Near Optimal Model Predictive Control of Thermal Energy Storage

Abstract

Efficient plant operation can be achieved by properly loading and sequencing available chillers to charge a thermal energy storage (TES) reservoir. TES charging sequences are often determined by heuristic rules that typically aim to reduce utility costs under time of use rates. However, such rules of thumb are in most cases far from optimal even for this task. Rigorous optimization, on the other hand, is computationally expensive and can be unreliable as well if not carefully implemented. Model-predictive control (MPC) that is reliable, as well as effective, in TES application must be developed. The goal is to develop an algorithm that can reach ∼80% of achievable energy efficiency and peak shifting capacity with very high reliability. A novel algorithm is developed to reliably achieve near optimal control for charging cool storage in chiller plants. Algorithm provides a constant COP (or cost per ton-hour) for 24-hr dispatch plan at which plant operates during most favorable weather conditions. Preliminary evaluation of this novel algorithm has indicated up to 6% improvement in plant annual operating cost relative to the same plant operating without TES. TOU rate used in both cases charges 7.4cents/kWh during off peak hours and 9.8cents/kWh during peak hours (Peak hours are 10 am to 10 pm).