Micro-grid Short-term Load Forecasting Based On Data Mining And GSA-BP Multi-model Neural Network

With the rapid development of China's economy,the problem of energy depletion become more and more serious.Micro-grid as a clean and friendly energy source can take advantage of local resources to effectively to solve the remote areas of China's central and western areas of electricity difficult,high energy transport costs,low utilization rate.Micro-grid short-term load forecasting is becoming a hotspot in microgrid system,which is attracting more and more attention.To effectively forecast microgrid load,can provide a guarantee for the operation of the micro-grid energy-saving and efficient operation,and provide the basis for the power dispatching department to develop the power generation plan.Therefore,it is important to strengthen the microgrid load forecasting both for the micro-grid system itself and for the large power grid.According to the micro-grid load characteristics,this thesis puts forward a multimodel neural network micro-grid short-term load forecasting model based on data mining and genetic simulated annealing algorithm(GSA).The main research work and innovation are as follows:Firstly,this thesis analyzes the factors such as weather,day type and actual historical load,which affect the microgrid load.Based on these factors,the preliminary data of microgrid load forecasting is established,and the data is excavated by data mining.The above basic steps are used to establish the basic prediction model.The specific approach is as follows:(1)The rough set attribute reduction processing is applied to the input of the forecasting network,and the core factors influencing the micro-grid load forecast are found through this step and used as the input of the prediction model;(2)According to the characteristics of micro-grid load fluctuation and difference,the sample data are clustered into several classes based on the fuzzy clustering analysis,and the corresponding BP neural network prediction model is established for each sample;(3)In the prediction of the micro-grid load,the relevant network is searched by the pattern recognition technology,and the network load is forecasted by using this network.The basic model of micro-grid short-term load forecasting based on multi-model BP network is established by the above steps.The simulation results show that the forecasting model can achieve the desired forecasting results.Secondly,considering the shortcomings of BP neural network iteration speed and easy to fall into local extreme,a prediction model of BP multi-model network optimized by GSA algorithm is proposed.The GSA algorithm combines the global search capability of the genetic algorithm(GA)and the probabilistic sudden jump characteristics of the simulated annealing(SA)algorithm,which is used with the BP network to predict the load.The optimized model improves the accuracy of micro-grid load forecasting.The advantages of GSA-BP multi-model forecasting model are verified by comparison with other prediction algorithms.Finally,through the analysis of foreign micro-grid load operation found that realtime price factors will affect the micro-network load size.Therefore,this thesis introduces the real-time electricity price factor into the forecasting model,and uses the fuzzy control algorithm to correct the predicted micro-grid load.The experimental results show that the algorithm can effectively correct the prediction results of real-time electricity price factors.