Hourly Optimization of a Cogeneration Plant

The thesis presents the development and application of an optimization model for the hourly management of a cogeneration plant (CHP) installed in a hospital facility. The study was carried out within EnergyNet during an internship period and aims to identify, for each hour of the day, the optimal operating load factor of the cogenerator that minimizes total energy costs while ensuring thermal comfort and operational reliability. The hospital’s trigeneration system simultaneously produces electricity, heat, and cooling energy. It is based on a Jenbacher JMS 320 GS-N.L engine powered by natural gas, with an electrical output of 1063 kWe and a total thermal output of approximately 1200 kW. The work focuses on designing a predictive–optimization tool developed in Python that integrates forecasting models, market data, and plant characteristics. The optimization operates on an hourly basis, consistent with the electricity tariff indexed to the Italian Day-Ahead Market price (PUN). Forecasts of electricity, heating, and cooling demand for the following day are obtained using artificial intelligence techniques: Meta’s Prophet model for electrical demand, and Random Forest regressors for thermal and cooling demands, which correlate energy consumption with external temperature trends. Meteorological data are automatically retrieved from OpenWeatherMap to ensure continuously updated predictions. The economic model incorporates all relevant cost components, including electricity purchase and sale prices, natural gas costs indexed to the PSV, fixed and variable charges, taxes, maintenance expenses, and potential incentives for High-Efficiency Cogeneration (CAR). Technical parameters, such as electrical and thermal efficiencies as functions of load factor and external temperature, are used to calculate hourly energy balances and total operating costs. The optimization algorithm evaluates all feasible load factor values (from 0.5 to 1.0, plus 0 for shutdown) and selects the one that minimizes total costs. The developed software automatically performs these calculations and exports the results to an Excel sheet shared daily with the hospital, indicating the optimal hourly load factor. The tool also provides a detailed breakdown of cost components, energy balances, and performance indicators, allowing both operational control and post-analysis of energy management strategies. Results show a strong correlation between the optimal load factor and the hourly trend of the PUN. When electricity prices rise, the cogenerator operates at full load to maximize on-site generation and potential export revenues; when prices fall, the system reduces its output, prioritizing self-consumption and avoiding uneconomic operation. The model identifies five main operating modes, ranging from full-load production with heat surplus to complete shutdown, depending on market conditions and energy demand levels. The adoption of the proposed optimization strategy significantly improves the economic performance of the hospital’s energy system, leading to a more efficient use of natural gas and a reduction in overall energy expenditure. In conclusion, the developed methodology demonstrates how data-driven optimization and predictive control can effectively enhance the management of cogeneration plants, ensuring both economic savings and greater sustainability in energy-intensive facilities such as hospitals.