Minimizing Gas-Fired Power Plant Operating Cost and CO2 Emissions With Optimized Scheduling of Gas Turbines and Batteries via Mixed Integer Programming

Abstract Power generation plants are integral parts of modern society. While the world works through the energy transition, the energy trilemma must be considered with each step forward. Maintaining a resilient power supply, accessible to all, is highly important. Gas turbine power plants offer industry-leading reliability and availability with around half the typical emissions output from a coal-fired plant. As global energy demand continues to grow, there are additional opportunities to reduce emissions by utilizing batteries for transient peak shaving, allowing for lower spinning reserve requirements. In certain installations, batteries can also be discharged to mitigate the need to turn on additional gas turbines and can be recharged during off-peak times when the energy supply is abundant and generated by renewable sources. In this paper, it is investigated how gas turbines and batteries can be combined through optimized load scheduling at a behind-the-meter power plant to reduce plant operating cost and CO2 emissions based on observed data in Poland. Mixed integer programming model was developed to optimize the scheduling of the power supply mix for a given demand while imposing constraints that reflect the reality of the operation of the system. The model accounts for gas turbine fuel flow, power, and efficiency by using high-fidelity thermodynamic models combined with near real-time data as well as day-ahead demand forecasts. These models and data are used to parameterize the objective and constraints of the mixed integer program. To demonstrate the effectiveness of the model, a method is applied to a real industrial power plant’s historical data and show the potential cost savings as compared with importing all needed power from the local grid. A comparison of historical operational data with an optimized solution shows the optimized solution can minimize operating costs while maintaining historical system reliability. The paper ends with a discussion of details around how such a system can be implemented.