Deep Learning for Energy Systems Optimisation

Optimising the renewable energy management system (REMS) in smart grids (SG) is highly important to increase operational efficiency, lower energy costs, and guarantee grid stability. Renewable energy sources (RES) exhibit high variability and uncertainty. As a result, it has become extremely challenging to efficiently manage such systems with the rapid expansion of RES in power system networks. Conventional analytical techniques require a lot of computations and time to process the massive amount of data from smart grids enabled by RES and advanced metering infrastructure. Probabilistic deep learning (PDL) methods, such as Bayesian neural networks enabled by appropriate optimization techniques can be used to predict different aspects of REMS as well as analyse the enormous amount of smart grid data with a greatly decreased computational complexity.

In this project, probabilistic deep leaning algorithms based on recurrent and convolutional neural networks will be developed to capture complex relationships between renewable energy such as wind and solar generation, demand, and external factors dynamically. The deep learning models with be integrated with different optimization algorithms, which will utilize the predictions from the deep learning models to make real-time decisions on energy storage charging/discharging, and demand response. To be specific, the project will aim to address the following research questions:

• How can deep learning algorithms be utilized to model the uncertainty in energy consumption behaviour of the end users?

• How can deep learning algorithms predict the operational patterns of renewable energy generation assets including solar, wind and battery energy storages?

• How can deep learning algorithms be integrated with energy management mechanisms to develop optimal energy use strategies for end users along with operational strategies for generation assets?

Expected skills for potential PhD students include prior knowledge on developing python scripts, applying machine learning algorithms for complex multi-dimensional time-series datasets, critical thinking as well as academic writing experiences.