Research Of Intelligent Power Demand Response Based On The Source-Grid-Load Situation

The risk of massive power shortage and cascading failures increases with the reality of large scale renewable energy integration,long-distance high-capacity power transmission and high percentage of power electronic devices.In order to maintain power balance in power system,it is necessary to make full use of the supporting capacity of resources in source,grid and load side,respectively.The situation awareness of source,grid and load is beneficial to formulate effective control strategy and to realize the power flow optimization dispatch and control in the whole scale of power system.Especially with the development of communication technology,it becomes an important research area that how to obtain the demand response(DR)potential by situation awareness of load side and further propose the intelligent power DR technology based on the source-grid-load situation.Funded by the National Natural Science Foundation(Theories and Methods for Load Preventive and Emergency Control of Cyber Physical Power Systems,No.51577030),the Projects of International Cooperation and Exchanges National Natural Science Foundation(Prediction Theory and Technology of Intermittent Renewable Energy Based on Trajectory Characteristics and Pattern Classification,No.51561145011),and the National Key Research and Development Program(Basic Theory and Method of Cyber Power Physical System Analysis and Control,No.2017YFB0903000),focused on the demand response issues in the new development situation in power system,this paper presents the performance of an in-depth study of research on intelligent power DR based on the source-grid-load situation.The main content and innovative contributions of this dissertation are as follows:The development characteristics and technological difficulties of situations are analyzed for the source side,grid side and load side respectively.To solve the problem in power prediction caused by the lack of observability of distributed photovoltaic(DPV)on the source side,an ultra-short-term master-slave forecast method for DPV generation based on the incomplete information is proposed: the numerical fitting for the spatial correlation relationship between target DPV and reference DPV is realized based on K-means clustering and least squares support vector machine(LS-SVM),and the impact of time delay characteristics on the corresponding spatial mapping relationship is studied,hence realizing the short-term generation forecast for target DPV plants.To solve the problem of quantitative evaluation of the effect of power flow control method with the perspective of grid global information,the notion of load handling capacity is brought out to quantify the power flow control situation base on global information.To solve the problem of whole-scale situation awareness load side based on partial information,firstly k-means clustering method is applied to cluster customers' daily load curves;secondly some typical loads are selected to provide precise measurements,and on that basis the whole-scale load composition estimation technique is proposed based on mean estimate.The two key issues to be solved in the application of intelligent DR technology are control framework and DR potential quantitative evaluation.Based on the analysis of dynamic operation characteristics of air conditioner,water heater and electric vehicle,the aggregate model of smart home appliance is built that satisfies the group distribution characteristics.Considering the time-varying feature of DR potential with the changing operating status of loads,the mathematic model and aggregate model of DR potential are built for smart appliances,according to the relationship between response potential and operating condition.A bi-level coordinated optimization strategy is proposed considering consumer' comfort and the timevarying characteristics of DR potential,realizing power balance with distributed DR through load aggregators using the multi-agent technology.The possible net power fluctuation caused by prediction errors of load and wind power generation is analyzed,and the technological principle to consume net power fluctuation with coordinated control of DR and AGC is explained.With overall consideration of generator ramping rate constraints,power flow limitation and smart appliance comfort constraints,the real-time coordinated dispatch model of source-grid-load is established,with the objective to maximize wind power consumption,and the model solution strategy considering DR is proposed.The presented coordinated method can make full use of AGC regulating capability on the source side and DR potential on the load side,as well as improve wind power consumption ability,by utilizing real-time source-grid-load control.The dissertation shows the difference of prediction accuracy of wind power generation and DR capability in different time scales,and the uncertainties in customer participation in DR process.On that basis,a framework for multi-time-scale source-grid-load coordinated control technology is proposed,covering the time scale of day-ahead,intraday and real-time control,and considering user regret and different DR characteristics.Corresponding to the different time scale of load participation in DR,the source-grid-load coordinated models are built for day-ahead,intraday and real-time scheduling for different DR loads.Considering the timevariant nature of customer behavior and different effects during DR process,the influence of DR and grid constraints on wind power consumption and economic costs in all time scales is analyzed.