Abstract:Efficient operation of combined cycle power plants (CCPP) involves solving mixed-integer nonlinear optimization problems, constrained by thermodynamic differential equations, thermal constraints, and temporal dependencies between variables. The optimization problem aims to maximize profit for day-ahead operation by minimizing short-term and long-term costs, and accounting for volatility in ambient conditions and energy price. In operational terms, the challenge lies in developing an efficient and robust optimization algorithm capable of determining optimal values for thousands of temporal variables. It must also have the capacity to evaluate a multitude of scenarios encompassing various weather and market conditions. To address this challenge, this paper proposes a hybrid optimizer that utilizes deep learning models trained on historical operational data of a real CCPP, along with a surrogate physics-driven piecewise-linear model of CCPP, to determine the optimal values of thousands of control setpoints, with strong temporal co-dependencies, within a two-minute runtime required by industry constraints. A multi-resolution architecture is proposed to efficiently implement this optimizer. The algorithm sequentially solves the problem in coarse and fine resolutions to accelerate convergence. Final results are checked for feasibility using pre-trained deep learning models, ensuring accuracy and validity. The study shows that the proposed model provides comparable results to the exact solution while reducing computational costs, up to 95%. This research has practical implications for power plant system operators, as well as traders and dispatchers, enabling them to make optimal decisions for day-ahead contracting purposes and CCPP operation in a reasonable time, ultimately leading to more efficient and cost-effective power plant management. Furthermore, the methodology introduced in this work could potentially be extended to other industrial applications that require large-scale, complex optimizations, where the precise governing principles of a process can be approximated by piecewise-linear functions.

A Physics Informed Machine Learning Method for Power System Model Parameter Optimization

Global Optimal Polynomial Approximation for Parametric Problems in Power Systems

Gradient-Enhanced Physics-Informed Neural Networks for Power Systems Operational Support

Review of Learning-Assisted Power System Optimization

Application of Physics-Informed Machine Learning Techniques for Power Grid Parameter Estimation

Physics-Informed Neural Networks for Power Systems

Learning the Optimal Strategy of Power System Operation With Varying Renewable Generations

Generalized Parameter Estimation Method for Model-Based Real Time Optimization

Capturing Power System Dynamics by Physics-Informed Neural Networks and Optimization

Physics-informed machine learning with optimization-based guarantees: Applications to AC power flow

Comparsion of Two Kinds of Optimization Algorithm for Load Model Parameter Identification

Parameter Estimation of Power Electronic Converters With Physics-Informed Machine Learning

Computationally Efficient Electromagnetic Transient Power System Studies using Bayesian Optimization

Physics-inspired machine learning for power grid frequency modelling

On Physics-Informed Neural Network Control for Power Electronics

Regularised Learning with Selected Physics for Power System Dynamics

Physics-informed deep learning and linear programming for efficient optimization of combined cycle power plants

Online power system parameter estimation and optimal operation

Data-driven AC Optimal Power Flow with Physics-informed Learning and Calibrations

Neural Network Parameter Optimization for Model Predictive Direct Speed Control of PMSM

Transient Stability Analysis with Physics-Informed Neural Networks