Power Quality Comprehensive Evaluation In Electric Traction Systems

Electric traction load is a large capacity asymmetric load which is nonlinear, fluctuational and has a shock impact to the system. The power quality problems caused by it is quite particular. Researching these characteristics and their influence on power system is the basics of the power quality assessment of electric traction systems.According to the national standards and current detection level of power quality, this paper establishes a power quality evaluation index system for the power quality assessment of electric traction systems, and discusses ways to determine the weight. This assessment index system consists of6indexes of continuous power quality including voltage deviation, voltage fluctuation and flicker, harmonic voltage distortion rate, three-phase voltage unbalance and frequency deviation. This paper also discusses the index system’s further enrich direction. Determination of weight in comprehensive assessment system plays an important role. however, the subjective and objective weighting methods both have advantages and disadvantages, and some assessment methods which are not obviously have weight are also have weight problems, and these problems are all discussed in details in this paper.This paper proposes an electrical railway traction load power quality comprehensive assessment model based on subjective-objective combination weighting method. The model weight is determined comprehensively by AHP and entropy method, containing not only experiences and judgments, but also dealing with the characteristics of evaluation data. The comprehensive assessment result is more scientific, reasonable, and has a clear interpretation for it. Application examples prove the assessment model is simple, effective, and easy to program.This paper also proposes an electrical railway traction load power quality comprehensive assessment model based on artificial neural network. The training samples are generated by the corresponding national standards of power quality. This model is established without weight, and the trained artificial neural networks can figure out the objective comprehensive assessment according to the individual assessment index. Application examples prove that the proposed model is simple, strong learning ability, high solving efficiency, and can deal with large numbers of samples fast and accurately, so the model is proved feasible.