For improving both accuracy and real-time of reactive power compensation of power system ofelectric arc furnace

the dynamic reactive power compensation method for the power supply system ofelectric arc furnace is studied deeply.A prediction method of RBF neutral network

which is optimized bythe PSO-GA

is proposed

the problem which is difficult to define by the parameters of RBF neutral networkprediction algorithm is solved with combination of the advantages of the PSO and GA algorithm and self-adaptive adjustment prediction parameters.The topological structure of dynamic reactive power of SVC andSVG hybrid compensation arc furnace is to predict the three-phase imbalance equivalent electricalsusceptance of SVC and perform dynamic dead beat reactive power compensation

and the cooperativecontrol method of SVC and SVG is designed.Both the feasibility and superiority of the hybrid reactive powercompensation method which is predicted on the basis of neutral network is verified by the simulation result.The simulation results verify the feasibility and superiority of the hybrid reactive power compensationmethod based on neural network prediction.