Optimal demand-supply energy management in smart grids
Lancaster University, Doctor of Philosophy, 2022
Supervisors: X. Ma, C.J. Taylor and N. Javeid
Everything goes down if you do not have power: the financial sector, refineries and water. The grid underlies the rest of the country’s critical infrastructure. This thesis focuses on four specific problems to balance demand-supply gap with higher reliability, efficiency and economical operation of the modern power grid. The first part investigates the economic dispatch problem with uncertain power sources. The classic economic dispatch problems seek thermal power generation to meet the demand most efficiently. However, this project exploits two different power sources such as wind and solar power generation into the standard optimal power flow framework. The stochastic nature of renewable energy sources (RES) is modeled using Weibull and Lognormal probability density functions. The system-wide economic aspect is examined with additional cost functions such as penalty and reserve costs for under and overestimating the imbalance of RES power outputs. Also, a carbon tax is imposed on carbon emissions as a separate objective function to enhance the contribution of green energy. The calculation of best power dispatch is proposed using a cost function.
The second part investigates demand-side management (DSM) strategies to minimize energy wastage by changing the time pattern and magnitude of utility load at the consumer side. The main objective of DSM is to flatten the demand curve by encouraging end-users to shift energy consumption to off-peak hours or to consume less power during peak times. It is more appropriate to follow the generation pattern in many cases instead of flattening the demand curve. It becomes more challenging when the future grid accommodates the penetration of distributed energy resources in a greater manner. In both scenarios, there is an ultimate need to control energy consumption. Effective DSM strategies would help to get an accurate balance between both ends, i.e., the supply-side and demand-side, effectively reducing power demand peaks and more efficient operation of the whole system. The gap between power demand and supply can be balanced if power peak loads are minimized.
The third part of the thesis then focuses on modeling the consumption behavior of end-users. For this purpose, a novel artificial intelligence and machine learning-based forecasting model is developed to analyze big data in the smart grid. Three modules, namely feature selection, feature extraction and classification are proposed to solve big data problems such as feature redundancy and high dimensionality to generate quality data for classifier training and better prediction results.
The last part of this thesis investigates the problem of electricity theft to minimize non technical losses and power disruptions in the power grid. Electricity theft with its many facets usually has an enormous cost to utilities compared to non-payment because of energy wastage and power quality problems. With the recognition of the internet of things (IoT) technologies and data-driven approaches, power utilities have enough tools to combat electricity theft and fraud. An integrated framework is proposed that combines three distinct modules such as data preprocessing, data class balancing and final classification to make accurate electrical consumption theft predictions in smart grids.
The result of our solution to balance the electricity demand-supply gap can provide helpful information to grid planners seeking to improve the resilience of the power grid to outages and disturbances. All parts of this thesis include extensive experimental results on case studies, including realistic large-scale instances.