Research On Intelligent Retrospective Reasoning And Evaluation Technology For Key Indicators Of Power Grid Operation

Power grid dispatching shoulders the important responsibility of achieving safe,high-quality,and economical operation of the power grid.On the premise of ensuring the safe operation of the power grid,improving the lean level of power grid operation is an important task of current work.The current theoretical research results are mainly focused on the pre-planning and real-time control of power dispatching.However,with the development of power system,the operating characteristics are becoming increasingly complex and the power grid operation control becomes more difficult,there is a lack of effective post-evaluation means to objectively evaluate the operation of power grid.In addition,there is a lack of automated and intelligent evaluation methods for the change law of grid operation indicators,the location of abnormal indicators and the analysis of causes,and the feedback and improvement of the results of the post-assessment are not involved.Starting from the needs of post-operation evaluation of power dispatching,this paper studies post-dispatch evaluation techniques,retrospective reasoning techniques for abnormal indicators,and optimization and improvement techniques for evaluation results,aiming to provide support tools for quantitative evaluation and closed-loop feedback of power dispatch effectiveness,and promote the effective improvement of scheduling lean level.The main research contents of this paper are as follows:Firstly,the effectiveness of power grid dispatching operation is evaluated quantitatively.Based on safety,economy,energy saving,environmental protection and fairness,a set of post-operation evaluation index system of power grid dispatching is established and comprehensive evaluation of indexes is carried out.In this paper,the subjective and objective weighting algorithms are used to calculate the weights of all levels of indicators,and the weights are synthesized based on the principle of minimum information discrimination,and the weights of key indicators are revised.This set of methods can effectively improve the correctness of expert weighting decisions,reflect the difference and consistency of indicator data and the pertinence of key important indicators.The approach ideal method is used to comprehensively evaluate the primary and secondary indicators of typical days.This method can intuitively help dispatching operators to grasp the operation status of the previous day in time,and provide effective support for the subsequent study of specific abnormal indicators,as well as the discovery of weak links that limit the lean level of power grid dispatching and the study of improvement measures.Then,taking the regional power flow operation index as an example,the paper analyzes the abnormal index of power grid dispatching operation by tracing reasoning.In this paper,a candidate set of correlation factors for specific indicators and scenes is established.Pearson correlation coefficient method is used for correlation analysis,Granger causality analysis is used for causality analysis,and decision tree algorithm is used for correlation analysis for discrete correlation factor sequences.A complete Bayesian network knowledge expression is formed by using the data discretization algorithm based on k-means and the parameter learning algorithm based on maximum likelihood estimation The forward reasoning technology based on Bayesian network is used to analyze the index risk probability,the key risk factors based on probability importance and key importance index,and the risk causes based on backward reasoning technology.In an example,the control variable method is used to test the retroactive results,which shows that the proposed method can accurately find out the main causes of abnormal indicators.Finally,on the basis of the results of retroactive reasoning after abnormal regional power flow operation indicators,combined with index causality analysis,the traditional day-ahead scheduling model and strategy are improved.This paper analyzes the direction of optimizing and improving the traditional scheduling by the results of index traceability reasoning,and the importance of causal analysis between indicators.It uses the convergence cross mapping algorithm to explore the causal relationship between indicators,and further analyzes that the day-ahead scheduling strategy can be improved from the perspective of "improving standby availability".The measures to improve the availability of reserve are analyzed,and it is clear that the distributed scheduling model can promote the rational allocation of reserve,and the participation of flexible load can improve the reserve potential and reserve capacity of the system.Based on the extreme scenario method to deal with the fluctuation of wind turbine and load output,a distributed dispatching model of power grid with flexible load is constructed.Finally,the synchronous alternating direction multiplier method is used to solve the whole network distributed optimization problem alternately.The example shows that the proposed method can effectively alleviate the abnormal trend of the studied indicators,thus verifying the correctness of the analysis process in this paper.