Research On Harmonics Analysis Virtual Instrument Based On Advanced Neural Network

With the steady progress of industrial and agricultural modernization in ourcountry, power conversion device is widely applied in all the departments of thenational economy. However, they cause harmonic current in power system in therealization of power control and processing, which seriously reduce the voltagestability at public access point. Neural network has obtained the good applicationeffect in harmonic component identification with its outstanding nonlinear fittingability and adaptive characteristics; while ant colony algorithm has the advantage ofautomatic learning solution space distribution and free for control center’s guidance,and it can reach the global optimal solution at last. This paper comes up with a neuralnetwork harmonic analysis method optimized by ant colony, which is able to analyzeharmonic component information when combined with virtual instrument technology.Firstly, BP neural network is chosen as the neural network type in this paperthrough comparative analysis, and it’s studied from two aspects: network architectureand mathematical representations of learning algorithm. Then this paper points out theadvantages and disadvantages of BP algorithm. Against disadvantages of slowlearning process, momentum item is joined in to improve the training speed, and themethod is validated to reduce the network convergence time effectively throughsimulation. In-depth analysis indicates that the method failed to avoid network outputinto local extremum, and further optimize research direction is given.Secondly, this paper presents the characteristics of ant colony algorithm andcompares the similarities and differences between artificial and natural ants, andestablishes classical ant colony algorithm model in solving traveling salesmanproblem. Aimed at excessive search and choice question, an improved ant colonyalgorithm is proposed based on elitist ants and optimal ordering strategy. Number ofelitist ants is changed to study different control parameters’ influence on the algorithmconvergence and rapidity, which verifies that the improved ant colony algorithm willsolve combinatorial optimization problems more efficiently. Finally neural networkmethod improved by ant colony is applied to harmonic analysis, and outcome showssignificant enhance in its convergence speed, local extremum avoidance and saturatedstag leap. Finally harmonic analysis virtual instrument with login window, ant colonyalgorithm research function, harmonic analysis and data archiving function isdesigned in this paper. Then single-phase current harmonic components in controlledrectifier experiment is analysed and three methods mentioned above are compared,which points out the superiority of ant colony combined with BP algorithm inquickness and local minima avoidance.