An Improved ANFIS Control For VSC-HVDC

VSC-HVDC means the high voltage direct current transmission based on voltage souerce converters. It is widely utilized in power transmission and distribution because of the two advantages. The first is having an economic capacity of several MW, and the second is the quality of flexible control with two control variables. Along with the improvement of power electrionics, many shortcomings of VSC-HVDC will be correctified gradually,as superfluous energy dissipation and scant capacity.And some of novel development orientations,like power regional interconnection, clean energy grid connection and microgrid technology, will trend to accept VSC-HVDC as the first scheme. VSC-HVDC is a nonlinear coupling system, but controlled by linear PI controller in practical engineering, which has problems of tuning parameters, excess number of subcontrollers, and so on. Adaptive Neuro-Fuzzy Inference System(ANFIS) composed of fuzzy theory and neural network theory,which are main research emphasies in nonlinear control field, becomes a new focus. This paper presents a novel improved ANFIS controller for VSC-HVDC, with the function of fulfilling direct power control method.Firstly, an analysis about the nonlinear characteristic of VSC-HVDC and PI controller principle is completed in this paper. The reason of that PI control can be used for nonlinear systems in some degree is also studied from stability theory. Some brief introduction of fuzzy theory, neural networks theory and Particle Swam Optimizer(PSO) comes after the above analysis.ANFIS theory is emphatically introduced here.Then, according to the restrictive factors of ANFIS control performance:the structure of fuzzy neural networks and the back propagation(BP) learning algorithm, two corresponding improvement method are presented and studied in this paper. To ameliorate the structure of ANFIS,adding PID processing modules to the input layer,so as to form the ANFIS controller with PID functions(PIDANFIS);then an algorithm applicable to high dimensional optimizations and based on treble cooperative strategy is presented,which has the name of treble cooperative particile swam optimizer(TCPSO). Simulation of these new innovations is respectively performed.The results show: the PIDANFIS controller has a superior performance over conventional ANFIS for typical nonlinear systems; and the TCPSO algorithm also has a overwhelming performance than traditional algorithms for high dimensional optimizations.Finally, through the reference of the fusion method in which intelligent algorithms and neural networks are merged to perform better,TCPSO is utilized to optimize the process of PIDANFIS offline learning,in order to get a set of network parameters surpassing BP algorithm.By inserting this set of parameters to PIDANFIS, a novel ANFIS controller based on TCPSO is created. This new controller’ module established by S-function of Matlab will be embedded in VSC-HVDC Simulink model, together with PI control as a comparision.And the simulation results prove that the novel controller is able to perform the function of direct power control(DPC) and better than PI control.