Study On Dynamic Demand Response Pricing Based On The Analysis And Prediction Of Residents Electricity Consumption Characteristics

With the expansion of wind and solar power generation,the issue of new energy dissipation has been increasingly prominent for the purpose of mitigating the occurrences of wind and solar abandonment.Consequently,against this background,how to crack such a dilemma has now become a major research topic among related fields.Considering that residents’ power consumption is on the increase annually over recent years and has formed a certain scale in the electricity market,furthermore,some high-power equipments such as air conditioners and thermal storage water heaters have a definite potential to be regulated.Nevertheless,as these power-using equipments are characterised by their large scale and dispersion,it is a large investment to adopt a centralised control approach and difficult to manage them in an administrative manner.For this reason,it is essential to undertake the regulation of residents’ electricity loads by means of tariffs in the electricity market,which has the advantages of being less difficult to manage and more effective to implement.One of the most critical issues is the analysis and forecasting of residents’ electricity consumption characteristics and their dynamic pricing on the basis of demand response characteristics.Obviously,a reasonable pricing scheme can achieve multiple management purposes such as balancing supply and demand on the power grid,increasing the utilisation of renewable energy,as well as savings in tariff costs for residents.To begin with,this thesis provides an overview of the foundations of the study of demand response dynamic pricing strategies.In particular,this thesis first introduces the basic concepts and classification of power demand response;analyses the types and operating characteristics of residents’ power equipments,as well as the patterns and characteristics of power consumption in different seasons,weekdays and non-working days;and also introduces and compares the commonly used methods for power load forecasting.Additionally,the thesis compares pricing methods within the electricity demand response environment and illustrates the advantages of reinforcement learning-based dynamic demand response pricing methods.Then,this thesis conducts a study on residents’ short-term load forecasting based on the minimal-redundancy-maximum-relevancy(m RMR)technique combined with gated recurrent unit(GRU).On the basis of the working principle of artificial neural networks,the improvement of the structure of recurrent neural networks together with gated recurrent unit,as well as the changes in computational performance resulting from the computational mechanism,are analysed;thereafter,by applying the m RMR technique,the input variables of the residents’ daily load forecasting model are screened,and a forecasting process based on m RMR-GRU for residents’ daily load forecasting is given;subsequently,a simulation analysis is conducted by adopting the electricity consumption data of the residents of a small town in Brussels,Belgium,which reveals that the forecasting method proposed in this study has higher accuracy and better applicability compared with single forecasting methods such as BP and SVM.Ultimately,this thesis presents and validates a dynamic pricing method for the residents’ demand response.Based on the relationship between the primary market participants in the retail electricity market,this study constructs a residents’ demand response dynamic pricing model that combines the residents’ cost of electricity and energy revenue with an optimal objective function,and takes into account various constraints such as customer satisfaction requirements and price difference restrictions between retail and wholesale prices.In the process of solving the model,the dynamic pricing problem of residents’ demand response is converted into a finite Markov decision process to achieve a superior solution with less computational expenditure,and the Q-learning algorithm is incorporated to achieve the solution of the dynamic pricing model.Lastly,the retail tariffs of the selected residents are priced through practical simulation examples.Besides,the reliability of the constructed model as well as the dynamic pricing method is verified in this thesis by increasing the number of customers,the time dimension and using CPLEX to calculate the same customer’s revenue and perform the simulation separately...