A Multi-core Parallel Algorithm For Medium-term Thermal Power Plants Starting Optimization And Its Application

The starting optimization of thermal power plants is vital to medium-term plan making of power grid, which refers primarily to the reasonable starting mode of all the power stations in power grid during the given computing period, and should take various constraints into the comprehensive consideration, such as electric power system load demand, operation parameters of power plants et al. The starting optimization of medium-term thermal power plants touches upon the unit commitment problem, which has the characteristics of high dimension, discrete and nonlinear, it's very difficult to find out a practical algorithm which can not only take various constraints into the comprehensive consideration, but also get ideal computing speed and optimal solution precision to solve this problem. The spring up and development of parallel processing hardware and software throws ample light on solving the thermal-unit commitment problem.Aiming at the multi-status, multi-stages characteristics in the decision-making optimization problem of medium-term thermal power plants starting, a multi-core parallel algorithm which explicitly relates the Fork/Join framework based on divide-and-conquer strategy is put forward The main contents of this paper are outlined below.(1)An optimal model taking equal capacity utilization hours of all thermal power plants as objective is established, and the detail solving process is given. Initial solution space which may meet the constraints of starting sets, load rate requirement is obtained first, then multiple initial feasible solution which meets the peak and valley duration constraint is selected by applying Heuristic Search to the Initial solution space for many times. Next, the Progressive Optimality Algorithm (POA) is employed to search the optimal solution, the global optimal solution is found out from multiple final solutions at last.(2)Feasibility Analysis for the optimization problem is conducted, and a multi-core parallel based implementation strategy is put forward. In light of the characteristics of Fork/Join framework, initial solution space is obtained parallelly according to the computing period, and the optimization problem parallel solved according to load rate for many times. The Heuristic Search is employed to initial feasible solution first, and then the POA is applied to search the optimal solution. The solving process is assigned to multi-core of the computers to compute parallelly by Fork/Join framework, and the optimal solution which may meet the actual demands of the projects is gained finally. (3)The multi-core parallel algorithm is applied to medium-term thermal power starting optimization system of Yunnan Power Grid. Taking the 27 thermal-units and 184 periods of Yunnan Power Grid as the basic data, the algorithm convergence is verified, and then speedup and efficiency of the algorithm are tested on the processor of single-core, dual-core,8 core,16 core respectively. The optimization result suggests that the multi-core parallel algorithm can make full use of the multi-core resources of the computers and significantly improve the computational efficiency and the quality of optimal solution.