Large-scale Unit Commitment Algorithm Research Based On Ordinal Optimization Theory

Unit Commitment problem belongs to areas of electricity economic dispatch and is the important part of optimal operation of power system. In the scheduling period, reasonable unit commitment scheme makes a big difference in security and stability of power grid, reduction of energy consumption, realization of energy-saving power generation and so on.For this status quo of multi-type complex power in Guangdong, in this paper, the weighted sum of coal consumption cost, power purchase cost and SO2 emissions cost is proposed as the objective function of unit commitment, subject to the constraints including time-coupled system operating constraints and unit features constraints. And this dynamic unit commitment optimization model proposed in this paper can deal with large-scale power system during the daily 96 periods, with all kinds of units considered, such as thermal, hydro, nuclear, biomass, gas units and purchased electricity. The total cost to society of the units generation is considered comprehensively in this model.From the mathematical point of view, Unit Commitment problem is ahigh-dimensional, non-convex with continuous-discrete variables mixed-integer nonlinear programming problems in typical NP-Hard qualities. It is difficult or impossible to obtain a global optimal solution of its theoretical. The multi-type complex power unit commitment problem model constructed in this paper is broken down into unit commitment scheme and economic dispatch these two sub-problems. Then ordinal optimization theory is for the first time put forward to solving unit commitment problem with generalized reduced gradient algorithm. Ordinal optimization theory is an effective tool to solve large complex optimization problems. With less computing time and higher probability to obtain good enough solutions, the complexity of optimization problems can be simplified by ordinal optimization theory. Under the ordinal optimization theory, firstly, the goal is softened, and good enough solution will be obtained instead of global optimal solution. Then dimensionality-reducing decoupling method and artificial neural network prediction method are proposed to rapid assess and filter a great number of unit commitment schemes. And the advantages and disadvantages in these two methods are analyzed. At last generalized reduced gradient algorithm is used to solve the units economic dispatch to achieve the complete solution of Unit Commitment problem.At last, this paper is under Matrix Laboratories R2010 b and General Algebraic Modeling System mixed programming environment, and based on the example of Guangdong province power system, the feasibility and practicality of ordinal optimization theory to solve large-scale unit commitment problem is verified by the comparison between ordinal optimization theory and, branch and bound method in mixed-integer nonlinear programming under the GAMS-BARON solver.