Research On Electricity Sales Forecast For Distribution Grid Development Planning Of DL Power Supply Company

Distribution grid is an important part of the electricity grid and is responsible for connecting the power generation side to the demand side.Distribution grid planning is the basis for distribution grid construction.Scientific and rational planning of the distribution grid can provide direction for the subsequent construction of the distribution grid and provide a theoretical basis for the safe and economic operation of the distribution grid.It is therefore important that power supply companies pay attention to the distribution grid development planning.The electricity sales forecast is the basic work of distribution grid planning and has an irreplaceable position in distribution grid planning.Its forecasting accuracy is related to the rationality and economy of planning,and is an important support for the operation and development of power supply companies,the formulation of regional grid development plans,the guarantee of rapid regional economic development and the maintenance of the quality of life of local residents.Therefore,this thesis takes DL Power Supply Company as the research object and conducts a study on electricity sales forecasting and distribution grid development planning,which will not only optimize and improve electricity sales forecasting and distribution grid planning techniques,but also provide strong support for the sustainable development of DL Power Supply Company.This is of great theoretical and practical significance.This thesis investigates the problem of electricity sales forecasting and distribution network development planning for DL Power Supply Company based on ARMA-LSTM.It introduces the general process of ARMA model in electricity sales forecasting and uses the model to forecast the electricity sales time series.It’s found that it can fit the training and test sets well,and the errors in the forecast are relatively small,demonstrating the effectiveness of the model.However,the model is unable to capture the characteristic factors affecting electricity sales.Therefore,based on the analysis of the characteristic factors affecting electricity sales,an LSTM forecasting model is constructed,using the MSE function as the cost function of the model and adding an L2 regularisation function to prevent over-fitting of the model,using the Adam optimization method to iteratively update the parameters,and inputting the feature set(season,average monthly temperature change,average monthly weather conditions)into the LSTM grid for modeling and forecasting.The errors in the forecast are all relatively small,verifying that the LSTM model is equally capable of making reasonable forecasts of electricity sales.However,compared to the ARMA forecast,the LSTM forecast is too flat and doesn’t match the growth trend of the observed values.Therefore,this thesis further constructs the ARMA-LSTM combination model for prediction and compares the forecasting results of ARMA,LSTM and ARMA-LSTM models.The results showes that the MAPE value of the forecasting results of ARMA-LSTM model on the training set of electricity sales is only 3.618%,which is better than the forecasting results of ARMA and LSTM alone,proving that the combined forecasting model has good generalization ability and can be better applied to electricity sales forecast.Based on electricity sales forecast,the ideas and objectives of the distribution network development plan are clarified,and the specific measures of the distribution network development plan in the DL region are proposed in combination with the planning requirements and principles,and the final expected results are given,so as to effectively increase the electricity sales revenue and improve the distribution network structure.To ensure the effective implementation of development planning,some reasonable safeguards and policy recommendations have been put forward.