Control method for power quality compensation based on Levenberg-Marquardt optimized BP neural networks

Unified Power Quality Conditioner (UPQC) has the function of improving voltage supply, compensating load reactive power, suppressing harmonic current and increasing power factor; however, tradition control method has a certain extent limitation for such multiple input, multiple output, close coupling nonlinear issue. Artificial Neural Networks (ANN) can deal with data for multiple objective learning in parallel continuous way, the control of complex object is achieved through interactions between nerve cells. Levenberg-Marquardt algorithm optimized back propagation neural network has, the characteristic of efficient learning and faster convergence; ANN outputs control signals for voltage and current compensation to UPQC through weights training. Simulation model is built in Matlab, load which is three phase unbalanced and has badly distorted current is simulated under the case of voltage sag. Simulation experiment indicates its compensation effectiveness is much more satisfying than traditional control method.