Session: 01-02: AI for Energy Sustainability II

Paper Number: 123705

123705 - AI for Energy Intensive Industry: A Hybrid Optimization Approach for Flexibility Service Providers 

Abstract: 

Today, our energy systems rely heavily on conventional power plants, such as gas turbines and combined heat and power generators when it comes to balancing generation and demand. With the advancing decarbonization of our power system, they will be replaced by highly unpredictable and inflexible renewables. This will inevitably lead to imbalances and congestions in the power grid. A sustainable energy system is therefore in dire need of flexibility. Promising candidates to provide the necessary flexibility to the power grid are companies in the energy intensive industry, since they represent some of the largest electricity consumers. They can act as Flexibility Service Providers (FSPs) and provide Demand Response (DR), which will play a crucial role in the energy transition.

FSP can provide flexibility by participating on energy markets, such as balancing or redispatch markets. Acting on these markets is far from trivial and requires some sort of strategy and planning. This paper presents a method based on a novel approach to formulate a bidding strategy for industrial energy systems to participate on these markets and help enable the energy transition.  

Bidding strategies are often formulated as two-stage or multi-stage stochastic optimization programs due to the sequential nature of energy markets. In the first stage, a bidding strategy is determined, which consists of bids, prices, and volumes. After the realization (market closure) bids are accepted or rejected. Based on this new information about accepted bids, a second optimization problem is solved, which can include (re-)scheduling of production processes or a Unit Commitment (UC) problem to optimize the operation of the plant. These problems are connected since the bidding strategy constrains the optimal operation of the production process.

The combinatoric nature of the first stage leads to high computation times, when solved with mathematical optimization such as Mixed Integer Linear Programming (MILP). Reducing the runtime to an acceptable size may only be achieved by oversimplifying of the original problem. Due to the advancements in AI, heuristic optimization techniques have proven to be a powerful tool when solving highly combinatorial problems in a reasonable amount of time. We therefore propose a hybrid approach to decompose the problem in the following way: The highly combinatorial first problem of choosing an optimal combination of bids is solved using common heuristics, while the following second UC problem of operating the plant while maintaining all constrains is solved using MILP.

We apply our proposed hybrid method to an industrial company from the energy intensive industry participating in the European Power Exchange (EPEX) and the Austrian balancing market, specifically focusing on automatic Frequency Restoration Reserve (aFRR). To demonstrate the method, a use case from the food industry is used, with on-site production of electrical and thermal power generation. A Combined Heat and Power (CHP) plant in combination with a thermal energy storage serves as a flexible asset that can provide DR. The second optimization consists of a UC problem of the optimal energy management and operation of the on-site energy supply.

We apply several heuristic optimization techniques and compare computation times and optimality of our hybrid approach.

A hybrid approach shows significant potential in decomposing large optimization problems that industrial FSP face when acting on sequential energy markets. This will lead to practical solutions for increasing flexibility and enabling industrial prosumers to support a sustainable energy system. 

Presenting Author: Martin Fischer TU Wien

Presenting Author Biography: Martin Fischer is working as a project assistant at the Institute of Energy Systems and Thermodynamics at TU Wien since 03/2022. Previously he worked for a large Austrian energy company, as well as a consultant in the energy sector. He studied technical physics and physical energy and measurement engineering at TU Wien, finishing his MSc in 2019.
In his PhD he is working on Optimization problems in the field of industrial energy systems, focusing on flexibility of industrial consumers for redispatch and ancillary services to enhance the resilience of power grids.

Authors:

Martin Fischer TU Wien
Rene Hofmann TU Wien