Research On Monthly Electricity Consumption Forecasting Approach For Virtual Customer Clusters Under High Penetration Of Distributed Photovoltaic

Under a competitive retail electricity market,electricity consumers can decide their own retailers.The retailer participates in the wholesale electricity market instead of the customers,and then sells electricity to the customers in the retail market,earning the difference in price.In the medium-and long-term market,the retailer needs to forecast the monthly electricity consumption of the agent customers in advance,and the accuracy of the forecast will directly affect the ultimate profitability of the retailer.The large-scale access of behind-the-meter distributed photovoltaic(BTM DPV)on the customer side has brought new challenges to the forecasting work.To address these difficulties,this paper investigates the monthly electricity consumption forecasting method for virtual customer clusters under high penetration DPV,which has implications for retailers’ applications in electricity sales and purchase decisions,mainly as follows:1)With the increase of DPV penetration,the monthly electricity consumption of customer clusters containing DPV has changed significantly,and mastering the electricity consumption characteristics in the new situation can play a positive role in guiding the planning and business plans of retailers.Based on this,the characteristics and key influencing factors of DPV,the actual electricity consumption characteristics and key influencing factors of customer clusters,and the characteristics of customer clusters considering different time scales and DPV penetration levels are analyzed respectively,and the differences in the stochastic volatility of monthly electricity consumption of DPV power generation and customer clusters are compared,and the indicators to identify the electricity consumption sequence containing DPV are constructed.These will lay the foundation for the next forecasting work.2)With the increasing maturity of DPV power generation technology and the decreasing cost of photovoltaic modules,the installed capacity of household DPV is growing rapidly driven by policies.Based on the analysis of key influencing factors of DPV,a monthly DPV power consumption forecasting method considering weather type distribution and DPV capacity growth is proposed.First,the weather conditions are classified by using the similarity of DPV capacity curves under different weather conditions,based on which a support vector machine based weather type distribution prediction model is proposed.Then,the PV capacity prediction model based on online time series method is established.Finally,a monthly DPV power generation forecasting model based on artificial neural networks is established,taking weather type distribution and DPV capacity growth into account.The simulation results show that the method effectively improves the monthly DPV forecasting accuracy.3)On the electricity sales side,for residential customers with DPV,the monthly electricity consumption forecast by retailers is actual electricity consumption(AEC)minus distributed photovoltaic power generation(DPVPG),which is called net electricity consumption(NEC).In order to cope with the electricity consumption forecasting problem caused by the large-scale access of DPV and the unpredictability of BTM DPV,a monthly NEC forecasting method is proposed.First,using the indicators that can identify the sequence of electricity consumption containing DPVPG,and the customers are classified into two categories,solar customers and non-solar customers,according to whether they have installed DPV or not.Then,the AEC sequences of non-solar customers are tested online to determine whether DPV is newly installed.For solar customers,the AEC data in the historical period without DPV is decoupled from the NEC data in the period with DPV to separate the two components of AEC and DPVPG,and the monthly forecasting models for AEC and DPVG are established separately.For non-solar customers,a AEC forecasting model based on the time series method is established.Finally,the forecasting results of DPVG of solar customers are then subtracted from the forecasting results of AEC of all customers to obtain the forecasting results of NEC of all customers.The simulation results prove that the forecasting accuracy of the method proposed in this paper is higher than that of the direct prediction method under the high penetration of DPV scenario.