• Login
• Admin

Goh, Shu Mei (2024) A prototype for a real-time power fluctuation events prediction and mitigation engine for photovoltaic (PV) grid-tied systems. PhD thesis, University of Nottingham Malaysia.

PDF (Thesis - as examined) - Repository staff only until 8 March 2026. Subsequently available to Repository staff only - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader Available under Licence Creative Commons Attribution. Download (9MB)

Abstract

Power fluctuation events could occur within a minute in solar photovoltaic (PV) grid-tied systems due to the fast-moving clouds, which leads to power quality issues in the power network and decreases the reliability of PV systems. Although a battery energy storage system (BESS) coupled in PV grid-tied system can be controlled to smoothen PV power generation, the occurrence of power fluctuation events is unknown to the BESS. Frequent charges/discharges will have a substantial impact on its cost-effectiveness, and inadequate power supply could deteriorate the power system. These issues can be improved with the integration of a very short-term PV power and power fluctuation event prediction engine alongside a reliable BESS-based mitigation controller. The prediction engine predicts power fluctuation events before they occur, prompting the battery controller to take action and smooth the power fluctuation events at the grid. Herein, this work proposed a novel artificial intelligence (AI) supported hardware-based prediction engine to enable automatic real-time prediction and mitigation of power fluctuation events. It takes in three inputs which are solar irradiance, air temperature and PV power to predict PV power and power fluctuation event in the forecast horizon of 30-second. The prediction outcomes, obtained in real-time before the power fluctuation occurrence, are used to control a BESS-based rule-based controller to mitigate the events in a custom-built laboratory-scale PV grid-tied system. The overall system forms a closed hardware-in-the-loop (HIL).

This research work focuses on implementing unsupervised incremental AI in Field-Programmable Gate Array (FPGA) to achieve real-time PV power and power fluctuation events prediction. To verify the implementation of unsupervised incremental AI in FPGA for real-time HIL system, the first part of the work focused on the hardware architecture of the time-series self-organizing incremental neural network (TS-SOINN). TS-SOINN was chosen because its simulated work had shown high performance in very short-term prediction and mitigation of power fluctuation events based on collected data. The developed HIL system employed with the FPGA-based TS-SOINN predicted 95.85% power fluctuation events with a false positive rate (FPR) of 32.65% and mitigated 83.33% of the power fluctuation events over four weeks of real-time experiment. Following that, a better improved unsupervised incremental prediction model, known as the perturbed time-series self-organizing incremental neural network (PTS-SOINN) was proposed. While PTS-SOINN has a similar structure to TS-SOINN, it has the added advantages of having separate different characteristics nodes in different node pools and has a perturbation function to correct falsely predicted events. Over four weeks of simulation, PTS-SOINN predicted an average weekly accuracy of 95.92% for power fluctuation events, with a corresponding weekly FPR of 30.97%. It outperformed other tested unsupervised neural networks such as SOM, E-SOINN, M-SOINN and TS-SOINN. Furthermore, the proposed PTS-SOINN exhibited the least battery redundant time in mitigating power fluctuation events, ensuring that the battery controller is activated only when necessary. Given that the battery lifespan is intricately tied to the performance of the prediction accuracy and battery controller, it is postulated that PTS-SOINN is likely to prolong the battery lifespan compared to other tested algorithms.

Next, a modified HIL system was employed with FPGA-based PTS-SOINN to perform a week of real-time experiment due to an unforeseen malfunction of the programmable bi-directional inverter in the unmodified HIL system. The FPGA-based PTS-SOINN predicted 94.34% of power fluctuation events with a FPR of 29.89%, which is similar to the average weekly simulation results. Moreover, it mitigated 87.59% of power fluctuation events when integrated with the BESS. From both real-time experiments of FPGA-based TS-SOINN and PTS-SOINN in the developed HIL systems, the battery was able to maintain within the healthy battery SOC range of 30%-100% for power fluctuation events mitigation. The prediction process of both hardware-based TS-SOINN and PTS-SOINN was less than one second, which is suitable for very short-term forecast horizon and real-time prediction. The area occupation of the proposed FPGA-based PTS-SOINN can be easily employed in low-end series Intel FPGA. Thus, it offers a low-cost solution to the HIL PV grid-tied system for real-time PV power and power fluctuation events prediction. To summarize, the proposed hardware PTS-SOINN, with its incremental learning capability was able to predict very short-term PV power and power fluctuation events relatively well. Furthermore, it can be integrated into a BESS-based mitigation engine to form a HIL in mitigating power fluctuation events in PV grid-tied systems in a real-time manner.

Item Type:	Thesis (University of Nottingham only) (PhD)
Supervisors:	Tan, Michelle Tien Tien Maul, Tomas Wong, Yee Wan
Keywords:	power fluctuation events, battery energy storage system (BESS), artificial intelligence (AI), Field-Programmable Gate Array (FPGA), predictive modeling
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering
Faculties/Schools:	University of Nottingham, Malaysia > Faculty of Science and Engineering — Engineering > Department of Electrical and Electronic Engineering
Item ID:	77289
Depositing User:	GOH, SHU MEI
Date Deposited:	09 Mar 2024 04:40
Last Modified:	09 Mar 2024 04:40
URI:	https://eprints.nottingham.ac.uk/id/eprint/77289

Actions (Archive Staff Only)

Edit View