Solution quality improvement in chiller loading optimization

In order to reduce greenhouse gas emission, we can energy-efficiently operate a multiple chiller system using optimization techniques. So far, various optimization techniques have been proposed to the optimal chiller loading problem. Most of those techniques are meta-heuristic algorithms such as genetic algorithm, simulated annealing, and particle swarm optimization. However, this study applied a gradient-based method, named generalized reduced gradient, and then obtains better results when compared with other approaches. When two additional approaches (hybridization between meta-heuristic algorithm and gradient-based algorithm; and reformulation of optimization structure by adding a binary variable which denotes chiller’s operating status) were introduced, generalized reduced gradient found even better solutions.

► Chiller loading problem is optimized by generalized reduced gradient (GRG) method.
► Results are compared with meta-heuristic algorithms such as genetic algorithm.
► Results are even enhanced by hybridizing meta-heuristic and gradient techniques.
► Results are even enhanced by modifying the optimization formulation.