Carbon-Adjusted Dispatch Optimization for Princeton’s Campus Energy Plants

Abstract

Existing optimization algorithms for combined heating, cooling, and power (CHCP) cogeneration plants primarily take a mathematical approach to optimization, modeling plants as systems of thousands of variables and equations, which are then plugged into equation solvers. This requires immense computing power, leading existing models to exhibit one or both of two key limitations: (1) Oversimplifications of complex plant components and/or (2) Extremely long runtimes that make models unsuitable for practical use or analysis of extended optimization periods.

This thesis presents a novel mechanical approach to optimization that overcomes both these limitations by modeling a CHCP plant as a series of energy flows, rather than a large number of abstract equations. This mechanical approach was used to develop an optimization algorithm for Princeton University’s campus energy plants. The model is designed to answer key questions from plant leaders while factoring in the CO2 impacts of its decisions.

This model achieves exceptional results. Its immense efficiency relative to traditional mathematical approaches enables runtimes thousands of times faster than many literature examples, while simultaneously modeling plant dynamics far more accurately. Its modular design also enables it to flexibly test a wide variety of potential operating cases. Carbon-weighted optimization of the plant’s 2018-19 operations, split into hourly time steps, yields an average 6.5% reduction in CO2 emissions (equivalent to taking 1,085 cars off the road) while saving $1.70M/year in operating costs using only Princeton’s existing energy infrastructure. Once Princeton’s new TIGER heat pump plant comes online, these average savings improve to a 36.3% CO2 reduction (equivalent to 6,129 cars), alongside $2.07M/year in cost savings.