Research On The Application Of Demand Response Based On Modeling Of Electrical Behavior And Parameter Identification

The effective resource interaction between the power supply side and the power generation side can be realized by the demand response to improve efficiency and reliability of system. Comparing with the traditional solution by adding peak-load regulation power, demand response will greatly reduce the pressure of investment power grid facilities. From now on, demand response projects based on incentive are mainly designed for large users, but as a big group of electricity users, the residents have great demand response potential which are much more suitable for demand response projects based on price. As a result, under the circumstance of smart power, it is meaningful to the application of demand response by analyzing the electrical behavior of residents and adjusting the behavior with price signal. From the monitor results of power usage of individual resident, multi-dimensional load classification system is constructed in this thesis to accomplish the classification and decomposition of load. And then the electrical behavior of multiple residents is analyzed to get different clustering with different price sensitivity. To the clustering of users, the real-time power market environment is constructed to achieve the dispatching by using the signal of price.Firstly, the necessity and feasibility of the participation in demand response of residents are analyzed and current status of the development and application of demand response at home and abroad are summarized. The research results of three aspects including price regulation, analysis of electrical behavior and model parameter identification are summarized as the foundation for the follow-up study of this thesis.Secondly, considering the shortcomings of current methods of load decomposition and the usage and characteristics of common equipment, multi-dimensional load classification system including function dimension, time dimension and power dimension is constructed for the individual user. Based on fuzzy membership function, the model of load decomposition is constructed to characterize the reliability of results of load decomposition by fuzzy membership degree.Then, the clustering model of electrical behavior is constructed to analyze the behavior of using electricity for multiple users by choosing power consumption proportion of different periods and load ratio as the power consumption characteristics for multiple users based on improved K-Means clustering algorithm under the circumstance of peak-valley price. According to the difference between the electricity consumption before and after the price change point, the target of demand response project which is sensitive based on price is determined by the model of screening users.Finally, the calculation model of demand response price is constructed for the aggregated users under the real-time power market environment based on consumer psychology by using support vector machine to realize model parameter identification along with the analysis of error and application scenarios. By comparing the error obtained by artificial neural network and regression analysis, these three parameter identification methods are combined to construct the integrated scheduling model.