Research On Distributed Generators Optimization Configurations Using Genetic Algorithm Based On The Extreme Learning Machine

With the development of social economy, the importance of energy and environmental issues become increasingly prominent. Due to its low carbon environmental protection, low investment and flexible power generation methods etc., distributed generators has been widely recognized and applied. However, a large number of wind power, photovoltaic power generation and other types of distributed generator are connected to the power grid, which brings more challenges to safety, reliability and economic operation of power system. Therefore, how to make planning of the distributed power system reasonably and efficiently becomes particularly significant.In this paper, the development and current situation of distributed generators are firstly demonstrated, and the interconnection of the main distributed generators is introduced in detail. In addition, the impact on the electric grid in the aspect of network loss, power quality, reliability and trend of distribution is systematically analyzed. The analysis results of the existing algorithm for distributed power source planning problem indicates that traditional algorithm extensively has the weakness of low speed and easy to fall into local optimum, etc. And Improved Genetic Algorithm based on the Extreme Learning Machine is proposed to solve this sort of problem. The algorithm applies a new single hidden layer feed forward neural grid algorithm—Extreme Learning Machine (ELM) to improve the fundamental Genetic algorithm. Meanwhile, it utilizes excellent ability of nonlinear mapping to simulate the evolutionary process of two generations of the population and combines with traditional genetic algorithm. Through the reasonable parameter settings, the global searching ability and rate of convergence is improved. In comprehensive consideration of environmental benefits and economic benefits, the Economic Planning Model which has the least investment and operation cost, the lowest expense of grid loss and the most environmental benefits is built. Aiming at selecting the candidate for node which has lacked in traditional Distributed Power Source Planning, the method in which calculating the Apparent Secondary Precise Torque Value of each node, sorting the value and choosing the optimal candidate installation node has been proposed.The simulation is performed in a 35-node system, the case analysis shows that the adopted algorithm is better than traditional Genetic Algorithm in the aspects of computational accuracy, rate of convergence and optimization ability..The reasonable and reliable distributed generation planning scheme, can be obtained. Moreover, it can verify that the dimensions of the variables can be greatly reduced with decreasing amount of calculation, and thus the efficiency of the algorithm can be enhanced computing Apparent Secondary Precise Torque Value of each node.