Electric Vehicle Charging Load Forecasting Considering The Impact Of Non-conventional Events

Electric vehicles are considered as one of the effective ways to solve air pollution and a strategic choice under the "double carbon" goal,and have been developed rapidly in recent years.The large scale EV load has strong randomness,which is a great challenge for the safety and stability of power system,and in this context,the EV charging load prediction is of great significance.However,the occurrence of unconventional events can have a great impact on people’s production and life,such as the COVID-19 epidemic,heavy rainstorms,and large mass events,which affect the EV charging load.Therefore,this paper considers the characteristics of unconventional events and establishes a high-precision electric vehicle charging load prediction model that can quickly respond to such events.Firstly,the set of influencing factors under different unconventional events was constructed,and a candidate factor set consisting of conventional and unconventional factors was formed.On this basis,a specific analysis of the relationship between various types of unconventional event influencing factors and EV charging load was carried out.In the arithmetic analysis with the data related to the COVID-19 epidemic,the epidemic-related data such as the number of new confirmed cases and the number of medium and high risk areas were personalized,and the mapping interval of the epidemic-related data was increased to improve the data differentiation and the sensitivity of each subseries to the changes of the epidemic,making it more suitable for the prediction of EV charging load.By calculating the permutation entropy to select the appropriate base learner for each subsequence after the secondary decomposition,and using the experimental error method and IPSO to optimize the number of hidden layers and Dropout parameters in the LSTM,the superiority of ESMD for such problems and the advantages of the secondary decomposition are verified by example analysis.Secondly,the prediction method of secondary decomposition and integrated learning is proposed.Considering that the EV charging load under the influence of unconventional events has obvious trend changes,the original EV charging load sequence is decomposed using extremepoint symmetric mode decomposition(ESMD),and the decomposed sequence has more significant trend and periodicity,and the remaining terms can reflect the changes of unconventional events,and then the variational mode decomposition(VMD)is applied to redecompose the high-frequency components after ESMD to reduce the complexity and volatility.By calculating the permutation entropy to select the appropriate base learner for each subsequence after the secondary decomposition,and using the experimental error method and IPSO to optimize the number of hidden layers and Dropout parameters in the LSTM,the superiority of ESMD for such problems and the advantages of the secondary decomposition are verified by example analysis.Finally,a hybrid prediction model considering the effects of unconventional events is established.The partial autocorrelation coefficient function(PACF)is used to establish a feature set for each subseries after secondary decomposition,and the maximum information coefficient(MIC)is used to select the best feature set from the established feature set,and the corresponding best feature set is added to the prediction of each subseries,and finally the prediction results of each subseries are integrated to obtain the final prediction results.The prediction results of different single prediction models,different influencing factors and different forecasting periods are analyzed by simulation.The results show that the proposed model in this paper improves the prediction accuracy by 20.59%compared with the secondary decomposition-integrated learning model that only considers conventional factors,which verifies the stability of the model.The results show that compared with the secondary decomposition-ensemble learning model and the secondary decomposition-ensemble learning model considering only conventional factors,the prediction accuracy of the proposed model is improved by 7.37%and 1.69%,respectively,which verifies the stability of the proposed model.In summary,the EV charging load prediction model considering the influence of unconventional events proposed in this paper can improve the accuracy of EV charging load prediction,which is of practical significance for safe and stable operation of the power grid,intraday demand-side response and optimal dispatching of distribution networks containing EVs.