Search

Short-term precipitation forecasting plays a critical role in the areas of meteorology and hydrology. To address the limitations of traditional forecasting methods in dealing with complex meteorological phenomena and the problem of cumulative errors in image sequence prediction by recurrent neural networks, a short-term precipitation forecasting method called multi-scale attention encoding-dynamic decoding network (MAEDDN) has been presented. It predicts future precipitation by learning the spatiotemporal features of the input data. Within the encoding process, convolutional blocks with spatial and channel attention are utilized for encoding, and a multi-scale fusion module is employed to address the challenge of capturing both small-scale and large-scale information in precipitation distribution simultaneously. In short-term precipitation forecasting, the method effectively addresses weather systems’ generation and dissipation. During the decoding process, a dynamic decoding network is put forward to flexibly choose the decoding process according to the learned intensity distribution and change trends derived from past input data. Experiments are carried out by utilizing the precipitation data from the open-source SEVIR dataset, and comparisons are made with the currently best methods reported. The experimental results reveal that: (1) The CSI evaluation metrics for MAEDDN in high-intensity precipitation areas have shown significant improvement, with CSI-160 at 0.3645, CSI-181 at 0.3077, and CSI-219 at 0.2330. These results indicate that the predictive capability of MAEDDN in high-intensity precipitation areas has been significantly enhanced. (2) MAEDDN outperforms other models in terms of the resolution of predicted image sequences. The constructed multi-scale attention encoding captures the complex relationships in meteorological data more effectively, while the dynamic decoding adapts the decoding process based on different scenarios, resulting in more accurate prediction outcomes. [ABSTRACT FROM AUTHOR]

Accurately predicting solar power to ensure the economical operation of microgrids and smart grids is a key challenge for integrating the large scale photovoltaic (PV) generation into conventional power systems. This paper proposes an accurate short-term solar power forecasting method using a hybrid machine learning algorithm, with the system trained using the pre-trained extreme learning machine (P-ELM) algorithm. The proposed method utilizes temperature, irradiance, and solar power output at instant i as input parameters, while the output parameters are temperature, irradiance, and solar power output at instant i+1, enabling next-day solar power output forecasting. The performance of the P-ELM algorithm is evaluated using mean absolute error (MAE) and root mean square error (RMSE), and it is compared with the extreme learning machine (ELM) algorithm. The results indicate that the P-ELM algorithm achieves higher accuracy in short-term prediction, demonstrating its suitability for ensuring accuracy and reliability in real-time solar power forecasting. [ABSTRACT FROM AUTHOR]

Photovoltaic (PV) power is greatly uncertain due to the random meteorological parameters. Therefore, accurate PV power forecasting results are significant for the dispatching of power and improving of system stability. This paper proposes a hybrid forecasting model for one-day-ahead PV power forecasting under different cloud amount conditions. The proposed model consists of an improved artificial neural network (ANN) algorithm and a PV power conversion model. First, the ANN model is designed to forecast the plane of array (POA) irradiance and ambient temperature. Backpropagation, gradient descent, and L2 regularization methods are applied in the structure of the ANN model to achieve the best weights, improve the prediction accuracy, and alleviate the effect of overfitting. Second, the PV power conversion model employs the forecasted results of POA irradiance and ambient temperature to determine the PV power produced by a PV module. In addition to the basic temperature factor, environmental efficiency and a reflection efficiency are incorporated into the conversion model to account for real PV module losses. The performance of the proposed model is validated with real weather and PV power data from Alice Springs and Climate Data Store. Results indicate that the model improves the forecast accuracy compared to four benchmark models. Specifically, it reduces root mean square error (RMSE) and normalized RMSE (nRMSE) by up to 25% under cloudy conditions and offers a 3% shorter training time compared to extreme gradient boosting. [ABSTRACT FROM AUTHOR]

To solve the problem of photovoltaic power prediction in areas with large climate changes, this article proposes a hybrid Long Short-Term Memory method to improve the prediction accuracy and noise resistance. It combines the improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN) and kernel principal component analysis (KPCA) algorithm. The ICEEMDAN algorithm reduces the instability of the environmental factor sequence. The KPCA algorithm reduces the input dimensions of the model. LSTM performs dynamic time modeling of the multivariate feature sequences to predict the output PV power. The adaptability of the ICEEMDAN-KPCA-LSTM model is assessed with datasets from a PV plant in west China and evaluated by root mean squared error (RMSE), mean absolute percentage error (MAPE), and R-squared metrics. Using 70% of the datasets for output PV power estimation, the results show a good performance, with an RMSE of 4.3715, MAPE of 8.9264%, and R-squared value of 89.973%. By comparing with other prediction models, the ICEEMDAN-KPCA-LSTM photovoltaic output power model outperforms other models. [ABSTRACT FROM AUTHOR]

High-quality short-term forecasts of wind farm generation are crucial for the dynamically developing renewable energy generation sector. This article addresses the selection of appropriate gradient-boosted decision tree models (GBDT) for forecasting wind farm energy generation with a 10-min time horizon. In most forecasting studies, authors utilize a single gradient-boosted decision tree model and compare its performance with other machine learning (ML) techniques and sometimes with a naive baseline model. This paper proposes a comprehensive comparison of all gradient-boosted decision tree models (GBDTs, eXtreme Gradient Boosting (XGBoost), Light Gradient-Boosting Machine (LightGBM), and Categorical Boosting (CatBoost)) used for forecasting. The objective is to evaluate each model in terms of forecasting accuracy for wind farm energy generation (forecasting error) and computational time during model training. Computational time is a critical factor due to the necessity of testing numerous models with varying hyperparameters to identify the optimal settings that minimize forecasting error. Forecast quality using default hyperparameters is used here as a reference. The research also seeks to determine the most effective sets of input variables for the predictive models. The article concludes with findings and recommendations regarding the preferred GBDT models. Among the four tested models, the oldest GBDT model demonstrated a significantly longer training time, which should be considered a major drawback of this implementation of gradient-boosted decision trees. In terms of model quality testing, the lowest nRMSE error was achieved by the oldest model—GBDT in its tuned version (with the best hyperparameter values obtained from exploring 40,000 combinations). [ABSTRACT FROM AUTHOR]

This paper compares four time series forecasting algorithms—ARIMA, SARIMA, LSTM, and SVM—suitable for short-term load forecasting using Advanced Metering Infrastructure (AMI) data. The primary focus is on evaluating the applicability and performance of these forecasting models in predicting electricity consumption patterns, which is a critical component for implementing effective demand response (DR) strategies. The study provides a comprehensive analysis of the predictive accuracy, computational efficiency, and scalability of each algorithm using a dataset of real-time electricity consumption collected from AMI systems over a designated period. Through extensive experiments, we demonstrate that each algorithm has distinct strengths and weaknesses depending on the characteristics of the dataset. Specifically, SVM exhibited superior performance in handling nonlinear patterns and high volatility, while SARIMA effectively captured seasonal trends. LSTM showed potential in modeling complex temporal dependencies but was sensitive to hyperparameter settings and required a substantial amount of training data. This research offers practical guidelines for selecting the optimal forecasting model based on data characteristics and application needs, contributing to the development of more efficient and dynamic energy management strategies. The findings highlight the importance of integrating advanced forecasting techniques into smart grid systems to enhance the reliability and responsiveness of DR programs. This study lays a solid foundation for future research on integrating these forecasting models into real-world AMI applications to support effective demand response and grid stability. [ABSTRACT FROM AUTHOR]

Accurate and efficient forecasting of energy consumption is a crucial prerequisite for effective energy planning and policymaking. The BP neural network has been widely used in forecasting, machine learning, and various other fields due to its nonlinear fitting ability. In order to improve the prediction accuracy of the BP neural network, this paper introduces the concept of forecast lead time and establishes a mathematical model accordingly. Prior to training the neural network, the input layer data are preprocessed based on the forecast lead time model. The training and forecasting results of the BP neural network when and when not considering forecast lead time are compared and verified. The findings demonstrate that the forecast lead time model can significantly improve the prediction speed and accuracy, proving to be highly applicable for short-term energy consumption forecasting. [ABSTRACT FROM AUTHOR]

Abstract This research aims to explore more efficient machine learning (ML) algorithms with better performance for short-term forecasting. Up-to-date literature shows a lack of research on selecting practical ML algorithms for short-term forecasting in real-time industrial applications. This research uses a quantitative and qualitative mixed method combining two rounds of literature reviews, a case study, and a comparative analysis. Ten widely used ML algorithms are selected to conduct a comparative study of gas warning systems in a case study mine. We propose a new assessment visualization tool: a 2D space-based quadrant diagram can be used to visually map prediction error assessment and predictive performance assessment for tested algorithms. Overall, this visualization tool indicates that LR, RF, and SVM are more efficient ML algorithms with overall prediction performance for short-term forecasting. This research indicates ten tested algorithms can be visually mapped onto optimal (LR, RF, and SVM), efficient (ARIMA), suboptimal (BP-SOG, KNN, and Perceptron), and inefficient algorithms (RNN, BP_Resilient, and LSTM). The case study finds results that differ from previous studies regarding the ML efficiency of ARIMA, KNN, LR, LSTM, and SVM. This study finds different views on the prediction performance of a few paired algorithms compared with previous studies, including RF and LR, SVM and RF, KNN and ARIMA, KNN and SVM, RNN and ARIMA, and LSTM and SVM. This study also suggests that ARIMA, KNN, LR, and LSTM should be investigated further with additional prediction error assessments. Overall, no single algorithm can fit all applications. This study raises 20 valuable questions for further research.

This research aims to explore more efficient machine learning (ML) algorithms with better performance for short-term forecasting. Up-to-date literature shows a lack of research on selecting practical ML algorithms for short-term forecasting in real-time industrial applications. This research uses a quantitative and qualitative mixed method combining two rounds of literature reviews, a case study, and a comparative analysis. Ten widely used ML algorithms are selected to conduct a comparative study of gas warning systems in a case study mine. We propose a new assessment visualization tool: a 2D space-based quadrant diagram can be used to visually map prediction error assessment and predictive performance assessment for tested algorithms. Overall, this visualization tool indicates that LR, RF, and SVM are more efficient ML algorithms with overall prediction performance for short-term forecasting. This research indicates ten tested algorithms can be visually mapped onto optimal (LR, RF, and SVM), efficient (ARIMA), suboptimal (BP-SOG, KNN, and Perceptron), and inefficient algorithms (RNN, BP_Resilient, and LSTM). The case study finds results that differ from previous studies regarding the ML efficiency of ARIMA, KNN, LR, LSTM, and SVM. This study finds different views on the prediction performance of a few paired algorithms compared with previous studies, including RF and LR, SVM and RF, KNN and ARIMA, KNN and SVM, RNN and ARIMA, and LSTM and SVM. This study also suggests that ARIMA, KNN, LR, and LSTM should be investigated further with additional prediction error assessments. Overall, no single algorithm can fit all applications. This study raises 20 valuable questions for further research. [ABSTRACT FROM AUTHOR]

High-quality short-term forecasts of electrical energy generation in solar power plants are crucial in the dynamically developing sector of renewable power generation. This article addresses the issue of selecting appropriate (preferred) methods for forecasting energy generation from a solar power plant within a 15 min time horizon. The effectiveness of various machine learning methods was verified. Additionally, the effectiveness of proprietary ensemble and hybrid methods was proposed and examined. The research also aimed to determine the appropriate sets of input variables for the predictive models. To enhance the performance of the predictive models, proprietary additional input variables (feature engineering) were constructed. The significance of individual input variables was examined depending on the predictive model used. This article concludes with findings and recommendations regarding the preferred predictive methods. [ABSTRACT FROM AUTHOR]

Copyright of China Rural Water & Hydropower is the property of China Rural Water & Hydropower 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.)

With the extension of lead time for extrapolation, the attenuation of radar echoes becomes increasingly pronounced, and the predictive performance for intense echoes rapidly deteriorates. These constitute two quintessential characteristics of the current radar extrapolation results inaccuracy. To address the issues mentioned above, a novel methodology called DiffREE (diffusion radar echo extrapolation algorithm) was introduced. This algorithm skillfully fuses the spatial and temporal information from past radar echo frames using a conditional encoding module. It employs a Transformer encoder to automatically extract spatiotemporal features from the echoes, which are then used as conditions to drive the diffusion model in reconstructing the current radar echo frames. Experimental results demonstrate that this method can produce high-precision, high-quality radar forecast frames, achieving significant improvements of 42. 2%, 51. 1%, 49. 8%, and 39. 5% in CSI, ETS, HSS, and POD, respectively, compared to the best-performing baseline algorithm [ABSTRACT FROM AUTHOR]

Copyright of Przegląd Elektrotechniczny is the property of Przeglad Elektrotechniczny 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.)

Photovoltaic (PV) power is greatly uncertain due to the random meteorological parameters. Therefore, accurate PV power forecasting results are significant for the dispatching of power and improving of system stability. This paper proposes a hybrid forecasting model for one-day-ahead PV power forecasting under different cloud amount conditions. The proposed model consists of an improved artificial neural network (ANN) algorithm and a PV power conversion model. First, the ANN model is designed to forecast the plane of array (POA) irradiance and ambient temperature. Backpropagation, gradient descent, and L2 regularization methods are applied in the structure of the ANN model to achieve the best weights, improve the prediction accuracy, and alleviate the effect of overfitting. Second, the PV power conversion model employs the forecasted results of POA irradiance and ambient temperature to determine the PV power produced by a PV module. In addition to the basic temperature factor, environmental efficiency and a reflection efficiency are incorporated into the conversion model to account for real PV module losses. The performance of the proposed model is validated with real weather and PV power data from Alice Springs and Climate Data Store. Results indicate that the model improves the forecast accuracy compared to four benchmark models. Specifically, it reduces root mean square error (RMSE) and normalized RMSE (nRMSE) by up to 25% under cloudy conditions and offers a 3% shorter training time compared to extreme gradient boosting.

The output power of PV systems is influenced by various factors, resulting in strong volatility and randomness, which makes it difficult to forecast. Therefore, this paper proposes an Informer prediction model based on optimised VMD for predicting short-term PV power. Firstly, the temporal coding of the Informer model is improved and, secondly, the original sequence is decomposed into multiple modal components using VMD, and then optimisation of the results of VMD in conjunction with the optimisation strategy of SSA improves the characteristics of the time series data. Finally, the refined data are fed into the Informer framework for modelling and prediction, utilising the self-attention mechanism and multiscale feature fusion of Informer to precisely forecast PV power. The power of PV prediction data from the SSA-VMD-Informer model and four other commonly used models is compared. Experimental results indicate that the SSA-VMD-Informer model performs exceptionally well in short-term PV power prediction, achieving higher accuracy than traditional methods. As an example, the results of predicting the PV power on 24 April in a region of Xinjiang are 1.3882 for RMSE, 0.8310 for MSE, 1.14 for SDE, and 0.9944 for R2. [ABSTRACT FROM AUTHOR]

In recent times, renewable energy sources have gained considerable vitality due to their inexhaustible resources and the detrimental effects of fossil fuels, such as the impact of greenhouse gases on the planet. This article aims to be a supportive tool for the development of research in the field of artificial intelligence (AI), as it presents a solution for predicting photovoltaic energy production. The basis of the AI models is provided from two data sets, one for generated electrical power and another for meteorological data, related to the year 2017, which are freely available on the Energias de Portugal (EDP) Open Project website. The implemented AI models rely on long short-term memory (LSTM) neural networks, providing a forecast value for electrical energy with a 60-min horizon based on meteorological variables. The performance of the models is evaluated using the performance indicators MAE, RMSE, and R2, for which favorable results were obtained, with particular emphasis on forecasts for the spring and summer seasons. [ABSTRACT FROM AUTHOR]

Energy generation and pollutant emissions are two faces of the same coin, as the current energy sources i.e., fossil energy, are still considered to be major sources of greenhouse gases (GHG). Therefore, shifting to cleaner energy sources imposes itself as an inevitable solution to reduce this environmental cost. In this paper, a hybrid system based on the multi-agent approach and Long Short-Term Memory (LSTM) neural networks to forecast the energy production and its carbon dioxide (CO2) emissions and simulate the potential emission reduction in case of switching to renewable energy sources is presented. The proposed system's architecture consists of combining LSTM models with the agent-based technology, where multiple LSTM forecasting models were trained to forecast the production of each type of the studied energies and then estimate the equivalent emitted CO2 and calculate the influence of the renewable energy inputs on the carbon emissions and the fossil fuels consumption. The simulation process consists of two phases: firstly, each forecasting agent uses a specific LSTM model to forecast short-term energy production. Secondly, these agents send the forecasted values to the coordination agent who is responsible for calculating the total CO2 emissions and the benefits of the renewable energy inputs. [ABSTRACT FROM AUTHOR]

Abstract In this paper, NeuralProphet (NP), an explainable hybrid modular framework, enhances the forecasting performance of pandemics by adding two neural network modules; auto-regressor (AR) and lagged-regressor (LR). An advanced deep auto-regressor neural network (Deep-AR-Net) model is employed to implement these two modules. The enhanced NP is optimized via AdamW and Huber loss function to perform multivariate multi-step forecasting contrast to Prophet. The models are validated with COVID-19 time-series datasets. The NP’s efficiency is studied component-wise for a long-term forecast for India and an overall reduction of 60.36% and individually 34.7% by AR-module, 53.4% by LR-module in MASE compared to Prophet. The Deep-AR-Net model reduces the forecasting error of NP for all five countries, on average, by 49.21% and 46.07% for short-and-long-term, respectively. The visualizations confirm that forecasting curves are closer to the actual cases but significantly different from Prophet. Hence, it can develop a real-time decision-making system for highly infectious diseases.

Wind energy has emerged as a viable alternative to fossil fuels due to its clean and cost-effective nature. Pakistan, facing growing energy demands and the imperative to reduce carbon emissions, has invested significantly in wind power to supply electric power in rural and urban communities, particularly in the Thatta district of Sindh Province of Pakistan. However, the sustainability of wind energy generation is contingent upon consistent and sufficient wind resources. This study examines the wind potential of Thatta district from 2004 to 2023 to assess its suitability for large-scale wind power development. To evaluate the wind potential of Thatta district, seasonal wind speed and direction data were collected and analyzed. Wind shear at different heights was determined using the power law, and wind potential maps were generated using GIS interpolation techniques. Betz’s law was employed to assess wind turbine power density. Box–Jenkins ARIMA and SARIMA models were applied to predict future wind patterns. This study revealed that Thatta district experienced sufficient wind speeds during the study period, with averages of 9.7 m/s, 7.6 m/s, 7.4 m/s, and 4.8 m/s for summer, autumn, spring, and winter, respectively. However, a concerning trend of decreasing wind speeds has been observed since 2009. The most significant reductions occurred in summer, coinciding with Pakistan’s peak electricity demand. While Thatta district has historically demonstrated potential for wind energy, the declining wind speeds pose a challenge to the sustainability of wind power projects. Further research is necessary to identify the causes of this trend and to explore mitigation strategies.

In this paper, NeuralProphet (NP), an explainable hybrid modular framework, enhances the forecasting performance of pandemics by adding two neural network modules; auto-regressor (AR) and lagged-regressor (LR). An advanced deep auto-regressor neural network (Deep-AR-Net) model is employed to implement these two modules. The enhanced NP is optimized via AdamW and Huber loss function to perform multivariate multi-step forecasting contrast to Prophet. The models are validated with COVID-19 time-series datasets. The NP’s efficiency is studied component-wise for a long-term forecast for India and an overall reduction of 60.36% and individually 34.7% by AR-module, 53.4% by LR-module in MASE compared to Prophet. The Deep-AR-Net model reduces the forecasting error of NP for all five countries, on average, by 49.21% and 46.07% for short-and-long-term, respectively. The visualizations confirm that forecasting curves are closer to the actual cases but significantly different from Prophet. Hence, it can develop a real-time decision-making system for highly infectious diseases. [ABSTRACT FROM AUTHOR]

Timely short-term spatial air quality forecasting is essential for monitoring and prevention in urban agglomerations, providing a new perspective on joint air pollution prevention. However, a single model on air pollution forecasting or spatial correlation analysis is insufficient to meet the strong demand. Thus, this paper proposed a complex real-time monitoring and decision-making assistance system, using a hybrid forecasting module and social network analysis. Firstly, before an accurate forecasting module was constructed, text sentiment analysis and a strategy based on multiple feature selection methods and result fusion were introduced to data preprocessing. Subsequently, CNN-D-LSTM was proposed to improve the feature capture ability to make forecasting more accurate. Then, social network analysis was utilized to explore the spatial transporting characteristics, which could provide solutions to joint prevention and control in urban agglomerations. For experiment simulation, two comparative experiments were constructed for individual models and city cluster forecasting, in which the mean absolute error decreases to 7.8692 and the Pearson correlation coefficient is 0.9816. For overall spatial cluster forecasting, related experiments demonstrated that with appropriate cluster division, the Pearson correlation coefficient could be improved to nearly 0.99. [ABSTRACT FROM AUTHOR]

As a result of climate change, the difficulty in the prediction of short-term rainfall amounts has become a necessary area of research. The existing numerical weather prediction models have limitations in precipitation forecasting especially due to high computation requirements and are prone to errors. Precipitation amount prediction is challenging as it requires knowledge on a variety of environmental phenomena, such as temperature, humidity, wind direction, and more over a long period of time. In this study, we first of all present our Lake Victoria Basin weather dataset and then use it to conduct a rigorous analysis of machine learning algorithms to do short-term rainfall prediction. The rigorous analysis includes algorithm optimizations to improve prediction performance. In particular, we validate our weather dataset using various machine learning regression models which include Random Forest regression, Support Vector regression, Neural Network regression, Least Absolute Shrinkage and Selection Operator regression, Gradient boosting regression, and Extreme Gradient boosting regression. The performance of the models was evaluated using Mean Absolute Error (MAE), and Root Mean Square Error (RMSE). The findings demonstrate that, in comparison to other algorithms, Extreme Gradient Boost regression has the lowest MAE values of 0.006, 0.018, 0.005 for Lake Victoria basin weather data in Uganda, Kenya, and Tanzania respectively.

With the rapid development of the civil aviation industry, flight delays caused by convective weather are becoming increasingly severe. In terminal airspace with complex traffic environments, these delays can propagate to subsequent arrivals and departures, thereby affecting the operational efficiency of the terminal airspace and potentially disrupting the entire air traffic system. Accurate short-term forecasting of convective weather can help reduce flight delays, improve airspace utilization, and decrease the workload of air traffic controllers. This study, grounded in supervised deep learning, addresses civil aviation operational requirements by developing a short-term forecasting model for convective weather based on a Convolutional Neural Network-Transformer (CNN-Transformer). The model leverages eight types of weather products from civil aviation weather radar and ERA5, including basic reflectance, echo top, vertically integrated liquid, relative humidity, temperature, U-component of wind, V-component of wind, and vertical velocity. The training was conducted using 241 datasets, totaling 23,881 samples. To evaluate the model’s validity, ablation studies were performed on each parameter, and its performance was compared with the Centroid Method, Optical Flow, CNN-CNN, and CNN-Long Short-Term Memory (CNN-LSTM). According to six evaluation indicators, traditional radar echo extrapolation showed better forecast accuracy within 1 hour, while the CNN-Transformer-based short-term forecasting model for convective weather demonstrated superior performance for 2-6 hour forecasts.

The significant increase in the installed capacity of power plants with renewable energy sources and the imbalance of the financial system of the wholesale electricity market of Ukraine prompted the Ministry of Energy to develop an alter-native support mechanism for RES producers. The introduction of a feed-in tariff (FIP), which compensates for the difference between the actual sale price of electricity and the "green" tariff, can help producers receive more money immediately after the electricity is released. However, studies have shown that exiting a balancing group without form-ing a new one can lead to increased costs associated with forecasting error. It is important for manufacturers to form independent balancing groups to compensate for negative consequences. The findings of the article show that there is no single optimal balancing group for all manufacturers, but some groups are often repeated. Switching to a separate balancing group can have a significant economic effect for the manufacturer, reducing the cost of forecasting error compared to being solely responsible for the imbalance. However, the balancing group determined by the method of retrospective calculation of the cost of the forecast error is not stable in the long term. Ref. 8, fig. 2, tab. 3.

The article presents the results of the study of an artificial neural network model of the LSTM type for short-term fore-casting of the values of positive and negative imbalances of electric energy in the IPS of Ukraine. The analysis of fore-casting results obtained with the help of hyperparameter optimization models and different window lengths and com-bining them into an ensemble of models was performed. Conducted research based on actual data of the balancing market of electric energy of Ukraine showed the effectiveness of using the specified models to solve the given problem. Ref. 10, fig. 3, tab. 3.

The use of big data in deep neural networks has recently surpassed traditional machine learning techniques in many application areas. The main reasons for the use of deep neural networks are the increase in computational power made possible by graphics processing units and tensor processing units, and the new algorithms created by recurrent neural networks and CNNs. In addition to traditional machine learning methods, deep neural networks have applications in anticipating electricity load. Using a real dataset for one-step forecasting, this article compares three deep learning algorithms for short-term power load forecasting: LSTM, GRUs, and CNN. The statistics come from the Turkish city of Zonguldak and include hourly electricity usage loads and temperatures over a period of three years, commencing in 2019 and ending in 2021. The mean absolute percentage error is used to compare the performances of the techniques. Forecasts are made for twelve representative months from each season. The main reason for the significant deviations in the forecasts for January, May, September, and December is the presence of religious and national holidays in these months. This was solved by adding the information obtained from religious and national holidays to the modeling. This is not to say that CNNs are not good at capturing long-term dependencies and modeling sequential data. In all experiments, LSTM, GRUs and encoder-decoder LSTM outperformed simple CNN designs. In the future, these new architectural methods can be applied to long- or short-term electric charge predictions and their results can be compared to LSTM, GRUs and their variations.

Wind power forecasting is pivotal in promoting a stable and sustainable grid operation by estimating future power outputs from past meteorological and turbine data. The inherent unpredictability in wind patterns poses substantial challenges in synchronizing supply with demand, with inaccuracies potentially destabilizing the grid and potentially causing energy shortages or excesses. This study develops a data-driven approach to forecast wind power from 30 min to 12 h ahead using historical wind power data collected by the Supervisory Control and Data Acquisition (SCADA) system from one wind turbine, the Enercon/E92 2350 kW model, installed at Casa Nova, Bahia, Brazil. Those data were measured from January 2020 to April 2021. Time orientation was embedded using sine/cosine or cyclic encoding, deriving 16 normalized features that encapsulate crucial daily and seasonal trends. The research explores two distinct strategies: error prediction and error correction, both employing a sequential model where initial forecasts via k-Nearest Neighbors (KNN) are rectified by the Extra Trees Regressor. Their primary divergence is the second model’s target variable. Evaluations revealed both strategies outperforming the standalone KNN, with error correction excelling in short-term predictions and error prediction showing potential for extended forecasts. This exploration underscores the imperative importance of methodology selection in wind power forecasting.

Renewable energy sources (RES) are seen as a means of the fuel and energy complex carbon footprint reduction but the stochastic nature of generation complicates RES integration with electric power systems. Therefore, it is necessary to develop and improve methods for forecasting of the power plants generation using the energy of the sun, wind and water flows. One of the ways to improve the accuracy of forecast models is a deep analysis of meteorological conditions as the main factor affecting the power generation. In this paper, a method for adapting of forecast models to the meteorological conditions of photovoltaic stations operation based on machine learning algorithms was proposed and studied. In this case, unsupervised learning is first performed using the k-means method to form clusters. For this, it is also proposed to use studied the feature space dimensionality reduction algorithm to visualize and estimate the clustering accuracy. Then, for each cluster, its own machine learning model was trained for generation forecasting and the k-nearest neighbours algorithm was built to attribute the current conditions at the model operation stage to one of the formed clusters. The study was conducted on hourly meteorological data for the period from 1985 to 2021. A feature of the approach is the clustering of weather conditions on hourly rather than daily intervals. As a result, the mean absolute percentage error of forecasting is reduced significantly, depending on the prediction model used. For the best case, the error in forecasting of a photovoltaic plant generation an hour ahead was 9 %.

This article is devoted to the problem of forecasting electricity generation by photovoltaic power plants based on meteorological data from open sources using machine learning methods. To solve the problem proposed in the article, an overview of existing meteorological data sources and possible methods of processing them is given, as well as a simplified LSTM algorithm based on the architecture of the machine learning methodology for predicting solar energy generation a day earlier. [ABSTRACT FROM AUTHOR]

The paper focuses on the development of models for forecasting the electricity generation of industrial solar power plants using artificial neural networks and numerical weather prediction. The relevance of the research is driven by the need to reduce costs related to imbalances in electricity generation from renewable sources, which can sometimes reach 50% of the released electricity. Additionally, the imbalances of such producers are increasing in Ukraine's power system. Currently, the general imbalances of renewable energy producers in Ukraine have led to a 45% reduction in green electricity production, especially due to the damage or destruction of 75% of wind power plants and 15% of solar power plants in southern and southeastern regions as a result of hostilities. Increasing the accuracy and stability of electricity generation forecasts for such producers could significantly reduce costs associated with imbalances.. Various aggregation methods have been developed for 15-minute values of green energy generation to enhance forecasting accuracy for 1, 2, and 24-hour intervals. The study investigated the potential benefits of using numerical weather prediction (NWP) forecast values to enhance forecasting accuracy. The study revealed the significance of different factors for forecasting at each bias interval. The study employed two modern recurrent neural network models, LSTM and GRU, with varying time sequences. References 14, figures 5, table 2. [ABSTRACT FROM AUTHOR]

The well-documented decrease in the annual minimum Arctic sea ice extent over the past few decades is an alarming indicator of current climate change. However, much less is known about the thickness of the Arctic sea ice. Developing accurate forecasting models is critical to better predict its changes and monitor the impacts of global warming on the total Arctic sea ice volume (SIV). Significant improvements in forecasting performance are possible with the advances in signal processing and deep learning. Accordingly, here, we set out to utilize the recent advances in machine learning to develop non-physics-based techniques for forecasting the sea ice volume with low computational costs. In particular, this paper aims to provide a step-wise decision process required to develop a more accurate forecasting model over short- and mid-term horizons. This work integrates variational mode decomposition (VMD) and bidirectional long short-term memory (BiLSTM) for multi-input multi-output pan-Arctic SIV forecasting. Different experiments are conducted to identify the impact of several aspects, including multivariate inputs, signal decomposition, and deep learning, on forecasting performance. The empirical results indicate that (i) the proposed hybrid model is consistently effective in time-series processing and forecasting, with average improvements of up to 60% compared with the case of no decomposition and over 40% compared with other deep learning models in both forecasting horizons and seasons; (ii) the optimization of the VMD level is essential for optimal performance; and (iii) the use of the proposed technique with a divide-and-conquer strategy demonstrates superior forecasting performance. [ABSTRACT FROM AUTHOR]

Wind power forecasting is pivotal in promoting a stable and sustainable grid operation by estimating future power outputs from past meteorological and turbine data. The inherent unpredictability in wind patterns poses substantial challenges in synchronizing supply with demand, with inaccuracies potentially destabilizing the grid and potentially causing energy shortages or excesses. This study develops a data-driven approach to forecast wind power from 30 min to 12 h ahead using historical wind power data collected by the Supervisory Control and Data Acquisition (SCADA) system from one wind turbine, the Enercon/E92 2350 kW model, installed at Casa Nova, Bahia, Brazil. Those data were measured from January 2020 to April 2021. Time orientation was embedded using sine/cosine or cyclic encoding, deriving 16 normalized features that encapsulate crucial daily and seasonal trends. The research explores two distinct strategies: error prediction and error correction, both employing a sequential model where initial forecasts via k-Nearest Neighbors (KNN) are rectified by the Extra Trees Regressor. Their primary divergence is the second model's target variable. Evaluations revealed both strategies outperforming the standalone KNN, with error correction excelling in short-term predictions and error prediction showing potential for extended forecasts. This exploration underscores the imperative importance of methodology selection in wind power forecasting. [ABSTRACT FROM AUTHOR]

The prevalence of substantial inductive/capacitive loads within the industrial sectors induces variations in reactive energy levels. The imbalance between active and reactive energy within the network leads to heightened losses, diminished network efficiency, and an associated escalation in operating costs. Therefore, the forecasting of active and reactive energy in the industrial sector confers notable advantages, including cost reduction, heightened operational efficiency, safeguarding of equipment, enhanced energy consumption management, and more effective assimilation of renewable energy sources. Consequently, a range of specialized forecasting methods for different applications have been developed to address these challenges effectively. This research proposes a stacked ensemble methodology, denoted as Stack-XGBoost, leveraging three distinct machine learning (ML) methods: extra trees regressor (ETR), adaptive boosting (AdaBoost), and random forest regressor (RFR), as foundational models. Moreover, the incorporation of an extreme gradient boosting (XGBoost) algorithm as a meta-learner serves to amalgamate the predictions generated by the base models, enhancing the precision of the active/reactive energy consumption forecasting using real time data for steel industry. To assess the efficacy of the proposed model, diverse performance metrics were employed. The results show that the proposed Stack-XGBoost model outperformed other forecasting methods. Additionally, a sensitivity analysis was conducted to assess the robustness of the proposed method against variations in input parameters. [ABSTRACT FROM AUTHOR]

A time series is a sequence of time-ordered data, and it is generally used to describe how a phenomenon evolves over time. Time series forecasting, estimating future values of time series, allows the implementation of decision-making strategies. Deep learning, the currently leading field of machine learning, applied to time series forecasting can cope with complex and high-dimensional time series that cannot be usually handled by other machine learning techniques. The aim of the work is to provide a review of state-of-the-art deep learning architectures for time series forecasting, underline recent advances and open problems, and also pay attention to benchmark data sets. Moreover, the work presents a clear distinction between deep learning architectures that are suitable for short-term and long-term forecasting. With respect to existing literature, the major advantage of the work consists in describing the most recent architectures for time series forecasting, such as Graph Neural Networks, Deep Gaussian Processes, Generative Adversarial Networks, Diffusion Models, and Transformers. [ABSTRACT FROM AUTHOR]

The article presents the results of the study of models for short-term forecasting of overall electricity imbalances in the IPS of Ukraine. The analysis of forecasting results obtained using different types of autoregressive models and two forecasting models based on artificial neural networks was performed. Conducted research based on actual data of the balancing market of electric energy of Ukraine showed the effectiveness of using artificial neural networks for the specified task. It is shown that the application of the LSTM (Long short-term memory) artificial neural network model achieves the highest forecasting accuracy for both positive and negative electricity imbalances, respectively, compared to forecasting using autoregressive models. Bibl. 11, fig. 3, table.

This article presents original probabilistic forecasting models for day-ahead hourly energy generation forecasts for a photovoltaic (PV) plant, based on a semi-parametric approach using three deterministic forecasts. Input information of these new models consists of data of hourly weather forecasts obtained from a Numerical Weather Prediction model and variables related to the sun position for future instants. The proposed models were satisfactorily applied to the case study of a real-life PV plant in Portugal. Probabilistic benchmark models were also applied to the same case study and their forecasting results compared with the ones of the proposed models. The computer results obtained with these proposed models achieve better point and probabilistic forecasting evaluation indexes values than the ones obtained with the benchmark models.

In order to make an effective economic decision, it is necessary to have an idea of the possible future state of the decision-making object and its environment, which is obtained by means of forecasting. The more accurately forecasts are performed, the less uncertainty in the decision-making situation, and the more effective the decisions made. Therefore, improving the accuracy of economic forecasting is an important scientific task. One of the new directions in economic forecasting is forecasting with the help of vector autoregression models. But the practical application of these models is difficult, because with increasing dimensionality of the autoregression vector the number of model coefficients grows nonlinearly and there are serious computational difficulties in the construction of such models. We propose to use complex-valued vector autoregressions, which are simpler than vector autoregressions of real variables, because they contain half the number of coefficients, the values of which should be estimated by statistical methods. Using the example of the market of world prices for non-ferrous metals, we have formed an eight-dimensional vector of prices for non-ferrous metals, precious and non-precious. Two linear vector autoregressions of real and complex variables, as well as two nonlinear models of vector autoregressions of real and complex variables were constructed on the basis of statistical data of this vector. It is shown that the nonlinear complex-valued vector autoregression is the best model of these four models both from the positions of Bayesian information criterion and from the position of accuracy of short-term economic forecasting, which was verified using the latest statistics. It is recommended to use nonlinear complex-valued autoregressions for short-term economic forecasting of prices. The possibility of using complex-valued vector autoregressions in short-term forecasting of other economic indicators should be clarified through additional research using the methodology outlined in the article. Proving the effectiveness of using complex-valued vector autoregressions in short-term economic forecasting is the basis for further construction of complex-valued vector autoregression models of dimensions greater than 10, which is extremely difficult or impossible for vector autoregressions of real variables.

The use of big data in deep neural networks has recently surpassed traditional machine learning techniques in many application areas. The main reasons for the use of deep neural networks are the increase in computational power made possible by graphics processing units and tensor processing units, and the new algorithms created by recurrent neural networks and CNNs. In addition to traditional machine learning methods, deep neural networks have applications in anticipating electricity load. Using a real dataset for one-step forecasting, this article compares three deep learning algorithms for short-term power load forecasting: LSTM, GRUs, and CNN. The statistics come from the Turkish city of Zonguldak and include hourly electricity usage loads and temperatures over a period of three years, commencing in 2019 and ending in 2021. The mean absolute percentage error is used to compare the performances of the techniques. Forecasts are made for twelve representative months from each season. The main reason for the significant deviations in the forecasts for January, May, September, and December is the presence of religious and national holidays in these months. This was solved by adding the information obtained from religious and national holidays to the modeling. This is not to say that CNNs are not good at capturing long-term dependencies and modeling sequential data. In all experiments, LSTM, GRUs and encoder-decoder LSTM outperformed simple CNN designs. In the future, these new architectural methods can be applied to long- or short-term electric charge predictions and their results can be compared to LSTM, GRUs and their variations. [ABSTRACT FROM AUTHOR]

Electricity load forecasting is an essential operation of the power system. Deep learning is used to improve accurate electricity load forecasting. In this study, combining Long short-term memory and reinforcement learning are proposed to encourage the advantage of a single approach for forecasting. Importance input features, including the mutual feature of electricity load, are used to increase accuracy. First, multi-time series input can handle by Long short-term memory and the addition of features supports to the load feature will make the model better efficient. Because the LSTM model is quite complex, choosing a good set of hyperparameters is difficult. Therefore, the purpose of using reinforcement learning is to optimize hyper-parameters of the Long short-term memory model. The proposed model is the combination of Long-short term memory and reinforcement learning. The proposed model will be applied in two electricity load data sets, the real-life data of Vietnam Electricity and the other public data set. In one day ahead forecasting, the proposed model archives superior performance than the benchmark. [ABSTRACT FROM AUTHOR]

Lubrication is a crucial tool for industrial maintenance. By reducing friction between components, it also reduces wear on equipment. Industrial lubricants have specific properties and can operate efficiently within their operating temperature range. The aim of this study is therefore to forecast the temperature variation of lubrication oil using simple exponential smoothing. In this paper, we discuss one of the most popular forecasting methods in engineering and mechanical fields. The exponential smoothing method has been successful due to the quality of the results it presents. The present contribution aims to model the temperature variation of lubricating oil in a mechanical contact by the simple exponential smoothing method. The analysis and examination of the set of results made it possible to compare the smoothing models for different values of α and to choose the model closest to the actual temperature curve, the choice of the closest model serves to calculate the optimization criteria for the forecast; this method consists in choosing the α minimizing. The forecast results obtained by the SES model with α=0.9 are judged very satisfactory, and they prove the dominance of the SES model. [ABSTRACT FROM AUTHOR]

Solar irradiance forecasting plays an essential role in efficient solar energy systems and managing power demand sustainably. In present work, a new residual ensemble learning approach, which consists of two advanced base models, namely Deep Neural Networks (DNNs) and Recurrent Neural Networks (RNNs), is proposed for solar irradiance forecasting. A model performance depends on data utilized for modeling and the modeling approach employed on the data. This paper focuses on both these aspects of the forecast model by proposing a three module approach. Firstly, a mechanism is proposed for the collection and analysis of multiple-site data surrounding the target location. A hexagon gridding system based algorithm is proposed for selection of multiple sites neighboring the target location. Then, correlation and feature importance scores are utilized as measures for feature selection to choose the most relevant data for forecasting target solar irradiance. In the second module, a residual ensemble learning model is proposed to forecast solar irradiance. The proposed framework is inspired by the hybrid forecast mechanism that considers the linear and non-linear characteristics for modeling. Advanced DNN models of Recurrent Neural Networks are also exploited for developing an accurate and robust model. The last module performs the integration of the deep neural network information and predicts the future values of solar irradiance. For a reliable and comprehensive assessment, the proposed framework is validated with data from four different solar power sites obtained from NASA's POWER repository. The residual ensemble model is trained on past 36 years of data as input for forecasting one day ahead, four days ahead and ten days ahead values of solar irradiance. Performance evaluation is carried out by comparing the prediction results with other models, including benchmark persistence, deep neural networks, and recurrent neural network approaches on performance indexes of MSE and RMSE. The proposed model shows an improvement in forecast performance by approximately 2.5 percent in prediction error. The predictive performance and stability make the proposed residual ensemble learning approach a reliable solar irradiance prediction model. [ABSTRACT FROM AUTHOR]