Approcci Quantistici per le Smart Grid di Prossima Generazione

The renewable energy field is experiencing significant development and growth, which will continue in the coming decades, thus resulting in an increase in energy generation and data flowing along the electric grid. At the same time, Quantum Computing and Quantum Machine Learning research and investments are bringing these fields towards important growth and a computational speedup with respect to classical techniques. The aim of this work is to explore the next generation of smart grid systems, by focusing on their main issues and challenges, then by proposing some quantum computing and quantum machine learning models to deal with them. Quantum computing tools based on the gate model and quantum annealer are exploited, as well as neural networks and quantum machine learning features like quantum kernel, quantum encoding, and Quantum Random Access Memory (QRAM). After a brief description of the smart grid architecture with its main features, three issues are further analyzed - smart grid variable forecasting, optimization, and stability. With respect to the first issue, a quantum machine learning approach by means of quanvolutional neural network and quantum variational regressor is analyzed, observing that a quantum advantage might be possible considering the more complex kernels than those applied to the classical models and the use of devices such as QRAM. Regarding optimization, it is possible to formalize the problem as a Quadratic Unconstrained Binary Optimization problem (QUBO) and explore a possible implementation for both Coherent Ising Machines (CIM) and D-Wave annealers. Through observation, a quantum advantage may be obtained considering the current results, as well as a size increase and noise suppression in the next generation of devices. Finally, stability is studied through the definition of the swing equations, and a linear solver based on the Harrow-Hassidim-Lloyd (HHL) algorithm is analyzed, observing that a quantum advantage may be found with respect to the size of the input, while taking into account the model parameters. Generally, the feasibility of such quantum applications in the smart grid system will be closely correlated with an increase in the size of quantum systems as well as the quantum encoding feasibility, having to manage a huge flow of data. This work was completed as part of a project by the NTT Data research group. A big thank you to Alberto Acuto and Antonio Policicchio for their crucial support and supervision.