Research And Implementation Of Demand Response Game Strategy Based On Big Data Load Forecasting

With the development of modem power systems and smart grid,Electricity Information Collection Systems have been built successfully in many places,users'real-time power load data can be acquired.Since power cannot be stored in large quantities,power demand is dynamically changing.This requires load forecasting to gain how much power is needed in the future to balance supply and demand.Load forecasting has become one of the most challenging tasks in the smooth operation of the power industry.At the same time,load forecasting is the basis and premise of power system operation and development.On the basis of effective power load,peak load reduction can be realized by flexibly using intelligent demand response technology to achieve reasonable scheduling of power generation capacity.This makes the establishment of an effective load forecasting model and research on the interruptible load incentive pricing mechanism become a concern.The research history of power load forecasting is very rich,and various short-term power load models are very mature,and most of them have been put into operation.However,with the development of smart grids and the arrival of power big data,the forecast data has grown into massive data.Some of the original methods have not met the power demand in terms of processing speed and prediction accuracy.Therefore,a comprehensive load forecasting modela with higher accuracy to assist power companies in making decisionsto is needed.At the same time,due to the late start of domestic demand response,some provinces are still in the pilot and initial implementation stage of interruptible load.There is no unified and complete incentive pricing mechanism in China to regulate the transaction process between users and power companies.Therefore,it is urgent to adapt the incentive pricing mechanism of the existing power system to assist the electric power staff to implement the demand response..In this paper,the two parts of load forecasting and demand response incentive pricing mechanism are studied in depth:1.A short-term load forecasting model(Wavelet Decomposition-LST,WD-LSTM)combining influencing factors analysis,wavelet decomposition feature extraction,cubic exponential smoothing time series analysis and long-term and short-term memory networks is proposed.The model firstly uses wavelet decomposition to extract the main characteristics of power load change,analyzes its correlation with temperature,holidays and industry influence factors,and constructs corresponding temperature,holiday and industry adjustment factors.Secondly,it uses three indexes for each feature subsequence.The smoothing algorithm performs preliminary prediction separately,and the prediction result is fitted to each adjustment factor as the input of the long-term and short-term memory network.Finally,the prediction result is obtained by inverse wavelet transform.2.Based on the load forecasting results,in order to guide the user to adjust the power consumption mode,the incentive pricing mechanism in the interruptible load implementation is studied,and a user selection model based on multi-attribute sealed auction game bidding is proposed.In the single bidding model,the risk of the user's demand response is measured first,and then the user is classified based on the load characteristics.Based on this,a User Classification Selection algorithm(UCS)is proposed,which enables the user to be selected when selecting the user.At the same time,consider the power company's revenue and user risk,and enable the selected users to achieve balanced peak clipping.Aiming at the phenomenon of high degree of coincidence of selected users in multiple bidding,a fair mechanism based on points is proposed.Different types of users design different points increase rules,and consider user points when selecting users to form an increase point system.The user classification selection algorithm(S+UCS)improves the enthusiasm of small and medium-sized users.3.In-depth study of hadoop,spark big data processing platform,based on the research of load forecasting and interruptible load incentive pricing mechanism,combined with the understanding of power-related business,using springMVC architecture model to build electric intelligent decision management system,short-term load forecasting The results and requirements respond to the user selection model applied to the actual business,and the design and implementation of the system prototype is elaborated.Through the above research,this paper solves the load forecasting problem for massive electricity consumption data,verifies the accuracy of the model,and develops an incentive pricing mechanism,which can effectively guide users to participate in the imterruptible load response,and at the same time ensure the profit of the power company.Design and implement a power intelligent decision management system that combines load forecasting with demand response to make the results display more intuitive and easier to operate.