This research focuses on developing strategies for the optimal control of large-scale Combined Cooling, Heating and Power (CCHP) systems to meet electricity, heating, and cooling demands, and evaluating the cost savings potential associated with it. Optimal control of CCHP systems involves the determination of the mode of operation and set points to satisfy the specific energy requirements for each time period. It is very complex to effectively design optimal control strategies because of the stochastic behavior of energy loads and fuel prices, varying component designs and operational limitations, startup and shutdown events and many more. Also, for large-scale systems, the problem involves a large number of decision variables, both discrete and continuous, and numerous constraints along with the nonlinear performance characteristic curves of equipment. In general, the CCHP energy dispatch problem is intrinsically difficult to solve because of the non-convex, non-differentiable, multimodal and discontinuous nature of the optimization problem along with strong coupling to multiple energy components.
This work presents a solution methodology for optimizing the operation of a campus CCHP system using a detailed network energy flow model solved by a hybrid approach combining mixed-integer linear programming (MILP) and nonlinear programming (NLP) optimization techniques. In the first step, MILP optimization is applied to a plant model that includes linear models for all components and a penalty for turning on or off the boilers and steam chillers. The MILP step determines which components need to be turned on and their respective load needed to meet the campus energy demand for the chosen time period (short, medium or long term) with one-hour resolution. Based on the solution from MILP solver as a starting point, the NLP optimization determines the actual hourly state of operation of selected components based on their nonlinear performance characteristics. The optimal energy dispatch algorithm provides operational signals associated with resource allocation ensuring that the systems meet campus electricity, heating, and cooling demands. The chief benefits of this formulation are its ability to determine the optimal mix of equipment with on/off capabilities and penalties for startup and shutdown, consideration of cost from all auxiliary equipment and its applicability to large-scale energy systems with multiple heating, cooling and power generation units resulting in improved performance.
The case-study considered in this research work is the Wade Power Plant and the Northwest Chiller Plant (NWCP) located at the main campus of Purdue University in West Lafayette, Indiana, USA. The electricity, steam, and chilled water are produced through a CCHP system to meet the campus electricity, heating and cooling demands. The hybrid approach is validated with the plant measurements and then used with the assumption of perfect load forecasts to evaluate the economic benefits of optimal control subjected to different operational conditions and fuel prices. Example cost optimizations were performed for a 24-hour period with known cooling, heating, and electricity demand of Purdue’s main campus, and based on actual real-time prices (RTP) for purchasing electricity from utility. Three optimization cases were considered for analysis: MILP [no on/off switch penalty (SP)]; MILP [including on/off switch penalty (SP)] and NLP optimization. Around 3.5% cost savings is achievable with both MILP optimization cases while almost 10.7% cost savings is achieved using the hybrid MILP-NLP approach compared to the current plant operation. For the selected components from MILP optimization, NLP balances the equipment performance to operate at the state point where its efficiency is maximum while still meeting the demand. Using this hybrid approach, a high-quality global solution is determined when the linear model is feasible while still taking into account the nonlinear nature of the problem.
Simulations were extended for different seasons to examine the sensitivity of the optimization results to differences in electric, heating and cooling demand. All the optimization results suggest there are opportunities for potential cost savings across all seasons compared to the current operation of the power plant. For a large CCHP plant, this could mean significant savings for a year. The impact of choosing different time range is studied for MILP optimization because any changes in MILP outputs impact the solutions of NLP optimization. Sensitivity analysis of the optimized results to the cost of purchased electricity and natural gas were performed to illustrate the operational switch between steam and electric driven components, generation and purchasing of electricity, and usage of coal and natural gas boilers that occurs for optimal operation. Finally, a modular, generalizable, easy-to-configure optimization framework for the cost-optimal control of large-scale combined cooling, heating and power systems is developed and evaluated.