Applications of Regression Enhanced Self-Organizing Incremental Neural Networks (RE-SOINN) in an embedded intelligent energy control system for off-grid systems

Puah, Boon Keat (2024) Applications of Regression Enhanced Self-Organizing Incremental Neural Networks (RE-SOINN) in an embedded intelligent energy control system for off-grid systems. PhD thesis, University of Nottingham Malaysia.

[img] PDF (Thesis - as examined) - Repository staff only - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
Available under Licence Creative Commons Attribution.
Download (5MB)


Battery Energy Storage System (ESS) is a very important component in most of the off-grid standalone photovoltaic (PV) system as battery ESS can store excess unused solar energy harvested via solar panels for later use. However, normal household electronic appliances often cause irregular load demand, pushing the battery ESS to experience deep power discharge cycles. The intermittent nature of solar irradiance has also contributed to irregular charging patterns of battery ESS. Therefore, battery-supercapacitor Hybrid Energy Storage System (HESS) is introduced as the most promising solution to reduce the harms to the battery. Most of the control strategies in the literature focus on the optimization of the system operations. For instance, the control strategy is usually implemented to optimally manage HESS based on real-time operating conditions. Though there are applications of intelligent energy control system in the past literature, there are still limited studies on how solar irradiance prediction is being embedded into the energy management system (EMS). Furthermore, many solar irradiance forecasting solutions provided in the literature are based upon deep learning algorithms which require a lot of computational resources, impeding these forecasting solutions to be implemented in off-grid applications. Also, the EMS implemented in the literature are based upon Artificial Intelligence-based optimization algorithms which use iterative computations to reach to optimal decisions making. As a result, an incremental unsupervised learning-based EMS is developed to perform solar irradiance forecasting as well as to manage and control the system operations in an off-grid standalone PV Renewable Energy Power System (REPS). The objectives of this research study are to develop a computational efficient hourly incremental unsupervised learning solar irradiance forecasting model using basic features as well as to develop and to implement an incremental unsupervised learning EMS in an actual standalone PV system with battery-supercapacitor HESS.

The novelties of the incremental unsupervised learning solar irradiance forecasting model include less computational demanding compared to deep learning models. Moreover, the proposed model takes historical solar irradiance measurements and timestamps as its sole input. As a mean to improve the forecasting performance of the proposed model, the data is decomposed into low frequency and high frequency components to reduce the influence of noisy variation of solar irradiance on the learning of the model. The incremental nature of the model also allows the proposed model to learn from new data once a gradual change in the data trend is found. The proposed model outperforms Artificial Neural Networks (ANN) by 19% in terms of Root Mean Squared Errors (RMSE) and 34% in Mean Absolute Scaled Error (MASE).

On the other hand, the proposed EMS model implemented does not require any predefined mathematical representation of the entire REPS due to its incremental feature. Then, its lightweight advantage outperforms many controllers in the past literature in terms of computational time, training time and specifications requirement on embedded computational platform. The proposed model introduces significant less battery power oscillations in the REPS, especially at higher battery power amplitudes that are very damaging to the battery while assists the REPS with HESS to harvest extra 26.6% of solar energy compared to a battery only REPS.

Item Type: Thesis (University of Nottingham only) (PhD)
Supervisors: Begam, Mumtaj
Khan, Nafizah
Wong, Yee Wan
Keywords: battery energy storage system (ESS), hybrid energy storage system (HESS), solar irradiance forecasting, incremental unsupervised learning, energy management system (EMS)
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: 77358
Depositing User: PUAH, Boon
Date Deposited: 09 Mar 2024 04:40
Last Modified: 09 Mar 2024 04:40

Actions (Archive Staff Only)

Edit View Edit View