Research On Forecasting Method Of Day Ahead Electricity Price In Electricity Market

With the promulgation of the electricity market opening policy,a free,equal and competitive electricity trading platform has gradually taken shape.Among them,the current electricity price as the reference price of various power transactions has become the key core for each main body of the power trading platform to grasp the market supply and demand relationship and formulate trading strategies.As a result,accurate forecasting of electricity prices in the past few days plays a vital role in maximizing the profits of power trading entities.However,due to the integration of a large amount of renewable energy into the grid and the formation of a competitive power market mechanism,it has brought unprecedented challenges to the forecast of electricity prices.To this end,this article starts from three aspects and studies the current electricity price forecasting method.Compared with mid-to-long-term electricity price forecasting,day-to-day electricity price forecasting is more demanding for the selection of historical day data sets.If historical daily electricity price data that does not match the predicted daily characteristics are mixed into the forecast data set,not only will the calculation cost be increased,but also the accuracy of the day-to-day electricity price forecast will be reduced.In view of the above problems,a method for selecting similar days based on comprehensive factors is proposed.First,analyze and determine the characteristics of electricity price influencing factors and give the calculation method of similarity of relevant electricity price influencing factors;then,design and improve the brainstorming algorithm to assign corresponding weights to the similarity of each influencing factor to realize the comprehensive factor calculation of historical days;finally,select Similar days with larger comprehensive impact factors form a forecast data set.Most of the traditional electricity price forecasting models focus on point forecasting and interval forecasting.This kind of forecasting result can neither show the fluctuation range of electricity price nor show the probability that electricity price may appear in price.Formulating detailed bidding schemes and planning corporate profits will have an adverse effect.To solve the above problems,a dynamic network quantile probability density prediction model(QGDFNN)is proposed.This model combines the quantile regression algorithm with the generalized dynamic fuzzy neural network,generates fuzzy rules through the input sample set to determine the model constraints,and designs a gradient-based nonlinear optimization algorithm to solve the model to obtain the predicted objects in different quantiles The conditional quantile on the point.Based on the QGDFNN model for the day-to-day electricity price forecast,because the output of the model is the forecasted daily hourly electricity price condition quantile,it is necessary to combine the kernel density function to estimate the electricity price condition quantile and obtain the predicted result.However,the traditional kernel density function chooses a fixed bandwidth,which will make the kernel density estimation result lack of local adaptability,which leads to the excessive smoothness of the current electricity price probability density prediction curve and reduces the prediction accuracy.To this end,an electricity price prediction method with improved nuclear density estimation is designed to improve the accuracy of electricity price prediction.First,define the activation level parameters,and implement variable bandwidth adjustment by calculating the activation level parameters;second,introduce the reliability index to limit the bandwidth of the kernel density function and complete the improvement of the kernel density function.The comparison results of multiple experiments show that the prediction accuracy of the previous day electricity price prediction method proposed in this paper is improved by 5.8%compared with other classic electricity price prediction methods,and the prediction time is shortened by 33.4%.