Research On Monthly Electricity Sales Forecast Based On State Grid Corporation Of China

Whether it is the State Grid Corporation's peer assessment system,or the power marketing in the electricity market environment,or the State Grid Corporation of China and its upstream and downstream companies that control the balance of profits and the distribution of human and material resources,the forecast of sales volume will all play a role.Very important role.At the same time,the power sales forecasting work plays a fundamental role in the overall power grid planning.The stable and sound operation of the economy and the reliable power supply of the power system all use it as a precondition,and to a certain extent,a reasonable power grid construction investment and operation is forecasted by the power sales forecasting accuracy.Decision.Therefore,to improve the forecast accuracy of monthly electricity sales,the State Grid Corporation of China and its upstream and downstream companies and other electric power companies attach great importance.However,there are quite a number of qualitative factors and quantifiable and non-quantifiable factors that affect power forecasting.It is difficult to find an accurate forecasting method that can ensure that it is suitable for any city or city and that it can be applied to any situation.Expected forecast results.Therefore,it is necessary to select a model that suits the needs of local development based on actual conditions.At present,there are many researches on this topic.The commonly used methods mainly focus on the time series method and the gray forecasting method.These methods are simpler and more mature than other traditional methods.For some cases with less influence factors,the prediction results are also good.However,these methods have limitations.In this paper,new attempts are made to integrate historical research findings and existing problems,and corresponding improvement predictions are made.First of all,this article uses the conventional gray model theory and improvement,and combines the multi-factor grey model theory of the grey relational degree theory to forecast the electricity sales.When using the grey forecasting model,more attention is paid to the characteristics of the data sequence itself,and the original data sequence must have a certain degree of smoothness,and the consideration of other internal and external influence factors is not comprehensive.For these issues,we conducted targeted improvement trials to find the best predictive model.The smoothness of the conventional gray model data sequence is processed to isolate the seasonal factors of the original data sequence,and then data smoothing is performed to eliminate a certain amount of random factors.Since the factors affecting the power sales are intricate and changeable,and the above method cannot be used to obtain better prediction results,we consider introducing the influencing factors into the prediction model,introducing as many influencing factors as possible,and then testing the significance of these factors.Factors that have a significant effect on entering the model.The data of the provinces and cities of the State Grid Corporation of China was used as an example to test,and finally,the electricity data available in Beijing was used for modeling and forecasting.Obtaining the model improves the prediction accuracy.Then the paper also uses the support vector machine prediction method.The parameters of the support vector machine are optimized,the best model optimization method is selected,and then the appropriate kernel function is selected to construct a support vector machine algorithm adapted to the kernel function.Combining with the actual data simulation prediction,the method has the effect of improving the prediction accuracy.Finally,this paper combines the grey forecasting and the support vector machine forecast,and uses the combined model to forecast the electricity sales.Since the focus of single model prediction is not the same,a new model is combined by combining these two methods.The different prediction results of combined forecasting methods will also be different.The optimal weight combinations are tried,and the variance-covariance optimization combination is combined with the actual data.Training,choose the combination method of the best prediction results in the region.