A data-driven approach is utilized to model a chiller plant that has four chillers, four cooling towers, and two chilled water storage tanks. The chillers have varying energy efficiency. Since the chiller plant model derived from data-driven approach is nonlinear and non-convex, it is not practical to solve it by using the traditional gradient-based optimization algorithm. A two-level intelligent algorithm is developed to solve the model aiming at minimizing the total cost of the chilled water plant. The proposed algorithm can effectively search the optimum under the non-convex and nonlinear situation. A simulation case is conducted and the corresponding results are discussed.
- Chiller plant
- Data-driven model
- Neural network
- Two-level intelligent algorithm