Research On Voltage Stability Prediction And Control Strategy Based On Improved LightGBM-SHAP

Under the situation of "three highs","two peaks" and the construction of new power system,the running state of the electric power system in China is gradually approaching the stable limit state,and the frequency of power failure accidents caused by voltage instability around the world is also gradually increasing,therefore,the problem of power grid stability is increasingly prominent,and the maintenance of voltage stability is the primary problem to be solved in the reliable operation of the power grid.Based on the strict requirements of real-time online application of power system,traditional methods can not fully meet the requirements of the current power grid,and the current artificial intelligence has done a lot of research and attempts in this aspect,but also provides a new idea for the realization of online assessment,prevention and correction of voltage stability.This paper proposes a static voltage stability strategy based on machine learning.The main research contents are as follows:Firstly,aiming at the static voltage stability of the power system,this paper solves the running data of the power system and the corresponding static voltage stability index L index under the operation mode of normal state and N-1 fault through the power flow calculation.On this basis,considering the characteristics of high-dimensional and massive power system data,machine learning-Light Gradient Boosting Machine(LightGBM)algorithm is used to predict L index of the system.In order to improve the performance of the prediction model,5 fold cross validation,Bayesian tuning,and Stacking fusion techniques are used to improve the model.Then,the performance test is carried out on the IEEE-57 and IEEE-118 node systems.The test results show that the proposed fusion BSLGBM prediction model is superior to other models in accuracy and speed by comparison with other machine learning algorithms.Secondly,considering that Shapley additive explanations(SHAP)algorithm has the ability of global and local analysis,it makes full use of its strong interpretability to systematically analyze the characteristic variables of the system.First,a visual quantitative analysis is carried out on the importance degree and direction of each characteristic variable of a single sample to L index.Second,rank the importance degree between each characteristic variable and L index in the whole power system.Through the analysis of IEEE-57 node system,the results show that SHAP algorithm can fully and effectively analyze the operating state and power system,and can provide a theoretical basis for the subsequent calibration system.Finally,based on the fast and accurate characteristics of the fusion BSLGBM prediction model and the interpretability of the SHAP model,a control strategy for online evaluation and correction of static voltage stability is proposed,which is the fusion BSLGBM-SHAP prediction and correction model.The BSLGBM prediction model can be used to quickly calculate the stability of the system voltage,and the SHAP model can be used to calculate the influence degree and direction of each system characteristic variable affecting the index L.Therefore,the BSLGBM prediction model is used to judge whether the correction scheme is reasonable,and the SHAP value of the SHAP model is used to correct the system.The feasibility of the proposed strategy is verified on the IEEE-57 node system,and the results show that the proposed strategy can provide a correction control scheme,which can stabilize the voltage of the power system within the safety margin.The above studies show that the fusion BSLGBM-SHAP prediction and correction model built in this paper is of great significance for realizing the online voltage evaluation of power system and improving the voltage stability of power system.