6. Planning for Large-scale Expansion

Supervisors: Adrian Muntean (main, KAU), Grigor Nika (KAU), Andreas Theocharis (KAU)

Description: Complex energy systems are built upon a number of components interacting together via different mechanisms. The assessment of the performance of such complex systems, when user demand and energy prices vary sometimes drastically in time, require simulation tools that can cope with uncertainty. The starting point in this work is to investigate cases where the uncertainty in data either due to measurements or sudden occurrence of defects or by forecasting, becomes non-negligible. To account for these uncertainty effects, careful sensitivity studies of the results versus input data are needed. This project involves the development of micro-macro stochastic modeling tools, multiscale polynomial chaos expansions, numerical simulation and optimization of parameters defining the target solar energy management system(s), where data improves modeling and consequently a better forecast is reachable.

Supervisors: Andreas Theocharis (main, KAU), Andreas Kassler (KAU), David Lingfors (UU)

Description: Large-scale integration of distributed energy resources such as solar PV systems is accomplished by digitalized smart electricity grids. The observation of the electricity grid, the environmental conditions and the end-users habits via big-data time-series collections and IoT is a key feature for efficient operations and planning. The existence of partially and/or unobserved localities with limited available data due to reasons such as grid expansion, technical failures, economic and social interference, cybersecurity and legal issues have an important role on the prediction of the energy supply, the load demand, the associated uncertainties as well as on the overall system management and planning. In this project, we will investigate possible solutions about energy production and demand forecasting based on Machine Learning and Edge Compute. These are substantial features for the development of smart electricity grids and intelligent management in order to limit grid investments for grid reinforcement and data collection sites.

To enable an accurate forecast of multiple prosumers in the system, we will explore the application of Machine Learning (ML) techniques like Federated Learning in combination to edge computing aiming to deploy the system in each smart energy grid. Furthermore, we will investigating the possibility of making the ML algorithm more understandable with interpretable ML techniques. This is important for the grid operators because it makes the predictions more trustworthy. Lastly, constant evaluation of the current running predictor is an important factor in ML to make sure that it predicts in the best possible way. As such, we are going to use drift detection techniques.

Supervisors: Joakim Munkhammar (main, UU), David Lingfors (UU), Andreas Kassler (KAU)

Description: In order to reach the goal for 100% renewable electricity production in Sweden until 2040, a widespread expansion of renewable power generation is required over the next decades. A significant part of this will most likely be solar power and, judging from recent developments, in the form of large-scale solar parks. Integrating large amounts of solar power in the energy system requires both optimization of how the variable power is used and managed, and planning of where and how to locate and size solar parks. Therefore the PhD candidate will work on developing mathematical, statistical and machine learning methods for optimizing solar power utilization in local energy systems, using electricity management systems in combination with, e.g., battery storage, electric vehicle charging, heat pump systems, and providing grid ancillary services. In addition to this, the PhD student will develop and apply methods for optimal siting and sizing of solar parks, given a wide variety of alternative land uses and constraints, including the hosting capacity of the electricity distribution grid. The project involves case studies of solar utilization planning together with a number of Swedish municipalities and distribution system operators that are partners in SOLVE.