Research On Micro-Grid Short-Term Load Forecasting And Smart Energy Efficiency Management

Microgrid has received widespread attention as a highly autonomous and flexible power technology that includes renewable energy sources.In this thesis,based on HBA-based EEMD-LSTM user-side short-term load forecasting model and LS-SVM distributed power forecasting model based on Xgboost multi-weight optimization,this thesis proposes a Pareto frontier optimized environment-friendly Target Energy Optimization Scheduling Model.Specific research includes the following aspects:Firstly,taking the short-term load of micro-grid as the forecasting object,and aiming at the characteristics of micro-grid residents’ load as an important part,a prediction model based on human comfort index HBA-EEMD-LSTM is proposed.The human comfort index is effectively integrated into the prediction model,and the overall trend of the time series is determined using the EEMD data decomposition method.In addition,the LSTM based on the S2 S architecture can flexibly adjust load forecasting at different time scales.The experimental results show that the HBA-EEMD-LSTM model improves prediction accuracy by 6.8% in short-term load forecasting on the microgrid compared to a single LSTM algorithm.Secondly,the distributed power supply is used as the forecasting object.Based on the non-linear and non-stationary characteristics of photovoltaic power,Xgboost algorithm is used to evaluate the weights of each feature,and an xgboost-K-means-LSSVM prediction model based on multi-weight optimization is proposed.Experiments show that compared with the original algorithm and the unmodified k-means algorithm,the prediction accuracy is improved by 5.94% and 3.62%,respectively.Thirdly,aiming at the characteristics of microgrid scheduling,a multi-objective optimization algorithm for dynamic economic environment based on Pareto optimization is proposed.The algorithm establishes time series for individuals associated with the same reference point to predict populations in the new environment,and feeds back historical prediction errors to the current prediction to improve the accuracy of the prediction.Compared with the non-optimized NSGA-II scheduling model,the optimized model can reduce the emission of pollutants while reducing operating costs,better assist the micro-grid to fully absorb renewable energy,promote energy saving and emission reduction,and help the main network to reduce Pinggu,and improve the system robustness.The work of this thesis is practical,and proposes a dynamic multi-objective scheduling optimization model based on user-side prediction,which provides a practical and effective scheduling scheme for China’s micro-grid to achieve economical,environmentally-friendly,efficient,and reliable power supply,and has great practical significance.