Dynamic optimization approach to coordinate industrial production and cogeneration operation under electricity price fluctuations

Cristian Pablos, Alejandro Merino Gómez, Luis Felipe Acebes Arconada, José Luis Pitarch, Lorenz T. BieglerComput. Chem. Eng 2021 · DOI · OpenReview

Introduction

Industrial facilities are among the world's largest energy consumers. Many of these plants, particularly in sectors like sugar production, utilize Combined Heat and Power (CHP) or cogeneration units to meet their thermal and electrical needs. While these units offer significant efficiency, they also present an opportunity for Demand Response (DR): adjusting production and energy generation in response to fluctuating electricity prices to minimize costs or maximize revenue from power sales.

However, a significant hurdle in optimizing these systems is industrial "inertia." Large-scale chemical and food processes cannot change states instantaneously; a sugar factory may take several hours to transition between different production levels. Traditional steady-state models often ignore these dynamics, leading to schedules that are either physically impossible to implement or economically suboptimal. This paper presents a dynamic-integrated optimization framework designed to synchronize industrial production with cogeneration while respecting both physical inertia and strict legal efficiency requirements.

System Architecture and Methodology

The research focuses on a Spanish sugar factory with a processing capacity of 370 to 430 tons of beet per hour. The system is divided into two primary blocks: the sugar production process and the cogeneration plant.

Schematic of the integrated sugar production and cogeneration system.

To manage this complexity, the authors employed a top-down dynamic-integrated approach. Rather than relying on computationally heavy first-principles models, they utilized "grey-box" models to represent the closed-loop dynamics of the plant. The optimization problem was formulated using Dynamic Algebraic Equations (DAEs) and solved via orthogonal collocation, which transforms differential equations into a large-scale nonlinear programming (NLP) problem.

Methodological flowchart for system study and optimization implementation.

Key technical considerations included:

  • Legal Constraints: Integration of the Primary Energy Savings (PES) index as defined by European Directive 2012/27/UE to ensure the plant maintains its "high-efficiency" status.
  • Complementarity Conditions: Mathematical techniques used to model discrete-like control logic (e.g., split-range valves) within a gradient-based optimization environment.
  • The Turnpike Property: Implementation of a "Turnpike Cost" to ensure the plant reaches an economically viable steady state at the end of the optimization horizon, preventing the solver from making short-sighted decisions.

Key Insights and Results

The methodology was tested against real-world market data from the Spanish Day-Ahead market (OMIE) and standard Time-of-Use (TOU) tariffs. The results demonstrate that modeling inertia is not just a constraint but an opportunity for more stable and profitable operation.

  • Cost Reductions: The dynamic approach achieved energy cost savings of 1.39% under TOU tariffs and 3.96% in the Day-Ahead market. When production flexibility was increased by relaxing storage constraints, savings reached 5.41%.
  • Energy Efficiency: The optimizer reduced natural gas consumption by an average of 0.54% by managing the CHP unit more intelligently relative to the process steam demand.
  • Operational Feasibility: Despite the complexity of the model (over 25,000 variables), the solution time averaged 123 seconds. This speed is critical for industrial applications where operational plans must be submitted to energy markets daily before strict deadlines.

Conclusion

By bridging the gap between high-level production scheduling and low-level process dynamics, this framework allows industrial plants to navigate the "legal edge" of energy efficiency while maximizing profit in volatile electricity markets. Future work aims to incorporate stochastic optimization to better handle uncertainties in weather-dependent beet arrivals and electricity price forecasts, as well as extending the model to include discrete "on/off" decisions for individual turbines and boilers.

Summary updated

Abstract

Industrial processes working with cogeneration utilities can coordinate their operation to take advantage of Demand Response programs to reduce their cost, while making the grid more stable and secure. Simplified steady-state and transition models are usually employed in a mathematical scheduling fashion to deal with this problem in process plants with small inertia. Here, however, we propose the use of a dynamic-integrated optimization approach that considers the dominant process dynamics jointly with the heat and power coupling in the cogeneration unit. The methodology meets the related European legislation and it has been tested in a simulated sugar factory that has a cogeneration system with connection to the external grid. The operation under two tariffs (TOU and Spanish Day-Ahead prices) was compared to the traditional policy of maximum production and fixed electricity prices, showing that a reduction of up to 5.41% of the costs is possible for the considered case study.
BibTeX
@article{pablos2021dynamic,
  author = {Cristian Pablos and Alejandro Merino Gómez and Luis Felipe Acebes Arconada and José Luis Pitarch and Lorenz T. Biegler},
  title = {{Dynamic optimization approach to coordinate industrial production and cogeneration operation under electricity price fluctuations}},
  journal = {Comput. Chem. Eng.},
  year = {2021},
  doi = {10.1016/j.compchemeng.2021.107292},
}
Figure