Energy Transition strategies using Artificial Intelligence and Machine Learning for optimal operation of Grid Integrated Renewable Sources

Abstract

newlineRestructured power systems have emerged as a dynamic response to the changing newlinelandscape of the energy sector.The driving forces behind restructuring include the newlinepursuit of greater efficiency, enhanced competition, and integration of renewable energy newlinesources. In this thesis, role of power system restructuring in facilitating the integration newlineof renewable energy, energy storage, and demand response technologies is considered. newlineIt addresses the necessity of designing market mechanisms that appropriately value the newlinecontributions of these resources to grid stability alongwith environmental sustainability newlineis examined. newlineShort-term load forecasting (STLF) stands as a critical component within newlinemodern power system operations and planning, facilitating efficient resource allocation, newlinegrid stability, and economic viability. This thesis delves into the significance, newlinemethodologies, challenges, and applications of STLF in the context of electricity newlinedemand prediction.Challenges inherent in STLF are discussed, encompassing factors newlinesuch as the influence of weather conditions, seasonality, and special events on load newlinepatterns. The thesis addresses the complexities of capturing abrupt load changes, newlineaccounting for demand response initiatives, and adapting models to evolving consumer newlinebehaviour, thereby underlining the need for robust and adaptable forecasting strategies. newlineIn this thesis, STLF is implemented and simulated using four different strategies newlineviz Analysis using Pure Regression Methods, Classifier-Regression Formulation, newlineClassifier-Regression Mapping and Validation-Regression Mapping. The main newlineobjective of simulating all the scenarios is to find out the most suitable strategy when newlineintegrated with active power scheduler. After careful analysis, it was found out that newlinethe most suitable methodology for prediction of load is Classifier-Regression Mapping newlinewith least root mean square error of 0.083 and R2 of 0.85 and also its suitability with newlinescheduler is validated.Power scheduling is a crucial aspect of modern power system operation and newlinemanagement, focusing on the efficient allocation and coordination of generation newlineresources to meet electricity demand while ensuring grid stability and reliability. This newlineconcept involves the real-time adjustment of power generation outputs to maintain a newlinebalance between supply and demand, taking into account the factors like load variations, newlinegeneration constraints,and system limitations. This thesis centres on the analysis of an newlineoptimised day-ahead power schedule, which encompasses several market participants newlinesuch as Micro Grid (MG), Distributed Generation (DG), Distribution Storage (DS), and newlineVolt/Var control and Feeder Reconfiguration. This study examines several operational newlinemodes of Distribution Energy Resources (DER), encompassing Planned Islanding newlineevents. The objective function is designed to minimise operational costs throughout newlinethe duration of a day. Constraints are derived by behaviour of the participants inside newlinethe market. The problem is defined using the YALMIP and solved using GUROBI newlinesolver in the MATLAB platform for a system consisting of 33 nodes. It was observed newlinethat The optimisation of Distributed Energy Resources (DER) in non-islanded mode newlineresults in a reduction of total operating costs by 10.57%.The optimisation of distributed newlineenergy resources (DER) in islanded mode results in reduction of total operating costs by newline10.08% as compared to non-optimized cases.In comparison to the algorithms proposed newlinewithin this field, there is an average reduction of 3.37% in operational time.Furthermore, newlineit has been noticed that the algorithm under consideration successfully mitigates voltage newlinecongestion. newlineFinally in this thesis STLF (Classifier- Regression mapping) is integrated with newlineactive power scheduling algorithm and operation is managed effectively. It was newlineobserved that due to short term forecasted load the accuracy of active power scheduling newlinehas improved. It is shown that the overall operating cost of the system is reduced by newline11.1 % due to better planning and less dependence on ancillary services and overall newlinecomputation time is reduced by 18.75 % as compared to algorithm already existing in newlinethis area. newlineVarious case studies have been performed in this thesis in collaboration with CEA newline(Central Electricity Authority),India and Wroclaw University Poland for STLF . The newlineeffectiveness of the proposed classifier regression mapping has been demonstrated in newlinethe thesis newline