Multi-source Complementary Micro-grid Energy Intelligent Prediction And Optimized Management

Multi-source complementary microgrid is an organic part of the energy internet,and has great potential in promoting the consumption of renewable power,improving energy efficiency,and improving the flexibility and reliability of power system.Therefore,it has also received continuous attention from researchers.However,renewable power sources with random characteristics increase the uncertainty of the energy supply side of the multi-source complementary micro-grid,and also bring challenges to the energy optimization of it.This article focuses on energy prediction and optimization in a multi-source complementary microgrid.On the one hand,deep learning models are proposed to accurately predict the photovoltaic and wind power output in the multi-source complementary microgrid to provide the output curve for day-ahead energy optimization of the microgrid;on the other hand,day-ahead energy optimization analysis of a multi-source complementary microgrids in grid-connected communities is carried out.The main work is as follows:1)The photovoltaic power prediction in the multi-source complementary microgrid was studied.First,the key influencing factors were screened through the pearson correlation coefficient based on the photovoltaic data space,and then the genetic algorithm improved bidirectional long short-term memory neural network(GA-BiLSTM)is built to predict photovoltaic power at different time scales in the microgrid(5min,15min,30min,60min).Then,to verify the performance,the proposed model is compared with long and short-term memory neural network(LSTM),extreme learning machine(ELM),BP neural network(BPNN)and genetic algorithm optimized BPNN(GA-BPNN).The results show that the proposed model has the higher prediction accuracy under different time scales.2)The wind power prediction in the multi-source complementary microgrid was studied,and the influencing factors were screened through gray correlation analysis under two different normalization methods based on the wind data space,and then a hybrid wind power prediction model(bidirectional long-short-term memory neural network(Bi-LSTM)combined with convolutional neural network(CNN),BiLSTM-CNN)is proposed to predict wind power;in order to verify the prediction performance of the proposed model,the proposed model is compared with other single models(Bi-LSTM,LSTM,CNN)and hybrid models(LSTM-CNN,CNN-BiLSTM,CNN-LSTM).The results show that the proposed hybrid deep learning BiLSTM-CNN model has high prediction accuracy.3)A simulation case study of a multi-source complementary microgrid in a grid-connected community including wind,solar,storage and electric vehicles was set as target case,and a quantum particle swarm optimization model(QPSO)was established to allocate its optimal load ahead of a day.First,mathematical modeling of its system structure and components is established.Second,considering economic and environmental aspects,the objective is given including the transaction costs between the microgrid and the main grid,the depreciation costs of electric vehicles and energy storage systems,and pollutant treatment costs;besides,the constraints of the multi-source complementary microgrid are also expounded.Through the prediction models proposed before,the output curves of photovoltaic and wind power are obtained respectively.Third,the quantum particle swarm optimization algorithm(QPSO)is applied to the optimal load distribution under three different scenarios.Through simulation analysis,the feasibility of the optimization algorithm proposed in this paper is verified,as well as the important role of electric vehicles and energy storage systems in reducing the total cost of community microgrids.