The Research Of Transmission Line Audible Noise Prediction Model Based On Neural Network

Research on Audible Noise is a significant aspect of EHV and UHV transmission lines of electromagnetic environment impact. With the rising of transmission lines voltage rating, the research of Audible Noise prediction has become the hotspot in EHV and UHV transmission field. It is vital important to accurate prediction on Audible Noise which influences the choice of power transmission line's structure, the constructing cost cutting, the conveying voltage rating and environmental protection, etc. Present predictive methods predict the Audible Noise by influences factor such as potential gradient and lead structure. However, Audible Noise can still be influenced by environment and geography. The accuracy and generalization of the prediction will be directly affected by whether those factors are taken into account. How to accuratelypredict Audible Noise of transmission lines with conveying voltage grade, circuit structure, wires structure and environment factors, geographical factors is the key aspect of this thesis.This paper analyses the present Audible Noise prediction formulas, in view of the current formula only consider the potential gradient and wires structure, prediction deviation larger, the problem of Applicable scope limitation, according to the study of influence factors on transmission lines Audible Noise, a new view about Audible Noise prediction is proposed, we must consider comprehensively every influence factor which include conveying voltage grade, circuit structure, wires structure and environment factors, geographical factors, etc in Audible Noise prediction.BP neural network is used for Audible Noise prediction aimed at the multi-factor in Audible Noise prediction, the nonlinear combined influence of Audible Noise caused by multi-factor and the beyond description by regular formula problem in Audible Noise prediction. We use BP neural network algorithm which provide highly nonlinear fitting ability, flexible and effective training prediction ability to predict Audible Noise and carry out some simulation experiments so as to test the robustness of our algorithm, Well performance is achieved and the effectiveness of our algorithm based on neural network is proved.There are many factors that can influence Audible Noise, which makes the structure of the Audible Noise neural network prediction model complicated. Take this phenomenon into account, this thesis builds a PCA based Audible Noise neural network prediction model to predict Audible Noise so as to decrease the complexity of the prediction model and improve the accuracy of the model.A simulation experiment of Audible Noise prediction is carried out by Matlab software and the prediction results are analyzed.Finally, considering the real application problem of Audible Noise prediction, we use .net and Matlab computing software to achieve Web-based Audible Noise neural network prediction system. It manages the test data information, prediction model information and realizes the function of transmission line Audible Noise prediction.