Search

Integrating renewable energy is a promising solution for buildings to achieve the net-zero-energy goal. Expanding real-time matching between renewable energy generation and building energy demand can help realize more enormous zero-energy potential in practice. However, there are few studies to investigate the real-time energy matching in renewable energy building design. Therefore, in this study, a hybrid ensemble learning framework is proposed for analyzing and predicting zero-energy potential in the real-time matching of photovoltaic direct-driven air conditioner (PVAC) systems. First, the datasets of zero-energy probability (ZEP) are generated under the three main climate regions in China, which are with consideration of the load flexibility of air conditioners and based on six important design variables. Second, a novel ensemble learning method named Extreme Gradient Boosting (XGBoost) is selected to predict ZEP and the Bayesian Optimization (BO) is adopted to identify the optimal hyperparameters and further improve the prediction performance. The statistical analysis shows that ZEP distributions are very different from one region to another one and the PVAC systems in Beijing are the easiest to achieve the zero-energy goal. Among all the variables, PV capacity is the most significant and positively related to ZEP. The prediction results show BO-XGBoost achieves more than 99% accuracy and outperforms other benchmark models in the ZEP prediction of three cities. In a word, this paper reveals BO-XGBoost is the most effective model for ZEP prediction and provides the framework for designers to utilize zero-energy potential analysis and prediction for the first time.

Space cooling in buildings is responsible for massive energy consumption and carbon emissions. Accurate cooling load prediction can facilitate the implementation of energy-efficiency cooling control strategies in practice. In this paper, an improved attention-based deep learning approach is proposed for robust ultra-short-term cooling load prediction. First, a novel time representation learning is introduced to extract the periodicity and non-periodicity of cooling loads efficiently. Then, long short-term memory with an attention mechanism extracts properly the time steps by identifying the relevant hidden states and learns high-level temporal dependency. The approach additionally incorporates extreme gradient boosting through the error reciprocal method, enhancing the elimination of prediction errors and improving robustness. The study takes Guangzhou as an example and generates cooling loads using diverse occupancy schedules of five building types based on the Chinese National Standard and Typical Meteorological Year data. The approach is evaluated on datasets comprising the cooling loads, meteorological data, and contextual information. Through results analysis, the approach outperforms other models in terms of prediction accuracy and robustness across all building types. Additionally, model interpretation is provided regarding feature importance and attention matrixes, which enhances the understanding and transparency of the final prediction from the proposed approach.

Space cooling in buildings is responsible for massive energy consumption and carbon emissions. Accurate cooling load prediction can facilitate the implementation of energy-efficiency cooling control strategies in practice. In this paper, an improved attention-based deep learning approach is proposed for robust ultra-short-term cooling load prediction. First, a novel time representation learning is introduced to extract the periodicity and non-periodicity of cooling loads efficiently. Then, long short-term memory with an attention mechanism extracts properly the time steps by identifying the relevant hidden states and learns high-level temporal dependency. The approach additionally incorporates extreme gradient boosting through the error reciprocal method, enhancing the elimination of prediction errors and improving robustness. The study takes Guangzhou as an example and generates cooling loads using diverse occupancy schedules of five building types based on the Chinese National Standard and Typical Meteorological Year data. The approach is evaluated on datasets comprising the cooling loads, meteorological data, and contextual information. Through results analysis, the approach outperforms other models in terms of prediction accuracy and robustness across all building types. Additionally, model interpretation is provided regarding feature importance and attention matrixes, which enhances the understanding and transparency of the final prediction from the proposed approach.

Reducing the heating and cooling load through energy-efficient building design can help decarbonize the building sector. Heating and cooling load prediction using machine learning (ML) techniques become increasingly important in the rapid assessment of building design variables at the early design stage. However, when applying the ML techniques, it still requires expert knowledge and manually frequent intervention to improve the prediction performance. Hence, this study proposed an automated machine learning (AutoML)-based framework to automatically generate the optimal ML pipelines for heating and cooling load prediction. An experimental dataset of residential buildings was used to evaluate the proposed framework. The proposed framework achieved the best performance with R2 of 0.9965 and RMSE of 0.602 kWh/m2 for heating load prediction, and R2 of 0.9899 and RMSE of 0.973 kWh/m2 for cooling load prediction. The prediction results showed that the proposed framework outperformed the other improved ML models from the representative studies in the last five years. Further, an explainable analysis of the ML models was explored to reveal the relationships between design variables and heating and cooling load. The proposed framework aims at promoting the AutoML-based framework to designers for building energy performance prediction without excessive ML knowledge and manually frequent intervention.

Space cooling in buildings is responsible for massive energy consumption and carbon emissions. Accurate cooling load prediction can facilitate the implementation of energy-efficiency cooling control strategies in practice. In this paper, an improved attention-based deep learning approach is proposed for robust ultra-short-term cooling load prediction. First, a novel time representation learning is introduced to extract the periodicity and non-periodicity of cooling loads efficiently. Then, long short-term memory with an attention mechanism extracts properly the time steps by identifying the relevant hidden states and learns high-level temporal dependency. The approach additionally incorporates extreme gradient boosting through the error reciprocal method, enhancing the elimination of prediction errors and improving robustness. The study takes Guangzhou as an example and generates cooling loads using diverse occupancy schedules of five building types based on the Chinese National Standard and Typical Meteorological Year data. The approach is evaluated on datasets comprising the cooling loads, meteorological data, and contextual information. Through results analysis, the approach outperforms other models in terms of prediction accuracy and robustness across all building types. Additionally, model interpretation is provided regarding feature importance and attention matrixes, which enhances the understanding and transparency of the final prediction from the proposed approach.

Space cooling in buildings is responsible for massive energy consumption and carbon emissions. Accurate cooling load prediction can facilitate the implementation of energy-efficiency cooling control strategies in practice. In this paper, an improved attention-based deep learning approach is proposed for robust ultra-short-term cooling load prediction. First, a novel time representation learning is introduced to extract the periodicity and non-periodicity of cooling loads efficiently. Then, long short-term memory with an attention mechanism extracts properly the time steps by identifying the relevant hidden states and learns high-level temporal dependency. The approach additionally incorporates extreme gradient boosting through the error reciprocal method, enhancing the elimination of prediction errors and improving robustness. The study takes Guangzhou as an example and generates cooling loads using diverse occupancy schedules of five building types based on the Chinese National Standard and Typical Meteorological Year data. The approach is evaluated on datasets comprising the cooling loads, meteorological data, and contextual information. Through results analysis, the approach outperforms other models in terms of prediction accuracy and robustness across all building types. Additionally, model interpretation is provided regarding feature importance and attention matrixes, which enhances the understanding and transparency of the final prediction from the proposed approach.

Space cooling in buildings is responsible for massive energy consumption and carbon emissions. Accurate cooling load prediction can facilitate the implementation of energy-efficiency cooling control strategies in practice. In this paper, an improved attention-based deep learning approach is proposed for robust ultra-short-term cooling load prediction. First, a novel time representation learning is introduced to extract the periodicity and non-periodicity of cooling loads efficiently. Then, long short-term memory with an attention mechanism extracts properly the time steps by identifying the relevant hidden states and learns high-level temporal dependency. The approach additionally incorporates extreme gradient boosting through the error reciprocal method, enhancing the elimination of prediction errors and improving robustness. The study takes Guangzhou as an example and generates cooling loads using diverse occupancy schedules of five building types based on the Chinese National Standard and Typical Meteorological Year data. The approach is evaluated on datasets comprising the cooling loads, meteorological data, and contextual information. Through results analysis, the approach outperforms other models in terms of prediction accuracy and robustness across all building types. Additionally, model interpretation is provided regarding feature importance and attention matrixes, which enhances the understanding and transparency of the final prediction from the proposed approach.

In this paper, we focus on the task query-based video localization, i.e., localizing a query in a long and untrimmed video. The prevailing solutions for this problem can be grouped into two categories: i) Top-down approach: It pre-cuts the video into a set of moment candidates, then it does classification and regression for each candidate; ii) Bottom-up approach: It injects the whole query content into each video frame, then it predicts the probabilities of each frame as a ground truth segment boundary (i.e., start or end). Both two frameworks have respective shortcomings: the top-down models suffer from heavy computations and they are sensitive to the heuristic rules, while the performance of bottom-up models is behind the performance of top-down counterpart thus far. However, we argue that the performance of bottom-up framework is severely underestimated by current unreasonable designs, including both the backbone and head network. To this end, we design a novel bottom-up model: Graph-FPN with Dense Predictions (GDP). For the backbone, GDP firstly generates a frame feature pyramid to capture multi-level semantics, then it utilizes graph convolution to encode the plentiful scene relationships, which incidentally mitigates the semantic gaps in the multi-scale feature pyramid. For the head network, GDP regards all frames falling in the ground truth segment as the foreground, and each foreground frame regresses the unique distances from its location to bi-directional boundaries. Extensive experiments on two challenging query-based video localization tasks (natural language video localization and video relocalization), involving four challenging benchmarks (TACoS, Charades-STA, ActivityNet Captions, and Activity- VRL), have shown that GDP surpasses the state-of-the-art top-down models. © AAAI 2020 - 34th AAAI Conference on Artificial Intelligence. All Rights Reserved.

In this paper, we focus on natural language video localization: localizing (i.e., grounding) a natural language description in a long and untrimmed video sequence. All currently published models for addressing this problem can be categorized into two types: (i) top-down approach: it does classification and regression for a set of pre-cut video segment candidates; (ii) bottom-up approach: it directly predicts probabilities for each video frame as the temporal boundaries (i.e., start and end time point). However, both two approaches suffer several limitations: the former is computation-intensive for densely placed candidates, while the latter has trailed the performance of the top-down counterpart thus far. To this end, we propose a novel dense bottom-up framework: DEnse Bottom-Up Grounding (DEBUG). DEBUG regards all frames falling in the ground truth segment as foreground, and each foreground frame regresses the unique distances from its location to bi-directional ground truth boundaries. Extensive experiments on three challenging benchmarks (TACoS, Charades-STA, and ActivityNet Captions) show that DEBUG is able to match the speed of bottom-up models while surpassing the performance of the state-of-the-art top-down models. © 2019 Association for Computational Linguistics