Study On Improved Optimization Methods For Voltage Reactive Power Control

Voltage-reactive power control is of great importance for the safety, economic operation and satisfied power quality in power system. The optimization algorithm for voltage-reactive power, main subject on the reactive power optimization control, plays a significant role in the quality and efficiency of optimization process. Consequently, the optimization algorithm is studied as the primary content in this paper for the development of the real time reactive power optimization control system. In order to search the optimized solution quickly and reliably, a new global sequential quadratic programming (SQP) based on trust-region search algorithm is proposed for reactive power optimization. Combining trust-region search and SQP approach, this new method can obtain mathematic global optimized solution. It has not only high efficiency to deal with the restrictions, but also reliable convergence capability and high calculating precision. The inaccurate direction component decomposed method, with little demand on the initial solution, is used to compute trust-region sub-problems so that the consistent level of solution is raised gradually until satisfying the feasibility condition in the optimal point. Therefore, the infeasibility problem in the optimization of volt-reactive power can be solved effectively. For the character of variable in inequality constraints of reactive power optimization in distribution network, a new algorithm with boundary constraints for nonlinear optimization which uses a trust-region interior-point method is presented. This algorithm solves the boundary constraint directly to avoid growth of problem dimension during the optimizing search and improves the convergence rate. The trust-region approximate compute method based on truncated conjugate gradient method is used to calculate sub-problem in internal iterative and can solve the large-scale reactive power optimization problem efficiently. For the partial convergence and discrete variable problem, genetic algorithm is adopted, some parameters of which is modified to improve the searching performance and compute speed. For the random problem of optimized solution inherent in genetic algorithm, the control threshold is set, the solutions are evaluated and filtered, and calculating parameters are adjusted relatively to reset optimizing process. To improve search performance and to promote capability of global search, an improved hybrid strategy based on combining genetic algorithm (GA), simulated annealing (SA) and tabu search (TS) is introduced and improved. GA parallel search is used in the outer circulation iterative, temperature value is set in the interior circulation, SA operation is carried out on new species group and TA is introduced, to avoid invalid repeated search. For the reactive power optimization during whole dispatching period, a new model and simplified method considering cost and times of facility adjustment is proposed. Taking into account load condition of all the nodes in whole dispatching period, this approach separates compute process into two stages including pre-optimization, which considers the result of whole period, and real time optimization, which is used to real time optimal control. Active parameters of facility action are introduced into real time model and adjusting factors of control cost are set on various facilities in practical operation. To cooperate in the relationship between various objectives including operation network power loss of the whole system and facility control, and to improve the integrative control effect of the whole system, decision-making theory is introduced in the real time model for reactive power dispatch to distribute weights and good artificial assisting and real-time decision-making process is provided. Weight approach based on improved analytical hierarchy process and the penalty varying weight method based on local variable weight are proposed and used to multi objective combination. This weight method can obtain satisfied solutions in accordance with practical dispatch requirement for the power loss, action number and control cost in reactive power optimization. The theory research and practice operation show that the optimization algorithms, models and improved methods proposed have better innovation and practical value for voltage reactive power optimization control in power system.