QLSTM Neural Network Prediction And Its Economic Operation In EV-containing Microgrids Strategy Research

The high randomness and volatility in the renewable energy output of the microgrid.Along with the random loads brought about by the access of large-scale electric vehicles(EVs),this leads to greater randomness on the demand side as well.The randomness of both supply and demand will have an impact on the stable operation of the microgrid,and it is necessary to develop a scheduling strategy to improve the stability of the microgrid and reduce operating costs.The complex structure of EVcontaining microgrids makes it difficult to build accurate models,and traditional algorithms are prone to fall into local optimum solutions when solving them.In this paper,we combine the parallel capability of quantum computing to improve the long and short-term memory neural network(LSTM)and establish a quantum long and short-term memory neural network(QLSTM)wind speed prediction model for wind farms.Based on the wind speed prediction results,the Double Q-learning algorithm is used to solve for a stable and economically optimal scheduling solution for EVcontaining microgrids.The main research work in this paper is as follows:The thesis first introduces the research background and the current situation of the quantum neural network(QNN),and the characteristics of EV-containing microgrids and the shortcomings of traditional optimization strategies are described.After identifying the prediction performance advantages of quantum neural networks,the LSTM units are improved in the quantum framework through quantum gates to obtain QLSTM units that can give full play to the quantum parallel computing capability,which provides the basis for building wind speed prediction models for wind farms in the following.Secondly,due to the weak regularity of wind speed data series and the difficulty of prediction,traditional neural network prediction algorithms cannot guarantee both prediction accuracy and training rate.In this paper,we propose a QLSTM wind speed short-term prediction model for wind farms,using meteorological factors such as air temperature and humidity as input features,and train the model using historical data to compare the prediction errors with traditional neural network models(CNN,LSTM,CNN-LSTM)under different seasonal conditions.The experimental results show that the average MAE,RMSE and MAPE values of the QLSTM model are 0.274m/s,0.382m/s and 12.698% respectively in the four seasons,which are all lower than other models,and the training rate is improved by 18.24% compared to LSTM,verifying the effectiveness of the method.Then,the problem of large-scale EV access to microgrids affecting their economic and stable operation is addressed.A mathematical model is established for each microsource in the EV-containing micro-grid into which the charging characteristics and travel characteristics of EVs are considered,and the disorderly charging load of EVs on weekdays and holidays is modelled using the Monte Carlo method.A charging demand response model is established through the electricity tariff elasticity matrix to determine the EV charging load corresponding to the real-time tariff.Finally,in order to develop a more realistic economic dispatching scheme before the day,the objective functions and constraints for the economic operation of EVcontaining microgrid are established,the wind speed prediction results of the QLSTM model are fitted to the wind power as input data,and the Double Q-Learning algorithm is used to design an economic optimal dispatching strategy,which is simulated under two scenarios: weekdays and holidays.The experimental results show that the algorithm can solve the over-estimation problem of the Q-learning algorithm and can combine the current environmental information to obtain the optimal economic dispatching scheme.Compared with the strategy designed by Q-Learning,it results in an average reduction of 8.72% in the economic cost of the microgrid,and the improvement of wind power prediction accuracy by QLSTM prediction algorithm can effectively reduce the abandoned wind rate and load loss rate in the microgrid.