Short-term Power Load Forecasting Of Residential Users Based On Quantum Neural Network

The randomness and nonlinearity of power load will affect the stability of power system operation,and accurate power load forecasting is a basic task to ensure the stable operation of power system.Although the short-term forecasting of aggregated load data has been widely studied,with more and more distributed power sources connected,the residential users’ electricity consumption patterns become more complex,which still challenges the accuracy and adaptability of short-term load forecasting.This paper summarizes the common power load forecasting methods,and summarizes the existing load forecasting methods at home and abroad and the research of quantum neural network.According to the characteristics of power load,the influencing factors are analyzed to determine the input vector of the model.Combining the concept of quantum parallel computing with the long-term and short-term memory neural network(LSTM),a quantum long-term and short-term memory neural network(QLSTM)model is established,and it is applied to the daily time scale power load forecasting problem.Aiming at the nonlinear problem of microgrid distributed generation,the optimal dispatching strategy based on improved PSO algorithm and considering demand response is studied.The main research contents are as follows:(1)According to the historical load characteristics and the correlation between meteorological factors,the LSTM short-term power load forecasting model is established,which effectively solves the problem of long dependence of time series of RNN.Considering the meteorological influence factors of temperature,humidity and wind speed,the simulation experiment is carried out by using the actual data of Shanghai area.The results show that compared with BP,SVR and RNN models,LSTM neural network has better prediction effect in short-term load forecasting.(2)The structure algorithm and network update rules of quantum neural network are analyzed,and quantum variational circuit(VQC)and long-term and short-term memory neural network(LSTM)are combined into a new prediction model,which is QLSTM.The QLSTM model is used to forecast and simulate the power load in Shanghai.The experimental results show that the QLSTM model can effectively improve the forecasting accuracy of residential users’ power load in different seasons.The average MAPE value of QLSTM model in four seasons is 2.18%,which is 1.03%higher than that of LSTM model.(3)Aiming at the nonlinear problem of wind-solar storage microgrid,an improved PSO algorithm is introduced,and an optimal dispatching model considering both source and demand sides is established.Three scenarios(PSO algorithm,improved PSO algorithm and improved PSO algorithm considering demand response)are analyzed.The simulation results show that the improved PSO algorithm considering demand response model can effectively reduce the system operation cost,reduce the load peak-valley difference and improve the economy and stability of microgrid system.