Solar radiation prediction using recurrent neural network and artificial neural network: A case study with comparisons

Abstract With the rapid advancement of the high-performance computing technology and the increasing availability of the mass-storage memory device, the application of the data-driven models (e.g., the artificial neural network (ANN) model) for solar radiation prediction is appearing in abundance in the past decade. Although the performances of these models have been discussed in a large number of studies, how to further enhance the forecasting accuracies of these data-driven approaches to better facilitate the advanced controls in the building system such as model predictive control (MPC) in smart buildings remains a challenge. Deep learning, which is considered as a powerful tool to move machine learning closer to one of its original goals, i.e., Artificial Intelligence (AI), is a viable solution to this problem. In this study, an ANN model and a recurrent neural network (RNN) model are developed to investigate the performances of the deep learning algorithms for the solar radiation prediction. The actual meteorological data (AMY) from a local weather station in Alabama is used for the training process. Various scenarios, including different sampling frequencies and moving window algorithms, are included for a comprehensive evaluation of the accuracies and efficiencies. The results suggest that compared with the ANN model, the solar radiation prediction using the RNN model has a higher prediction accuracy, with a 47% improvement in Normalized Mean Bias Error (NMBE) and a 26% improvement in Root-Mean-Squared Error (RMSE). Besides, this forecasting accuracy could even be taken to a higher level by increasing the granularity of the data or adding a moving-window algorithm to the prediction model. By increasing the sampling frequency of the training data from 1 h to 10 min, the Root-Coefficient of Variation Mean-Squared Error (CV(RMSE)) of the ANN model dropped from 30.9% to 9.41%, while the CV(RMSE) of the RNN model dropped from 9.83% to 7.64%. For the RNN model, the NMBE was improved from 0.9% to 0.2% after the implementation of the moving-window algorithm. Besides, it was found that the cloud cover could have a significant impact on the prediction accuracy.

[1]  Guoqiang Peter Zhang,et al.  Time series forecasting using a hybrid ARIMA and neural network model , 2003, Neurocomputing.

[2]  Martin T. Hagan,et al.  Neural network design , 1995 .

[3]  Manfred Morari,et al.  Use of model predictive control and weather forecasts for energy efficient building climate control , 2012 .

[4]  Betul Bektas Ekici,et al.  Prediction of building energy consumption by using artificial neural networks , 2009, Adv. Eng. Softw..

[5]  Alberto Bemporad,et al.  Model Predictive Control (MPC) for Enhancing Building and HVAC System Energy Efficiency: Problem Formulation, Applications and Opportunities , 2018 .

[6]  Draguna Vrabie,et al.  Simulation and experimental demonstration of model predictive control in a building HVAC system , 2015 .

[7]  Theo Chidiezie Chineke,et al.  Equations for estimating global solar radiation in data sparse regions , 2008 .

[8]  Cyril Voyant,et al.  Numerical Weather Prediction (NWP) and hybrid ARMA/ANN model to predict global radiation , 2012, ArXiv.

[9]  S. S. Bedi,et al.  Weather Forecasting Using Sliding Window Algorithm , 2013 .

[10]  Philip Haves,et al.  Model predictive control for the operation of building cooling systems , 2010, Proceedings of the 2010 American Control Conference.

[11]  Chen Zhe,et al.  Extracting typical occupancy data of different buildings from mobile positioning data , 2018, Energy and Buildings.

[12]  Shuhui Li,et al.  Training Recurrent Neural Networks With the Levenberg–Marquardt Algorithm for Optimal Control of a Grid-Connected Converter , 2015, IEEE Transactions on Neural Networks and Learning Systems.

[13]  Athanasios Sfetsos,et al.  Univariate and multivariate forecasting of hourly solar radiation with artificial intelligence techniques , 2000 .

[14]  Nicholas Gayeski,et al.  Predictive pre-cooling control for low lift radiant cooling using building thermal mass , 2010 .

[15]  Kurt Roth,et al.  Advanced Controls for Commercial Buildings: Barriers and Energy Savings Potential , 2006 .

[16]  Jürgen Schmidhuber,et al.  LSTM: A Search Space Odyssey , 2015, IEEE Transactions on Neural Networks and Learning Systems.

[17]  Zheng O'Neill,et al.  Comparisons of inverse modeling approaches for predicting building energy performance , 2015 .

[18]  Victor M. Zavala,et al.  On-line economic optimization of energy systems using weather forecast information. , 2009 .

[19]  Cameron W. Potter,et al.  Building a smarter smart grid through better renewable energy information , 2009, 2009 IEEE/PES Power Systems Conference and Exposition.

[20]  Yacine Rezgui,et al.  Trees vs Neurons: Comparison between random forest and ANN for high-resolution prediction of building energy consumption , 2017 .

[21]  Francesco Borrelli,et al.  Implementation of model predictive control for an HVAC system in a mid-size commercial building , 2014 .

[22]  Jessica Granderson,et al.  Assessment of Automated Measurement and Verification (M&V) Methods , 2015 .

[23]  C. K. Chan,et al.  Prediction of hourly solar radiation using a novel hybrid model of ARMA and TDNN , 2011 .

[24]  Geoffrey E. Hinton,et al.  Deep Learning , 2015, Nature.

[25]  Zheng O'Neill,et al.  Uncertainty quantification and sensitivity analysis of the domestic hot water usage in hotels , 2018, Applied Energy.

[26]  Peng Xu,et al.  Application of mobile positioning occupancy data for building energy simulation: An engineering case study , 2018, Building and Environment.

[27]  Bing Dong,et al.  Integrated building control based on occupant behavior pattern detection and local weather forecasting , 2011 .

[28]  A. Mellit,et al.  A 24-h forecast of solar irradiance using artificial neural network: Application for performance prediction of a grid-connected PV plant at Trieste, Italy , 2010 .

[29]  Xiao Chen,et al.  Model predictive control for indoor thermal comfort and energy optimization using occupant feedback , 2015 .

[30]  Kamaruzzaman Sopian,et al.  Modeling of Daily Solar Energy on a Horizontal Surface for Five Main Sites in Malaysia , 2011 .

[31]  Steven T. Taylor,et al.  Energy savings and ventilation performance from CO2-based demand controlled ventilation: Simulation results from ASHRAE RP-1747 (ASHRAE RP-1747) , 2019, Science and Technology for the Built Environment.

[32]  Zheng O'Neill,et al.  The role of sensitivity analysis in the building performance analysis: A critical review , 2020 .

[33]  Defeng Qian,et al.  Nationwide savings analysis of energy conservation measures in buildings , 2019, Energy Conversion and Management.

[34]  Adel Mellit,et al.  Prediction of daily global solar irradiation data using Bayesian neural network: A comparative study , 2012 .

[35]  Mawloud Guermoui,et al.  Hybrid models for global solar radiation prediction: a case study , 2020 .

[36]  Yan Su,et al.  Real-time prediction models for output power and efficiency of grid-connected solar photovoltaic systems , 2012 .

[37]  Srinivas Katipamula,et al.  Energy savings potential from improved building controls for the US commercial building sector , 2018 .

[38]  Prashant J. Shenoy,et al.  Predicting solar generation from weather forecasts using machine learning , 2011, 2011 IEEE International Conference on Smart Grid Communications (SmartGridComm).

[39]  Wojciech Zaremba,et al.  Recurrent Neural Network Regularization , 2014, ArXiv.

[40]  Les E. Atlas,et al.  Recurrent neural networks and robust time series prediction , 1994, IEEE Trans. Neural Networks.

[41]  Wei Qiao,et al.  Short-term solar power prediction using a support vector machine , 2013 .

[42]  Nima Amjady,et al.  Solar energy forecasting based on hybrid neural network and improved metaheuristic algorithm , 2018, Comput. Intell..

[43]  Saad Mekhilef,et al.  Performance evaluation of hybrid adaptive neuro-fuzzy inference system models for predicting monthly global solar radiation , 2018 .

[44]  Xing Lu,et al.  Electricity demand response in China: Status, feasible market schemes and pilots , 2016 .

[45]  Weerakorn Ongsakul,et al.  Levenberg-Marquardt Recurrent Networks for Long- Term Electricity Peak Load Forecasting , 2011 .

[46]  Draguna Vrabie,et al.  A Wireless Platform for Energy Efficient Building Control Retrofits , 2012 .

[47]  Annamária R. Várkonyi-Kóczy,et al.  A Hybrid Machine Learning Approach for Daily Prediction of Solar Radiation , 2018, Recent Advances in Technology Research and Education.

[48]  Henrik W. Bindner,et al.  Application of Model Predictive Control for active load management in a distributed power system with high wind penetration , 2012, 2012 IEEE Power and Energy Society General Meeting.

[49]  Yan Chen,et al.  Saving Building Energy through Advanced Control Strategies , 2013 .