Quantum Computing for Optimal Power Flow and Short-Term Forecasting

Overview

With an increasingly nonlinear and stochastic electrical grid, accurate modelling of power flow problems will become even more essential for grid stabilisation and price minimisation in the near future. Due to the significant additional computational power required for modelling these more complex systems, Golestan et al. (2023) pose that current computational devices and algorithms will not be sufficient.

Hence, finding superior approaches to existing methods of power flow modelling is critical to an orderly transition to a decarbonised electrical power system. As a result of the theoretical advantages of quantum computing technologies with certain tasks, particularly optimisations, it is speculated as a potential solution. Morstyn (2023) exhibits that there is already value that can be provided through using hybrid solvers, combining quantum annealing and classical technologies, for solving optimal power flow problems.

The proposed research for this DPhil will build on the work by Dr Morstyn to investigate approaches for developing more complex optimal power flow models using quantum hardware. Additionally, it will investigate methods of applying the models to short-term forecasting of electricity pricing. This project falls within both the EPSRC quantum technologies and, energy and decarbonisation, research areas.

Methodology

Firstly, as a natural development on top of Morstyn's work in 'Annealing-Based Quantum Computing for Combinatorial Optimal Power Flow', we will investigate methods to model a grid with fully renewable and intermittent generation sources (assuming normalised renewable generation profiles) and conduct an optimal power flow analysis with discrete decision variables being the size of the generation, location of generation, size of battery storage and location of battery storage. To do this, we will investigate using a discrete quadratic model and a constrained quadratic model of the system.

The software that will be used for developing the quantum annealing model will be D-Wave's Ocean SDK and it will be tested using the Leap Quantum Application Environment for cloud-based annealing. Using hybrid solvers, we will also explore different distributed optimisation approaches, such as Alternating Method of Multipliers and Bender's Decomposition, and how this can generate sub-problems for quantum processors.

Developing on this model, we will then investigate methods to improve real-world modelling through the integration of factors including weather forecasting and historical demand data. We will investigate developing QUBO (Quadratic Unconstrained Binary Optimisations) or Isling models for these variables to allow classical data to be embedded into a quantum annealer.

We will also investigate combining quantum forecasting and optimisation, based on quantum annealing and gate-based quantum computing being relevant for both forecasting and optimisation algorithms. The formulation of gate-based programs will be developed in the Qiskit SDK and ran on the IBM Qiskit Runtime.

Outcomes

- Develop an optimal power flow model ran on quantum technology, which allows for the solving of problems with integer-based decision variables. - Develop a model to enable accurate modelling of high resolution electrical grids which use a nodal pricing

- Develop models to allow the embedding of classical data onto quantum states, to be used in power flow optimisation problems.

The wider value of this research is that the development of quantum computing will allow the modelling of electrical grids with higher fidelity, that were not possible with classical computation, and to solve existing problems in an accelerated period of time. This will allow for improved modelling of electricity grids and it will allow network operators and market participants to simulate and predict the grid more efficiently.