An Improved Bp For Short-circuit Current Prediction Based On Combinational Algorithm

With the continuous expansion of the scale of the grid,the load level and the number of power equipments in the grid continues to grow.Also,the structure of the grid is strengthening,which improved the power capacity of the main grid,and led to the short-circuit current level of hub site increased sharply.Based on the consideration of the safety,stability and economy of the large power grid,the short-circuit current level prediction of the key nodes and even all nodes in the power grid can provide guidance for reasonably balancing the resources of the short-circuit capacity and transmission capacity of the power grid and reducing the investment of the equipment,but also allow the planning and operation department of the entire grid to grasp the weakness of short circuit current,in order to take appropriate measures to provide timely reference to ensure the safe and stable operation of the power system.The traditional BP neural network has the disadvantages of slow training,sensitive to the initial weight and easy to fall into the local minimum point and difficult to determine the network structure.The network structure has a significant effect on the performance of the BP neural network,while there is no clear way to determine the structure.In order to predict the short-circuit current in the power grid,this paper proposed an improved BP for short-circuit current prediction based on CCGA and KPCA.Considering the synergistic relationship between network structure and connection weight,CCGA optimizes the network structure and initial weight of BP neural network.In view of the problem that CCGA is prone to premature convergence and slow convergence,this paper improves the CCGA performance by improving the coding mode and crossover operator,and using the adaptive crossover rate and mutation rate to optimize the BP neural network structure and the initial weight.The configuration plays a further role in optimizing the speed and accuracy of the prediction model.Based on the analysis of the influence factors of short-circuit current,the main factors of short-circuit current are taken as the input variables of BP neural network,and the power load,generator set,weather condition and date type are taken as the main factors.The input variables of BP neural network are reduced dimension by using KPCA,and the BP neural network structure is simplified to reduce the computational burden of BP neural network.The short-circuit current prediction model based on CCGA and KPCA optimization BP neural network is established by using the actual value of the calculated current based on the exact equivalent model calculation method as the output expectation value.According to the historical load data,the short circuit current prediction model proposed in this paper is simulated by MATLAB.There are three type of short circuit current prediction models used in this paper,which are the short-circuit current prediction models of the topology is determined by empirical formula,random and the combination of CCGA and KPCA.The three short-circuit current prediction models are compared with the same training input Training Results and Predictive Relative Errors in Sample Training.The simulation results showed that the short-circuit current prediction model based on CCGA and KPCA has good performance in training speed and prediction accuracy,and can provide decision-making reference for power grid operation.