Social Evolutionary Programming Based Unit Commitment

Unit commitment is an important aspect of optimal operation of power systems. Since it could mean significant annual financial savings in power generating cost, it continues to be the main optimization task in everyday power generation schedule. From the point of view of mathematics, it is a NP hard combinatorial optimization problem with many constraints and it is difficult to find the optimal solution in theory. Several solution techniques have been applied to this problem to find a good solution in a reasonable time. As one of these techniques, GA has been recently applied to solve the unit commitment problem and has made some remarkable achievements. Because its initial populations are generated at random, it can't guarantee feasible solutions because of the violation of constraints such as minimum up/down time constraint. In addition, in the processes of crossover and mutation the genetic operators are also operated randomly, it's difficult to guarantee satisfying minimum up/down times constraint. All these factors reduce GA's computational efficiency.This paper presents a new method—Social Evolutionary Programming to solve these problems. The new algorithm is based on a general Social Cognitive Model and has been successfully applied to the optimal planning of power distribution system. This paper first analyzes people's cognitive behavior in solving unit commitment problem, and then makes a detailed design in cognitive agents, cognitive regulation, "paradigm study and update" and "Breakthrough of existing paradigms" mechanisms. The minimum up/down time constraint is considered in cognitive regulation. Cognitive Agents imitate human decision-making behaviors to obtain the feasible solutions and the "paradigm study and update" mechanism can avoid the production of numerous infeasible solutions caused by crossover and mutation. Therefore the proposed algorithm has an advantage in the convergent stability and the computational efficiency. The proposed algorithm has been tested on systems of up to 100 units to demonstrate its advantages.