Application Of BP Neural Network And Genetic Algorithm On Electronic Load

In practical application scene, electronic load, which is based on the fundamentals of PWM voltage source rectifier(VSR), is often used to simulate high-power, nonline and time-variable load in multiple environment such as the unknown parameters of the controlled system and the stochastic disturbance which can not be ignored. In the situation, it is quite difficult to tune the three parameters k_p,k_i,k_d of the PID controller and reach the desired control effect.To solve the problem, BP neural network and Genetic Algorithms(GA) is adopted in this paper to adjust k_p,k_i,k_d adaptively so that the electronic load can meet the need on the spot quickly.Error-back-propagation Neural Network has strong nonlinear approximation and study ability, and its structure and study algorithm are simple and clear. So, the indirect self-tuning control strategy, which based on the combination of BP neural network and PID control, can automatically rectify the controller parameters to achieve the best combination of PID control and make the system perform better.Meanwhile, to prevent the improper selection of the neural network weights which would cause the BP NN's premature convergence sometimes the unable convergence and affect the controller's performance finally, Genetic Algorithms(GA) is applied to in this paper to optimize the initial weights of the BP NN. Lots of simulational results show that after BP Neural Network and Genetic Algorithms are applied to Electronic Load, its stability and rapid has been significantly improved, and the steady-state error is greatly reduced.