Energy Consumption Prediction Research Articles

Machine Learning Algorithms for Predicting Energy Consumption in Educational Buildings

In the past few years, there has been a notable interest in the application of machine learning methods to enhance energy efficiency in the smart building industry. The paper discusses the use of machine learning in smart buildings to improve energy efficiency by analyzing data on energy usage, occupancy patterns, and environmental conditions. The study focuses on implementing and evaluating energy consumption prediction models using algorithms like long short-term memory (LSTM), random forest, and gradient boosting regressor. Real-life case studies on educational buildings are conducted to assess the practical applicability of these models. The data is rigorously analyzed and preprocessed, and performance metrics such as root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) are used to compare the effectiveness of the algorithms. The results highlight the importance of tailoring predictive models to the specific characteristics of each building’s energy consumption.

Comparative assessment and policy analysis of forecasting quarterly renewable energy demand: Fresh evidence from an innovative seasonal approach with superior matching algorithms

In the pursuit of sustainable development, accurate renewable energy demand forecasting holds great significance for climate change mitigation and promoting sustainability. However, renewable energy forecasting has been consistently challenged by seasonality and nonlinearity. Identifying the periodic and nonlinear characteristics concealed within renewable energy sources accurately is still an unexplored problem. Consequently, an innovative nonlinear discrete seasonal grey model is proposed for renewable energy forecasting, which incorporates seasonal dummy variables and a power exponent term for handling the seasonality and nonlinear patterns in time series. Furthermore, an intelligent algorithm matching framework is proposed to augment the flexibility of the newly developed model. For practical purposes, the new methodology is contrasted against a range of benchmarks encompassing statistical, machine-learning, and traditional grey models in forecasting the quarterly total renewable energy consumption in the United States. The proposed model exhibits over 27% improvement rates over its counterparts, achieving the most superior predictive accuracies of 1.45%, 39.27, and 0.79 in MAPEP, RMSEP, and MASEP metrics, respectively. Furthermore, the probability density and sample size analyses are conducted to validate the robustness of the new model, confirming its adaptability and stability towards algorithm randomness and historical information volume. Consequently, the novel model is employed to forecast the short-to-long-terms renewable energy consumption in the U.S., showcasing an upward trend and seasonal fluctuations of the consumption for the forthcoming 24 quarters from 2023Q4 to 2029Q3. These insights can offer valuable implications to the stakeholders such as energy suppliers, utility managers, and policy advocates, highlighting actionable strategies for optimizing renewable energy consumption forecasting and aiding sustainable development initiatives.

Research on the Energy Consumption Influence Mechanism and Prediction for the Early Design Stage of University Public Teaching Buildings in Beijing

The public teaching buildings of universities have a large flow of people, high lighting requirements, and large energy consumption, which present significant potential for energy saving. The greatest opportunity for integrating “green” architectural design strategies lies in the design phase, especially the early stage of architectural design. However, current designers often rely on experience or qualitative judgment for decision-making. Thus, there is a pressing need for rational and quantitative green architectural design theories and techniques to guide and support decision-making for the design parameters of teaching buildings. This study, based on field surveys of 40 teaching buildings, constructs building archetypes regarding energy consumption including 28 typical values. Based on the “Rectangle”, “L”, “U”, and “Courtyard” archetypes, through batch energy consumption simulation and multiple regression methods, the influence mechanisms of nine energy consumption influencing factors on four types of building energy consumptions were explored, and energy consumption prediction models were derived. The findings of this research can serve as factor evaluation and selection in the early stage of architectural design for public teaching buildings at universities, and the prediction model can assist in the early estimation of energy consumption. This aims to enrich and supplement green architectural design methods by supporting the design of green public teaching buildings and providing reference and application for relevant engineering practices.

Predicting Liquid Natural Gas Consumption via the Multilayer Perceptron Algorithm Using Bayesian Hyperparameter Autotuning

Reductions in energy consumption and greenhouse gas emissions are required globally. Under this background, the Multilayer Perceptron machine-learning algorithm was used to predict liquid natural gas consumption to improve energy consumption efficiency. Setting hyperparameters remains challenging in machine-learning-based prediction. Here, to improve prediction efficiency, hyperparameter autotuning via Bayesian optimization was used to identify the optimal combination of the eight key hyperparameters. The autotuned model was validated by comparing its predictive performance with that of a base model (with all hyperparameters set to the default values) using the coefficient of variation of root-mean-square error (CvRMSE) and coefficient of determination (R2) based on the Measurement and Verification Guideline evaluation metrics. To confirm the model’s industrial applicability, its predictions were compared with values measured at a small-to-medium-sized food factory. The optimized model performed better than the base model, achieving a CvRMSE of 12.30% and an R2 of 0.94, and achieving a predictive accuracy of 91.49%. By predicting energy consumption, these findings are expected to promote the efficient operation and management of energy in the food industry.

Two processes based on a data-driven model combined with dynamic simulation for demand forecasting and providing energy saving measures

In this study, two processes of predictive model combining dynamic simulation and regression model were presented. First, (1) the demand forecasting process defined in this study is a method of predicting energy consumption based on the simulated energy consumption of a target building, and (2) the energy saving process is a method of setting an energy reduction target based on the predicted energy consumption and providing energy saving measures to achieve the target. Specifically, there are two functions based on the regression models which are predicting actual energy consumption by inputting the energy consumption simulation data, and inversely inputting the actual energy consumption and deriving the corresponding simulation conditions. As a result of predicting building energy consumption through the demand forecasting process, the performance in the heating season showed CV-RMSE was 14.622%, and in the cooling season CV-RMSE was 5.877%. In addition, in the case study using the energy saving process, it was possible to derive the heating and cooling set temperature for 5% energy saving as an energy saving method by using weather forecast data.

Building consumption anomaly detection: A comparative study of two probabilistic approaches

This paper investigates the performance of two probabilistic approaches, Ensemble batch Prediction Intervals (EnbPI) a conformal prediction approach and XGBoost Location, Scale and Shape (XGBoostLSS), in predicting building energy consumption and in detecting systemic anomalies with proposed alarm matrix. The research questions focus on the effectiveness of these models in providing both point and probabilistic predictions and their utility in identifying collective anomalies. Both models showed good point and distribution prediction performance. For example, the observed point prediction had CV-RMSE in the range of 9 to 17%, outperforming recommendations from ASHRAE. Furthermore, a post-processing stage, the alarm matrix, effectively flags collective, repetitive anomalies, thus shifting focus from conventional point anomalies. The EnbPI-based method yields higher recall rates, with a trade-off of having more false alarms, while XGBoostLSS-based method excels in precision, minimizing overlooked alarms. Moreover, a robustness analysis was carried out to evaluate how these models performed when faced with training datasets containing anomalies. The robustness analysis revealed that the EnbPI-based method was more prone to overfitting, meaning its performance actually improved when the training data included some noise. On the other hand, the XGBoostLSS-based method was more stable, performing well with low levels of noise with performance drop when the noise level was high. While the findings contribute significantly to building energy consumption prediction and anomaly detection, future research could address performance in dynamic environments for the methodology and explore continual learning strategies.

A GPU-accelerated adaptation of the PSO algorithm for multi-objective optimization applied to artificial neural networks to predict energy consumption

Optimization research often confronts the challenge of developing time consuming processes. This article introduces an innovative approach that leverages the computational power of Graphics Processing Units (GPUs) to speed up that optimization process. We present an innovative adaptation of Particle Swarm Optimisation (PSO) to meet the requirements of multiobjective optimization problems. This approach aims to leverage the strengths of a multi-objective approach to perform energy consumption prediction using neural networks. By employing GPU parallel techniques, our method not only speeds up the optimization process but also enhances the efficiency of neural network training execution. The main advantage of our approach lies in its dual ability to simultaneously optimizing neural network architectures by determining the minimum number of hidden neurons and fitting the weights of the networks in order to achieve the lowest error. Preliminary results suggest a notable enhancement in prediction accuracy of forecasting electric energy consumption, as a result of optimizing the architecture and parameters of the neural network using the proposed method. This PSO adaptation stands out for its ability to address complex problems, increase efficiency and produce accurate predictions, making it a novel solution in Machine Learning heuristic methods for application in the solution of advanced prediction problems with time constraints from time series.

Bi-scale car-following model calibration based on corridor-level trajectory

The precise estimation of macroscopic traffic parameters, such as travel time and fuel consumption, is essential for the optimization of traffic management systems. Despite its importance, the comprehensive acquisition of vehicle trajectory data for the calculation of these macroscopic measures presents a challenge. To bridge this gap, this study aims to calibrate car-following models capable of predicting both microscopic measures and macroscopic measures. We conduct a numerical analysis to trace the cumulative process of model prediction errors across various measurements, and our findings indicate that macroscopic measures encapsulate the accumulation of model errors. By incorporating macroscopic measures into vehicle model calibration, we can mitigate the impact of noise on microscopic data measurements. We compare three car-following model calibration methods: MiC (using microscopic measurements), MaC (using macroscopic measurements), and BiC (using both microscopic and macroscopic measurements)—utilizing real-world trajectory data. The BiC method emerges as the most successful in reconstructing vehicle trajectories and accurately estimating travel time and fuel consumption, whereas the MiC method leads to overfitting and inaccurate macro-measurement predictions. This study underscores the importance of bi-scale calibration for precise traffic and energy consumption predictions, laying the groundwork for future research aimed at enhancing traffic management strategies.

Applications of Fractional Time Delayed Grey Model in Primary Energy Consumption Prediction

The prediction of total energy consumption is crucial across various domains including the economy, environment, market, and geopolitics. Accurate forecasts can guide policy-making, investment decisions, and international strategies, contributing to sustainable development and energy security. Fractional models have been proven to better capture the long-term memory effects and complex dynamic characteristics of systems, with time delay playing a crucial role in capturing dynamic behaviors. Such models enhance the accuracy and reliability of predicting future trends and behaviors. For the prediction of primary energy consumption in South and Central America, the Middle East, and Africa, this study opts for the existing fractional time delayed grey model, optimizing the fractional order using the particle swarm optimization algorithm. Experimental results demonstrate that in most cases, the predictive capability of the fractional time delayed grey model surpasses that of other grey models. This indicates the effectiveness and reliability of the model in forecasting energy consumption, providing valuable references and foundations for decision-making in relevant fields.

Office building energy consumption forecast: adaptive long short term memory networks driven by improved beluga whale optimization algorithm

With the development of urbanization, buildings have become a major source of energy consumption. This research uses a data-driven approach to achieve accurate building energy consumption prediction by analyzing and modeling building energy consumption data. The model proposed in this paper uses improved beluga whale optimization algorithm (IBWO) to optimize long short-term memory networks (LSTM) for accurate energy consumption prediction. In order to enhance the ability of BWO in global search and local exploitation, a new method of dynamic adjustment of step factor as well as strategies such as nonlinear decreasing are introduced to improve BWO. For the first time, it is proposed to explore the accuracy of the number of hyper-parameters of LSTM on the prediction of energy consumption, and the improved beluga whale optimization algorithm is used to optimize the two, three, and four hyper-parameters of LSTM respectively. Then short-term prediction of historical energy consumption data of an office building in Xi`an is performed. Experiments show that the optimization of the four hyperparameters of LSTM using the IBWO of this paper can reduce the mean absolute error (MAE) of the pre-improvement model from 830.71 KW to 128.28 KW, the mean absolute percentage error (MAPE) from 12.32% to 1.38%, and the coefficient of variation (CV) from 7.5% to 1.2%.

Spatio-temporal prediction of total energy consumption in multiple regions using explainable deep neural network

Energy consumption forecasting is essential for energy system integration and management. However, existing studies mainly focus on temporal features of energy consumption, which neglects the spatial correlation of variables with time information. Capturing the spatio-temporal relationships helps to improve forecasting accuracy and further promote energy dispatch. To tackle this problem, an explainable Convolutional Neural Network-Long Short Term Memory forecasting model is employed to effectively predict the total energy consumption by capturing the spatial and temporal features of multivariate time series. In the model, the autoencoder is used to achieve the nonlinear dimensionality reduction and transfer the data to a low-dimensional space. Furthermore, a Convolutional Neural Network is used to extract more effective features from the decoded data, and long short-term memory is employed to identify the temporal dependencies between extracted features and total energy consumption. Shapley additive explanation is introduced to interpret the outputs of the black-box model. The superior performance of the proposed method with high accuracy and good adaptability is verified by the comparisons with conventional forecasting models. This method provides an insight into the regional energy consumption analyzing contributions of weather variables to energy consumption, which helps administers in understanding regional energy performance for enhancing energy efficiency.

Prediction method of regional carbon dioxide emissions in China under the target of peaking carbon dioxide emissions: A case study of Zhejiang

AbstractAll provinces of China respond to the central government, predict future carbon dioxide emissions, and formulate action plans detailing how the province intends to fulfill its target of carbon emission peaking before 2030. Based on the bottom‐up energy consumption prediction and top‐down goal verification, this paper constructs a set of regional carbon dioxide emission prediction methods. Compared to the traditional bottom‐up prediction method, this method could simplify the parameters while improving the prediction accuracy. This model is used to predict and analyze the process of carbon dioxide emission peaking in Zhejiang. The results show that the mean absolute percentage error of the retrospective prediction value is only 1.56%. Zhejiang will reach carbon dioxide emission peaking around 2029–2030, and the peak value will be 569.7 million tons. Different factors have different effects on the process of carbon dioxide emission peaking. There is a strong correlation between the peak time of carbon dioxide emission and the production time of major energy‐consuming projects in Zhejiang. Meanwhile, if the 16 nuclear reactors are not put into operation, Zhejiang will not be able to achieve the goal of carbon dioxide emission peaking. Besides, the basic data used in this model is mainly from the local statistical departments of the region. Thus, it can be applied to other provinces and regions conveniently.

Application of energy scheduling algorithm based on energy consumption prediction in smart home energy scheduling

To effectively optimize household energy management, enhance user comfort, and ensure the stability of the power grid, this paper directs its focus towards household equipment and presents inter-household and indoor energy scheduling algorithms. Grounded in a two-dimensional dynamic device priority framework and predictive device operation modeling, this study demonstrates that the utilization of scheduling algorithms can avert grid overload scenarios during periods of low electricity prices. This results in more consistent hourly electricity consumption and a substantial peak-to-average ratio reduction of up to 44.86%. Additionally, the application of the continuous wavelet neural network algorithm proves to be proficient in accurately forecasting home device operations. When the time was 17.5 hours, the actual energy consumption of the water heater was 7.12 kWh, which was 0.78 kWh more than the predicted energy consumption. Utilizing scheduling algorithms, as opposed to not employing them, led to notable reductions in electricity expenses, demonstrating a cost savings rate of 15.32% on the third day. Additionally, it facilitated decreased usage of the power grid. When the time was 18 o'clock, the power consumption of the grid using scheduling algorithms was 0.24 kWh, which was 0.25 kWh lower than that without scheduling algorithms. Following the implementation of the scheduling algorithm, minimal deviation in device startup times was observed during periods of high comfort, with offsets ranging from 0 hours. The research methods employed demonstrate efficacy in orchestrating energy distribution among households and indoor spaces, thereby striking a harmonious balance between energy conservation and user comfort.

Distributed Heterogeneous Transfer Learning

Transfer learning has proved to be effective for building predictive models even in complex conditions with a low amount of available labeled data, by constructing a predictive model for a target domain also using the knowledge coming from a separate domain, called source domain. However, several existing transfer learning methods assume identical feature spaces between the source and the target domains. This assumption limits the possible real-world applications of such methods, since two separate, although related, domains could be described by totally different feature spaces. Heterogeneous transfer learning methods aim to overcome this limitation, but they usually i) make other assumptions on the features, such as requiring the same number of features, ii) are not generally able to distribute the workload over multiple computational nodes, iii) cannot work in the Positive-Unlabeled (PU) learning setting, which we also considered in this study, or iv) their applicability is limited to specific application domains, i.e., they are not general-purpose methods.In this manuscript, we present a novel distributed heterogeneous transfer learning method, implemented in Apache Spark, that overcomes all the above-mentioned limitations. Specifically, it is able to work also in the PU learning setting by resorting to a clustering-based approach, and can align totally heterogeneous feature spaces, without exploiting peculiarities of specific application domains. Moreover, our distributed approach allows us to process large source and target datasets.Our experimental evaluation was performed in three different application domains that can benefit from transfer learning approaches, namely the reconstruction of the human gene regulatory network, the prediction of cerebral stroke in hospital patients, and the prediction of customer energy consumption in power grids. The results show that the proposed approach is able to outperform 4 state-of-the-art heterogeneous transfer learning approaches and 3 baselines, and exhibits ideal performances in terms of scalability.

Prediction of energy consumption for manufacturing Small and Medium-sized Enterprises (SMEs) considering industry characteristics

There has been a growing demand for energy consumption statistics in the manufacturing industry to establish national energy and greenhouse gas policies. Despite its importance, the Korean government faces significant challenges in collecting energy data at a facility level in a precise and timely manner. To address the lack of timely data, this paper employs machine-learning models to predict the annual total energy consumptions of each manufacturing facility. We first designed four prediction models that take into account the characteristics and energy consumption behaviors of industry sub-sectors. As input variables, these prediction models mainly included electricity consumption, employee size, energy types, gas consumption and other accessible data. Finally, we conducted numerical experiments on approximately 100,000 facilities and evaluated the prediction performance of various machine-learning algorithms such as linear regression, decision tree regression, random forest regression, gradient boost regression, and extreme gradient boosting regression. The numerical experiments provided insights into which model and algorithm offer the best prediction performance for each industry sub-sector. In addition, we identified the important variables for predicting total energy consumption, revealing that not only electricity but also various other energy sources and variables representing industry-specific characteristics play a crucial role in improving prediction performance.

Attention mechanism-aided model ensemble method of chiller energy consumption prediction

The chiller energy consumption prediction plays an important role in reducing the building energy consumption. Continuously improving the performance of the prediction model is the key to ensuring the accuracy of the chiller energy consumption prediction. As a result, a new improved ensemble model of the chiller energy consumption prediction is proposed, based on the idea of algorithmic ensemble aided by an attention mechanism (AM). To guarantee that the proposed model can see information from many spatial and structural viewpoints, the improved ensemble model incorporates the benefits of several base prediction methods. The suggested approach chooses crucial features to simplify the model input by utilizing AM, thereby reducing the complexity of the samples. The improved ensemble model was experimentally verified in an actual multi-chiller system, and the results showed that the established model could obtain good prediction results. Finally, the proposed model was compared with existing, well-known prediction models. To evaluate the accuracy of these models, the mean absolute error (MAE), mean absolute percentage error (MAPE), mean square error (MSE), variance accounted for (VAF), and coefficient of determination (R2) were used as the evaluation indices. With an MAE of 2.858, an MAPE of 1.2755%, an MSE of 20.1878, a VAF of 93.6058, and an R2 of 0.9923, the findings demonstrate that the improved ensemble model has a greater prediction accuracy than the other well-known prediction methods. The suggested model enhances the variety of empirical model algorithm libraries.

Short-term energy consumption prediction for university dormitory buildings based on energy consumption grouping by considering meteorological factors and teaching schedules

Accurate short-term energy consumption prediction is essential for studying renewable energy utilization in building technologies. University dormitory energy consumption is affected by meteorological factors and teaching schedules, presenting a more complex time series compared to residential buildings, rendering traditional methods unsuitable. Therefore, in the present study, we developed a model called energy consumption grouping-Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise-Bi-directional Long Short-Term Memory (ECG-ICEEMDAN-BILSTM) for predicting short-term energy consumption in university dormitory buildings. Model was tested using electricity consumption data from university dormitory buildings in cold regions of China. Based on the energy consumption characteristics, we identified two consumption patterns comprising high power energy consumption (HPEC) and low power energy consumption (LPEC). Using the ECG method, data with similar daily energy consumption distributions for HPEC and LPEC were grouped. Subsequently, individual prediction models were constructed based on ICEEMDAN-BILSTM for each specific energy consumption group. The experimental results showed the ECG-ICEEMDAN-BILSTM model obtained more accurate predictions compared with other neural network models. In addition, the model was more effective at forecasting the complex distribution of the HPEC pattern, with reductions of at least 16.21% and 15.82% in MAPE and PRMSE, respectively. Thus, study still can obtain valuable insights to address the problem of short-term prediction using complex time series.

Multi-source domain generalization deep neural network model for predicting energy consumption in multiple office buildings

The effective prediction of building energy consumption can be used to optimize building operating modes and reduce the overall energy consumption and carbon emission of the building. To predict the energy consumption of the same type of building in the same area, it is often necessary to train separate prediction models for different buildings, which is usually costly computationally in order to get better prediction results. Therefore, this study will combine deep neural networks and multi-source domain generalization in an encoder-decoder architecture to train a model that can be directly applied to predict the energy consumption of multiple buildings in same type. In order to validate the predictive effectiveness of the model, it will be tested on real office building energy consumption dataset and different comparative experiments will be designed on the source and target domains. The results show that the constructed multi-source domain generalization model is able to accurately predict the energy consumption trend of different buildings 1 h in advance. It also has some energy consumption prediction ability for the unknow training set of buildings.

Deep learning for prediction of energy consumption: an applied use case in an office building

Abstract Non-residential buildings are responsible for more than a third of global energy consumption. Estimating building energy consumption is the first step towards identifying inefficiencies and optimizing energy management policies. This paper presents a study of Deep Learning techniques for time series analysis applied to building energy prediction with real environments. We collected multisource sensor data from an actual office building under normal operating conditions, pre-processed them, and performed a comprehensive evaluation of the accuracy of feed-forward and recurrent neural networks to predict energy consumption. The results show that memory-based architectures (LSTMs) perform better than stateless ones (MLPs) even without data aggregation (CNNs), although the lack of ample usable data in this type of problem avoids making the most of recent techniques such as sequence-to-sequence (Seq2Seq). Graphical abstract

Towards Contrastive and Data Driven Cloud Data Center Energy Prediction using Regression Analysis

Abstract: Data centers are the backbone of todays Internet and cloud computing systems. Due to the increasing demand for electrical energy by data centers, it is necessary to account for the vast amount of energy they consume. Energy modeling and prediction of data centers plays a pivotal role in this context. In this study, we address the challenge of predicting energy consumption in cloud data centers, crucial for managing their significant electricity demand. Despite numerous existing methods, there remains a lack of robust methodologies. To fill this gap, we propose a novel approach that incorporates aleatoric uncertainty estimation. Our method utilizes regression distributions to model this uncertainty, with parameters derived from regressive techniques. This yields energy consumption predictions as random variables drawn from these distributions. Additionally, we illustrate how these random variables can be aggregated to form probabilistic forecasts for diverse data center portfolios. Our methodology achieves three key advancements: 1) introducing a simple multiple linear regression model for fundamental series; 2) devising a unique method that combines quantile regression and empirical copulas to estimate joint distribution; 3) enhancing prediction accuracy through a weighted correction technique based on constrained quantile regression.