A Hierarchical Decomposition Approach for the Optimal Design of a District Cooling System

Abstract

A district cooling system is a centralized cooling supply system providing air conditioning to a set of buildings located in the same district. Designing and sizing such a system is very complex, as both the initial construction cost and the operation cost of the cooling system during its entire life must be considered. We first propose a modeling approach aiming at formulating this combinatorial optimization problem as a mixed-integer linear program of tractable size. We then extend a previously published hierarchical decomposition technique in order to find the optimal solution in an efficient way. Finally, we provide preliminary computational results based on a real-life case study located in China. INTRODUCTIONA district cooling system (DCS) is a centralized cooling supply system. It consumes electricity to cool down water and distributes it through an underground pipe network to the buildings in the district to provide them with air conditioning. DCSs usually are highly energy-efficient cooling systems. Thus, according to the Electrical and Mechanical Services Department of Hong Kong (EMSD, 2020), using DCSs instead of traditional air-cooled air-conditioning systems results in energy savings of 35%. DCSs also compare well with individual water-cooled air-conditioning systems using cooling towers as the energy savings may be up to 1 20%. Furthermore, this lower energy consumption leads to lower greenhouse gas emissions, which helps reduce the environmental impact of the system.Designing a DCS involves choosing the type and number of chillers to be installed as well as the ice storage capacity. These decisions should take into account not only the construction costs, but also the operation costs of the system during its whole lifetime. Computing these operation costs is a challenging problem. Namely, the demand for cooling power is highly variable and features a daily, weekly and yearly seasonality together with random variations. Moreover, due to technical reasons owing to the chillers, these operations costs are not at all proportional to the produced cooling power. Thus, in order to accurately estimate them, a detailed schedule describing, on a hourly basis, the on/off status and the load allocation of each chiller should be built for an horizon spanning a whole year. Furthermore, the deployment of a district cooling system is usually not a one-shot decision but rather a process in which investment decisions are made step by step, following the development of the district and the upward trend of the average demand over the years. This implies that a multi-phase strategic deployment plan should be built.This optimization problem can be formulated as a mixed-integer program. However, its resolution poses several difficulties. The first one comes from the non-linearity of the chillers' performance curves. These performance curves give, for each chiller, the amount of electricity consumed as a function of the amount of produced cooling power and thus play a key role in the estimation of the system operation cos...