Research On Multi-time Scale Renewable Energy Generation Forecasting Methods And Power Scheduling In Microgrids

Due to the continuous consumption of fossil fuels and their pollution,renewable energy and distributed power generation have attracted more and more attention and research.In the development of distributed generation,the emergence of microgrid(MG)technology fits the needs of The Times.Among them,the integrated energy management in MG is the key technology to achieve the efficient utilization of renewable energy and the overall economy of MG.This is the key for MG to attracting users and promoting it in the power system.Therefore,it is of great significance to study the integrated energy management technology in MG.In this paper,the prediction method of renewable energy generation and the power scheduling strategy in MG in multi-time scale are studied.At present,there are two difficulties in MG energy management: firstly,the generating capacity of renewable energy is highly volatile;secondly,optimization objectives of power scheduling in the long and short time scale are different.In view of the first difficulty the data relationship between the generation capacity of renewable energy and meteorological variables in the long and short time scale is fully studied in this paper,and multiple methods are combined to mine effective information to obtain accurate predicted values.In the long time scale,a hybrid online adaptive learning model combining time-varying linear model and GABP is proposed,which can continuously improve the prediction accuracy through adaptively learning the new data.In the short time scale,a comprehensive time series prediction model based on LASSO and LSTM is proposed,which can obtain very accurate predicted value of renewable energy generation in the short term.Aiming at the second difficulty,a microgrid scheduling strategy in MG based on different time scales is proposed in this paper.In the long time scale,the robust day-ahead scheduling strategy based on particle swarm optimization and teaching-learning-based optimization(PSO-TLBO)is used to obtain the Pareto front with economy and environmental protection as the objective function under the condition of considering the prediction error of renewable energy generation capacity.In the short time scale,the Smooth Electric Power Scheduling for Deferrable Load(SEPS-DL)algorithm based on Majorization theory is used to carry out short-term scheduling,and a safe and reliable power scheduling strategy is obtained.The simulation verification based on real data on the prediction method and power scheduling strategy proposed in this paper are also carried out in MATLAB,which confirmes the accuracy and high efficiency of the above method.