Studies On Principle Analysis And Calculation Method Of Distribution Network Loss

With the enhancement of the economic sense of electric power enterprise, reducing the network loss became an increasingly concerned topic of power supply departments at various levels. Line loss calculation is the basic method to guide the loss reduction and save energy, but the calculation accuracy is subject to the cognition degree of structure parameters and operating data of the power grid. Since the distribution network loss shares the largest part of the whole loss, and the power distribution automation system in most areas in China is at low level, the collected operation data is limited. So proposing more precise calculation method has important application value and theoretical significance.This thesis firstly reviews the commonly used line loss calculation method of distribution network. Different methods have different applicable conditions, depending on the power grid structure or operation data. And with the accumulation of historical data and unceasing appearance of various kinds of data mining algorithm, the new calculation method which using the historical data for line loss regression has developed, but this method still has certain problems in engineering application which remain to be further discussed.In the section of transformer loss analysis, the comparison of the common transformer loss calculation method applied to living examples proves that shape coefficient can effectively improves the calculation accuracy. But the shape coefficient is not easy to calculate considering the lack of metering device while the load factor is easier to obtain. This paper analyze the effect of shape coefficient on transformers with different load factor, so that the load rate can be divided into different sections, and gives the approximate calculating method of shape coefficient in each interval through analysis and simulation. Example analysis proves the feasibility of this method.In view of the line loss, this paper applies support vector regression machine (SVR) for fitting line loss. To determine the independent variable of support vector machine (SVM), the thesis use factor analysis method to ascertain the influence of the multiple factors on the distribution network loss. But the precision is not high when the data samples are various. To solve the problem, the clustering method is introduced in group-dividing on the basis of predecessors’ clustering thought. The application of clustering algorithm divides distribution lines with similar structure into one subgroup, highlighting their characteristics. Example analysis shows that the effectiveness of the method.On the basis of above research, this paper further studies the way to optimize the parameters of SVM for improving the precision of fitting. Comparative analysis on the commonly used optimization method reveals that the particle swarm optimization (PSO) algorithm is superior. So that PSO is chosen as the basic optimization idea. In order to improve its optimization ability, this paper introduces the competition mechanism into PSO and put forward a hybrid optimum method which divides the particles into three types according to their evaluation index and uses different optimization strategy in different type. This search strategy balances the ability of both global optimization and local optimization. Commonly used optimization test function and application in line loss calculation proved the improved PSO had stronger optimal ability.All in all, subject to the limited operation data and distribution line information, this paper puts forward a more feasible and precise line loss calculation method on the basis of previous work.