Research On Optimal Dispatching Method Of Regional Power Grid Considering Wind Power Prediction And EV Flexibility

With the continuous increase in the number of electric vehicles,large-scale electric vehicles will be connected to the grid for disorderly charging,which will bring about problems such as increased network loss,decreased power quality,and increased difficulty in the optimization control of power grid operation.However,with the increase in the amount of renewable energy connected to the grid,the consumption of new energy,especially wind power,is becoming more serious,and the safe and stable operation of the power grid is under great threat.Due to its mobile energy storage characteristics,electric vehicles have great potential in vehicle network interaction,peak regulation,frequency modulation and other aspects,which has attracted wide attention.Fully exploiting the flexibility of electric vehicles,studying wind power forecasting,EV charging and discharging scheduling,and regional power grid optimization are of great significance to ensure the safe and stable operation of power grid and improve the proportion of new energy consumption.In this paper,from the two aspects of peak trimming and valley filling of regional power grid and the overall optimal dispatching of regional power grid,considering the peak regulating demand,wind power forecast,EV charging and discharging and other factors,the optimal dispatching method of regional power grid of new power system is studied.Firstly,the basic interaction model between EV and power grid considering is studied,and the basic theory of wind power prediction including empirical mode decomposition and neural network is studied.Secondly,based on Bidirectional Long Short-term Memory(BILSTM)neural network,an optimal scheduling method of multi-type electric vehicles participating in regional power grid peak adjustment is proposed.The proposed method analyzes the characteristics of the historical real data of multiple types of vehicles.The Gaussian distribution was used to generate different types of EV data,and the global optimization model of multi-type EV participating in the peak scheduling of the power grid was established to minimize the fluctuation of the power grid load curve and the expectation of the dispatching cost.The linear programming model was solved by using the solver.Based on the model scheduling results,the BILSTM network input matrix was optimized,and the network was trained after the optimization of the input matrix.In the real-time scheduling stage,the BILSTM network after training is called and the results are updated in real time.In the example part,the difference of load curves before and after the participation of multiple types of electric vehicles is compared and analyzed,and the influence of different neural network algorithms on the peak regulation effect and the income of aggregators are discussed.Finally,a multi-time scale scheduling method for regional power grid considering the flexibility of electric vehicles and wind power consumption is proposed to solve the problem that wind power/base load is difficult to accurately predict and lead to the decline of real-time control accuracy.The proposed method performs modal decomposition processing on historical data of wind power and base load.Furthermore,BILSTM algorithm was used to reconstruct the prediction of Intrinsic Mode Functions(IMF)components with different frequencies,and the prediction data were used in the day-day real-time model.In day-ahead optimization,the mixed integer nonlinear programming model is solved to obtain the output curve of day-ahead thermal power unit.In the intra-day real-time model,the wind power and load prediction errors from the current moment to some point in the future are optimized in real time by taking advantage of the charging and discharging flexibility of electric vehicles,so as to maintain the power balance.The effectiveness of the method is verified by an example simulation.