Particle Swarm Optimization for Optimal Operational Planning of a Cogeneration Plant

Recently, cogeneration systems (CGS) have been installed in various factories and buildings. CGS is usually connected to various facilities such as refrigerators, reservoirs, and cooling towers, and produces various energies for electric loads, air-conditioning loads, heating loads, and hot water loads (fig.1). Since daily load patterns of the various loads are different, optimal operational planning for CGS is a very important task for saving operational costs and reducing environmental loads. In order to generate optimal operational planning for CGS, various loads should be forecasted, and startup and shutdown status and input values for the facilities at each control interval should be determined using facility models (fig.2). Therefore, the optimal operational planning problem can be formulated as a mixed-integer linear problem (MILP) and mathematical programming techniques such as branch-and bound, decomposition method, and dynamic programming have been applied conventionally [9,14,15]. However, The facilities may have nonlinear input-output characteristic practically and operational rules, which cannot be expressed as a mathematical forms, should be considered in actual operation. Therefore, the problem should be formulated as a mixed-integer nonlinear problem (MINLP), and independent facility models should be developed for practical use and the method for solving the MINLP problem has been eagerly awaited. PSO is one of the evolutionary computation (EC) techniques [10]. The method is improved and applied to various problems [1-5,8,11,12]. PSO can be expanded to handle the whole MINLP by itself easily and naturally, and it is easy to apply to various problems compared with the conventional methods [7,13]. This paper proposes optimal operational planning for cogeneration system using particle swarm optimization. The proposed method is applied to typical cogeneration planning problems with