accuracy and efficiency as the prediction model, supplemented by a stability window mechanism to improve the accuracy and stability of the prediction model. Finally, with the third-party custom autoscaling framework, Custom Pod Autoscaler (CPA), we packaged our custom autoscaling algorithm into a framework and deployed the framework into the real Kubernetes cluster. Comprehensive experiment results prove the feasibility of our autoscaling scheme, which significantly outperforms the existing Horizontal Pod Autoscaler (HPA) approach. [ABSTRACT FROM AUTHOR]

Accurate multi-energy load prediction plays a very crucial role in integrated energy system management. To address the load characteristics of strong relational coupling, volatility and uncertainty in user-side integrated energy systems, this paper proposes a multi-task learning based short-term multi-energy load prediction method. Load participation factor is proposed to portray the proportion of different loads in the total load demand, and multi-task learning method is introduced to deeply explore the coupling relationships among them. Then, the proposed method is validated on a real-world dataset. The results show that the method has higher prediction accuracy than the existing methods, and the prediction accuracy is improved by at least 1.3%. [ABSTRACT FROM AUTHOR]

The construction of smart grids has greatly changed the power grid pattern and power supply structure. For the power system, reasonable power planning and demand response is necessary to ensure the stable operation of a society. Accurate load prediction is the basis for realizing demand response for the power system. This paper proposes a Pre-Attention-CNN-GRU model (PreAttCG) which combines a convolutional neural network (CNN) and gate recurrent unit (GRU) and applies the attention mechanism in front of the whole model. The PreAttCG model accepts historical load data and more than nine other factors (including temperature, wind speed, humidity, etc.) as input. The attention layer and CNN layer effectively extract the features and weights of each factor. Load forecasting is then performed by the prediction layer, which consists of a stacked GRU. The model is verified by industrial load data from a German dataset and a Chinese dataset from the real world. The results show that the PreAttCG model has better performance (3~5% improvement in MAPE) than both LSTM with only load input and LSTM with all factors. Additionally, the experiments also show that the attention mechanism can effectively extract the weights of relevant factors affecting the load data. [ABSTRACT FROM AUTHOR]

Electricity load prediction is an essential tool for power system planning, operation and management. The critical information it provides can be used by energy providers to maximise power system operation efficiency and minimise system operation costs. Long Short-Term Memory (LSTM) and Support Vector Machine (SVM) are two suitable methods that have been successfully used for analysing time series problems. In this paper, the two algorithms are explored further for load prediction; two load prediction algorithms are developed and verified by using the half-hourly load data from the University of Warwick campus energy centre with four different prediction time horizons. The novelty lies in comparing and analysing the prediction accuracy of two intelligent algorithms with multiple time scales and in exploring better scenarios for their prediction applications. High-resolution load forecasting over a long range of time is also conducted in this paper. The MAPE values for the LSTM are 2.501%, 3.577%, 25.073% and 69.947% for four prediction time horizons delineated. For the SVM, the MAPE values are 2.531%, 5.039%, 7.819% and 10.841%, respectively. It is found that both methods are suitable for shorter time horizon predictions. The results show that LSTM is more capable of ultra-short and short-term forecasting, while SVM has a higher prediction accuracy in medium-term and long-term forecasts. Further investigation is performed via blind tests and the test results are consistent. [ABSTRACT FROM AUTHOR]

With the rapid development of integration in blockchain and IoT, virtual machine consolidation (VMC) has become a heated topic because it can effectively improve the energy efficiency and service quality of cloud computing in the blockchain. The current VMC algorithm is not effective enough because it does not regard the load of the virtual machine (VM) as an analyzed time series. Therefore, we proposed a VMC algorithm based on load forecast to improve efficiency. First, we proposed a migration VM selection strategy based on load increment prediction called LIP. Combined with the current load and load increment, this strategy can effectively improve the accuracy of selecting VM from the overloaded physical machines (PMs). Then, we proposed a VM migration point selection strategy based on the load sequence prediction called SIR. We merged VMs with complementary load series into the same PM, effectively improving the stability of the PM load, thereby reducing the service level agreement violation (SLAV) and the number of VM migrations due to the resource competition of the PM. Finally, we proposed a better virtual machine consolidation (VMC) algorithm based on the load prediction of LIP and SIR. The experimental results show that our VMC algorithm can effectively improve energy efficiency. [ABSTRACT FROM AUTHOR]

An accurate and fast building load prediction model is critically important for guiding building energy system design, optimizing operational parameters, and balancing a power grid between energy supply and demand. A physics-based simulation tool is traditionally used to provide the building load demand; however, it is constrained by its complex model development process and requirement for engineering judgments. Machine learning algorithms (i.e., data-driven models) based on big data can bridge this gap. In this study, we used the massive energy data generated by a physics-based tool (EnergyPlus) to develop three data-driven models (i.e., LightGBM, random forest (RF), and long-short term memory (LSTM)) and compared their prediction performances. The physics-based models were developed using office prototype building models as baselines, and ranges were provided for selected key input parameters. Three different input feature dimensions (i.e., six-, nine-, and fifteen-input feature selections) were investigated, aiming to meet different demands for practical applications. We found that LightGBM significantly outperforms the RF and LSTM algorithms, not only with respect to prediction accuracy but also in regard to computation cost. The best prediction results show that the coefficient of variation of the root mean squared error (CVRMSE), squared correction coefficient (R2), and computation time are 5.25%, 0.9959, and 7.0 s for LightGBM, respectively, evidently better than the values for the algorithms based on RF (18.54%, 0.9482, and 44.6 s) and LSTM (22.06%, 0.9267, and 758.8 s). The findings demonstrate that a data-driven model is able to avoid the process of establishing a complicated physics-based model for predicting a building's thermal load, with similar accuracy to that of a physics-based simulation tool. [ABSTRACT FROM AUTHOR]

In the district heating systems, the historical operation data of the buildings in those areas would be partially or entirely missing. The traditional data-driven model is hard to predict the ground truth results because the historical data is not available for model training. However, utilizing the physics-based methods for load calculation takes a long time to process and encounters low accuracy issues. This paper investigates several hybrid models that integrate the data-driven model and the physics-based models with different fusion methods. The physics-based models calculate envelope load and infiltration load, based on Fourier's law and the grand canonical ensemble theory, respectively. After undergoing load processing, features fusion, and residual connection, the best advanced hybrid models generate 21.35%, 16.35%, and 12.73% better prediction results compared with the data-driven model. Moreover, the advanced hybride models also perform strong transferability across all the data quantity groups. In terms of practical application, the advanced hybrid models could be deployed with effective generalization in limited data scenarios and robust transfer capabilities. The selected best model constructed by hybrid modeling displays the highest performance and saves the total training costs with strong transferability. • The hybrid models integrating physics-based models and data are established. • The training costs, metrics, and transferability are evaluated for hybrid models. • The hybrid models' performances in different data quantity groups are discussed. [ABSTRACT FROM AUTHOR]

Building load prediction plays an important role in building energy savings and mechanical and electrical system optimization control. The dynamic energy consumption can be accurately calculated using the traditional physical energy simulation method that entails a complex setup and verification process due to the numerous input parameters required. It is also difficult to change the physical model once it has been determined. The data-mining method is fast in calculation and simple to use, but its prediction accuracy is limited by historical data quality. It is difficult to predict the load of new buildings without historical data. To solve these problems, this study proposes a hybrid building load prediction method for office buildings. The proposed method uses EnergyPlus to generate a building load database that includes than 25.14 million data cases, covering 35 types of building geometry and 2870 building examples. Based on the above database, the LightGBM algorithm was selected to extract feature variables that affect the load and build a load prediction model. The training results show that there are 24 key feature variables for office building load prediction. The hourly MAPE of the cooling load prediction model is 6.95 % and RMSE is 4.31 W/m2, and the hourly MAPE of the heating load prediction model is 7.09 % and RMSE is 11.64 W/m2 compared with EnergyPlus model. Two actual office buildings are selected as case studies to validate the model prediction accuracy. Results show that comparing predicted results with measured data, the hourly cooling load of the MAPE is 12.42 %. Coparing predicted results with actual heating load, daily MAPE is 7.97 %. • A case-rich office load simulation database over 25.14 million pieces is generated and extensible. • A high-precision load forecasting method based on LightGBM for office buildings is proposed. • Eliminate the geometric constraints imposed by simulation model and the reliance on historical data of data-driven model. • The method can be applied to other types of buildings by modifying the parameters. [ABSTRACT FROM AUTHOR]