A new coal dust particle recognition algorithm based on concave points extraction and ellipse fitting is proposed for the features of irregularities and particle overlap. The new algorithm includes contour processing and ellipse fitting in this paper. In the part of contour processing, the feature points are obtained with polygonal approximation on the edge of a binary dust particles image, and then concave points of overlapping particles are extracted by the method of angle combined with size, finally the edge is segmented by concave points. To solve the problem that direct least square ellipse fitting is easily affected by noise points, bare bones particle swarm optimization is introduced to find global optimum fitting parameters and the segmented edge is ellipse fitted. Experiment results show this proposed algorithm obtains better recognition performance.

This paper proposes the optimization method on reactive power compensation of distribution network. The objective is that total loss difference between before being compensated and after should get maximum. The method adopts back/forward sweep method for calculating power flow to obtain total active network loss. Then global neighborhood searching strategy and aspiration criterion for fixed parallel capacitor are established by tabu search algorithm. The optimal installation location and compensation capacity of fixed parallel capacitor are determined. Optimizing simulation on a 34 nodes distribution system is realized, and optimal planning scheme is obtained. Results show that the proposed method is feasible and effective in optimal planning on reactive power compensation of distribution network.