Power Grid Risk Assessment Algorithm Based On Improved Monte Carlo Method And LSSVM

With the continuous increase of grid interconnection scale,user load,and new energy grid-connected amount,the security and stable operation of the power grid has new challenges,and the accuracy and real-time requirements of the dispatching control methods have increased.The traditional method based on time domain simulation analysis is slower for the situation of large power grids and can no longer meet the current needs.With the construction of the grid information system,real-time data of the system and power components can be easily acquired.The new advances in machine learning methods and advances in computer technology provide new tools and make these real-time data practical applications for online security and stable operation.The concept of risk has both the probability and consequences of the event and is more suitable for describing the ability of the power system to reliably supply power.This article first studied the Monte Carlo method and improved it to obtain accurate and effective grid risk information.Lost load was selected as a risk indicator and Latin hypercube sampling techniques were used to generate a set of scenes to fully reflect the state distribution of the system.Perform repeated scene consolidation,and then perform risk consequential analysis on the scenario set,including three steps of topology analysis,line flow over-limit judgment,and interior point method optimal load shedding to obtain the risk consequences of each scenario.Finally,the number of statistical scenes and the corresponding consequences are obtained,and the overall risk value of the system is obtained.The use of machine learning can avoid the computational burden of traditional methods,transform the original nonlinear,high-dimensional problem into a simpler learning machine calculation model,thereby greatly improving the efficiency of risk assessment online applications.Among them,the least squares support vector machine(LSSVM)is one of the most commonly used and most effective methods.It has excellent generalization ability,and it can obtain much better results than other algorithms on a small sample training set.In this paper,the principle of the LSSVM and the parameter optimization method are studied.A LSSVM improved by Gaussian particle swarm optimization method based on is given.This paper proposes a fast algorithm for risk assessment of power system based on improved LSSVM.First,a data sample of grid risk is generated through the improved Monte Carlo method,and a common scene set method is used to avoid repeated calculation of scenes during sample generation.After the training is completed,the system online data can be used in the online risk analysis during the operation process to get the information such as the risk level of the grid timely,provide a reference for operators to maintain safe operation of the system from the aspect of system risk.Risk sensitivity reflects the influence of the changes of equipment parameters on grid risk.Using sensitivity indicators can effectively find equipment parameters that are closely related to grid risk,so as to find the key links that affect grid risk.In this paper,perturbation method is used to calculate the risk sensitivity of power system components.Combined with the risk assessment method based on improved LSSVM,the risk sensitivity calculation results can be obtained simply and quickly,and the equipment can be clustered and graded according to the risk sensitivity to obtain the maintenance level of the equipment.The application of the method in this paper can make the power system detect problems in real-time and make rapid response.It plays an important auxiliary decision role for the dispatching control of the safe and stable operation of the power grid.