Abstract
The operational strategy for district cooling systems (DCS) must be carefully designed and implemented to achieve real-time optimization. However, optimizing DCS operations is challenging due to the complex interplay of many nonlinear variables and practical constraints. Therefore, it’s essential to approach DCS optimization from a holistic perspective. This chapter introduces a real-time operational optimization framework for DCS, which integrates three key components: (i) cooling demand prediction and guidance, (ii) chiller system optimization, and (iii) cooling tower system optimization. The process begins with predicting cooling demand using artificial neural networks (NN). The predicted demand is then input into a mixed integer linear programming (MILP) model to determine the optimal chiller dispatch schedule and allocate cooling loads among the chillers and thermal energy storage tank for the next 24 hours. Following this, the chiller system and cooling tower are optimized every 30 minutes. Regression models are used to capture the operational characteristics of DCS components. These models are fed into an objective function aimed at minimizing energy consumption, with optimization performed using sequential quadratic programming. This methodology has been successfully implemented in three commercial district cooling plants, demonstrating the effectiveness of combining mathematical modeling, machine learning techniques, and optimization in practical DCS operations. Coupled with the optimal DCS design outlined in Chapter 3, this chapter provides readers with comprehensive knowledge on the design, operation, and optimization of DCS.