Dynamic Load Restoration Method Of Active Distribution Network

In recent years,the large-scale outages caused by extreme weather disasters have become more frequent,resulting in great political influence and economic losses,and even threatening the normal conduct of social life and the personal safety of the people.Taking the advantages of the flexible,clean and efficient local distributed generations to ensure the continuous service of critical loads under extreme outages is an effective means to improve the resilience of the distribution system.Currently,the research on the use of distributed generations for distribution system recovery at home and abroad is still in its infancy,and a set of recognized effective distribution system optimization decision-making methods has not been formed yet.In this paper,the dynamic fault restoration method of the active distribution system considering uncertainty of intermittent energies is studied.The main work and research results of this paper include:Firstly,the characteristics of extreme outage scenarios are analyzed,the recovery process of resilient systems under extreme outages is summarized and the existing configuration and characteristics of active distribution systems are discussed.Thus,the dynamic load restoration strategy aiming at improving the resilience of the distribution system is determined as the research content.Considering the uncertainty of intermittent energies,a multi-period load restoration strategy based on rolling plan is proposed.The output model of renewable energies based on real-time information is updated continuously,and the dynamic correcting load restoration scheme based on the single-period risk-limiting load restoration model is proposed as the research idea.Secondly,a unified modeling method for intermittent energy output uncertainty is proposed against the stochastic volatility of renewable energy resources.According to the modeling requirement analysis of historical data and the research of traditional GMM and EM algorithms,the traditional models and algorithms are improved.Hence,the WD-GMM and WD-EM algorithm which can meet the requirements of large-scale high-dimensional weighted historical data modeling are obtained.Based on the improved algorithm,the prior distribution of WT and PV history output prediction errors verifies the effectiveness of the improved model and algorithm.Thirdly,in order to deal with the renewable energy outputs which are difficult to predict after extreme disasters,the traditional EM algorithm is also improved based on the maximum a posteriori estimation.The historical prior information and the continuously acquired real-time measurement information are combined to update the GMM parameters online.As a result,a self-learning method for intermittent energy output distribution parameters is obtained based on improved SF-EM algorithm.On the basis of the model parameter self-learning algorithm,the real-time updated intermittent energy output prediction error uncertainty model is established,realizing real-time response to real-time information changes and improving the accuracy of uncertainty descriptions after extreme disasters.At last,according to the opportunity constrained programming,a multi-period load restoration optimization model is proposed to improve the resilience of the distribution network,considering the constraints of power energy balance and distributed device characteristics.In order to solve the established PCP model,three different risk index calculation methods are proposed,which transforms the PCP model into a easier-solved MILP model.Based on the multi-period load restoration optimization model of the active distribution system considering different risk-limiting levels,the improvement of resilient of the proposed load restoration strategy is verified from three scenarios.And the advantages and disadvantages of the three optimization scenarios are analyzed from the perspectives of economy,safety and reliability through different evaluation indicators.