25 December 2022
Quantum Computing in System Operation
Elia Group explored the capabilities and limitations of quantum solvers in a proof of concept to solve a complex optimization problem in system operation, measuring the accuracy and runtime of quantum solvers against classical ones.
Elia Group conducted a proof of concept to explore the capabilities and limitations of today's quantum solvers (e.g. quantum annealing), leveraging a platform to access different types of quantum hardware and software. The proof of concept was conducted for a complex MILP (Mixed Integer Linear Programming) optimization problem in system operation used for congestion management, e.g. for solving forecasted overloads within the transmission network. The objective was to define a formal QUBO-translatable model of the optimization problem, implement the QUBO formulation, and measure the accuracy and runtime of the resulting QUBO problem when solved using different quantum in comparison to classical solvers.
Why this project
Quantum computing is a cutting-edge technology that holds the potential to revolutionize the way we solve problems and process information. It uses the principles of quantum mechanics to perform operations on data in a fundamentally different way from classical computing. However, true quantum computing is still in its early stages of development, and there are significant technical and practical challenges that need to be overcome before it can become a mainstream technology.
Elia Group is encountering increasing solving times or loss of accuracy when dealing with complex calculations using classical computing solvers. These difficulties are mainly present in System Operations, where complexity arises from several factors :
- General development towards more detailed modelling of renewables and conventional generators, as well as grid constraints and optimization variables
- Tight process timings due to short-term nature of operational planning and realtime operation with a renewable driven system
- Complex market movements and operational constraints
Quantum technologies can address different categories of problems, where optimization problems are the most mature in terms of technology readiness . Finding the optimal variable combination, given a set of constraints and optimization criteria, from a virtually infinite number of combinations is considered as key feature of a quantum-applicable problem.
Hybrid-quantum such as quantum annealing, or quantum simulators (classical algorithms that simulate the behaviors of full-quantum computers) could help solve the optimization problem by providing assurance that solutions are optimal or near optimal and by reducing solve time.
By exploring the capabilities and limitations today in true and inspired quantum technologies, Elia Group intends to pave the way for a future where quantum computing can truly transform the way information is processed and used.
Elia Group has taken a proactive approach to explore the capabilities and limitations of quantum solvers using a platform that enables access to various types of quantum hardware and software. The aim of this initiative was to conduct a proof of concept (POC) for a complex MILP (Mixed Integer Linear Programming) optimization problem in System Operations used for congestion management, where the goal is to define the optimal preventive redispatching actions and transformer settings required to avoid overloads and ensure an adequate grid operation.
The POC was conducted in two phases. In the first phase, known as Minimal Viable Product (MVP) phase, the primary objective was to build and benchmark a minimal viable product that covers a minimal set of model parameters and complexity end-to-end. In the second phase, known as the Refinement phase, the objective was to incorporate further parameters and complexity based on the results obtained from the MVP.
The hypotheses that were tested during the PoC were as follows: (1) The Congestion Management optimization problem can be practically reformulated as a Quadratic Unconstrained Binary Optimization (QUBO) problem. (2) The QUBO problem can be solved and compared using different Quantum targets. (3) Some of the considered quantum computing (QC) or QC-inspired approaches can outperform the incumbent in terms of absolute runtime (equivalent to additional variables at the same time).
The first phase of the POC involved defining a formal QUBO-translatable model of the simplified but realistic MVP optimization problem, implementing the QUBO formulation, and benchmarking against classical solvers. Two metrics were used to assess the different solvers: accuracy and runtime. To achieve this, the problem had to be reformulated into a QUBO problem because the dominant approach for quantum computing optimization involves translating the mathematical optimization problem into a QUBO problem formulation.
The second phase of the POC mainly focused on improving the model’s scalability and adding more parameters to the problem. Overall, the Elia Group's approach was a proactive step towards exploring the potential of quantum computing and assessing its applicability in solving complex optimization problems in System Operation.
The hypotheses were formulated to test the feasibility of using QC for this task, and the results obtained are presented in this chapter.
Hypothesis 1: The Congestion Management optimization problem can be practically reformulated as a Quadratic Unconstrained Binary Optimization (QUBO) problem.
The first hypothesis was proven to be true. A simplified version of the optimization problem of Congestion Management could be reformulated as a QUBO problem.
Hypothesis 2: The QUBO problem could be solved and compared using different Quantum targets.
The second hypothesis was also proven to be true. The QUBO problem was solved using different quantum instances, such as a simulated annealing solver. A gate-based quantum computer could not have been tested to the full extent, as the simplest version of the QUBO problem exceeded the number of qubits available.
Hypothesis 3: Some of the considered quantum computing (QC) or QC-inspired approaches could outperform current solvers in terms of accuracy and absolute runtime.
The third hypothesis was proven to be wrong. None of the considered QC or QC-inspired approaches outperformed current solvers in terms of accuracy and absolute runtime. Simulated annealer solvers were not a viable alternative to classical solvers as mismatched constraints could not be compensated by faster runtimes.
Though, for larger size of the problem, the quantum-inspired solvers offered much better runtime behavior but led to unacceptable accuracy results. Despite the findings, the project showed that computing challenges, both from an opportunity and threat perspective, are an innovation priority. In conclusion, while QC and QC-inspired approaches may not currently be a viable alternative to classical solvers for grid modelling optimization problems, the potential of QC for computing challenges should not be overlooked. This project has provided valuable insights into the strengths and limitations of Quantum Computing for this type of optimization problem and highlights the importance of further exploration in this area. Therefore, Elia Group plans to continue its investigations into quantum computing as well as high-performance computing to address the growing complexity in grid optimization problems through technological innovation.