Comparison of two large scale NLP strategies for determining optimal pump station controls

To solve the general nonlinear problem which is not simplified using the uniform spatial demand patterns, a two-level hierarchial scheme has been identified. The purpose of this decomposition is to reduce space of possible pump combinations at each time period. Thus, a DP technique may be then easily applied to find out an optimal combinations. The first stage of the algorithm is to solve a nonlinear programming problem (NLP) for the optimal tank trajectories. The decision variables in this problem are each pump station's discharge and pumped head for each period. The second level, a DP method is applied to find the best combination for each operation time period based on the optimal tank trajectories. Within the scope of this paper, the authors only focus on the first level, e.g., the NLP module. The real-time pump station operation problem has been posed as a large scale NLP which can be efficiently considered using a problem reduction technique. This formulation is discussed in detail. A discussion of two methods to solve the resulting NLP is presented and conclusions regarding their advantages are discussed.