Research On Risk Prediction Of Active Distribution Network Based On Situational Awareness

As distributed energy resources inundate distribution networks in vast quantities,the power supply structure of these networks is transformed,rendering traditional grid management and fault prediction methods inadequate to fulfill the evolving demands of active distribution networks and directly impacting the reliability of power supply.By identifying weak links and potential risks in the operation of active distribution networks ahead of time and enhancing the predictability of system risks,the safe and stable operation of power systems can be ensured while providing technical support for the risk development trends of active distribution networks.However,active distribution networks feature a plethora of devices,complex topological structures,and strong randomness and weak causality of failure factors,making it challenging for conventional methods to accurately predict grid risks.Consequently,this paper,grounded in the concept of situational awareness,primarily concentrates on unveiling the inherent connections between failure risk factors in active distribution networks and investigating specific risk prediction algorithms.The main work completed in this paper includes:(1)A framework for actively predicting the risks in the distribution network based on situational awareness was established.In response to the inadequacy of existing models for actively predicting risks in distribution networks,this paper analyzed the basic principles of situational awareness and the indicators for actively predicting risks in distribution networks,and proposed a model that integrates situational awareness,situational understanding,and situational prediction for actively predicting risks in the distribution network system.Additionally,this paper provided ideas for the work of each of these three components.(2)A fault risk data situation understanding method for active distribution network based on feature selection is proposed.Based on the Variance Threshold-GARFECV algorithm,this study selects,eliminates,and evaluates the active distribution network fault features.It ultimately identifies the optimal subset of features that are strongly correlated with active distribution network fault risk.This approach enables the risk prediction model to determine reasonable input variables and enhance prediction accuracy.(3)A method for predicting the risk level of active distribution network failures based on situational prediction has been proposed.The study uses the PSO-SVM algorithm to construct a situational prediction model for active distribution network failure risk.In order to address the premature convergence and local optimal solutions of the PSO algorithm,the study introduces the concept of SA to enhance the algorithm’s exploration capability.A linear iteration strategy is used to control the learning factor,and a hyperbolic tangent function is used to control the weight coefficient to avoid premature convergence.This results in the ASAPSO-SVM failure risk prediction model.The optimal feature subset is verified through practical examples,and the prediction accuracy of this method is as high as 98.43%,which can meet the practical needs of active distribution network risk prediction and provide reliable predictive support for the safe operation of active distribution networks.(4)A situational awareness-based active power distribution network risk early warning prototype system has been constructed.By designing the architecture of the risk early warning system,adopting the concept of modular design,the early warning system is divided into data acquisition terminal module,data transmission module,and platform visualization module.In accordance with relevant laws,regulations,and technical standards in China’s power industry,the ASAPSO-SVM prediction model is comprehensively integrated to ensure the safety,reliability,flexibility,and rationality of the entire system.Functional design and technical analysis are carried out for the three modules,ultimately resulting in a proactive distribution network risk warning system that integrates data collection,model prediction,and situational visualization.