|
Components sizing and load scheduling approach based on usage preference for fuel cost optimization for off-grid PV system |
|---|---|
| รหัสดีโอไอ | |
| Title | Components sizing and load scheduling approach based on usage preference for fuel cost optimization for off-grid PV system |
| Creator | Phaschananpak Jungjariyanon |
| Contributor | Thanaruk Theeramunkong, Advisor |
| Publisher | Thammasat University |
| Publication Year | 2568 |
| Keyword | Artificial intelligence, Heuristic approach, Load scheduling, Optimization, Off-grid PV system, Particle swarm, Renewable energy management, Remote area, Sizing component |
| Abstract | An off-grid photovoltaic (PV) systems integrated with battery storage and diesel generators (PV-Battery-DGs) represent a critical solution for rural electrification, particularly in remote areas where conventional grid extension is economically unfeasible. With over 44% of the global population residing in rural areas and approximately 13% lacking electricity access, these hybrid renewable energy systems have gained significant attention for their potential to provide reliable and sustainable power in grid-isolated locations. How-ever, the inherent intermittency of solar energy production presents substantial challenges for effective energy management, particularly in optimizing operational costs while maintaining user satisfaction and system reliability. This thesis addresses the fundamental challengesof both system component sizing and load scheduling approach based on PSO in off-grid PV-Battery-DGs systems, with particular emphasis on integrating user preferences into the optimization while minimizing fuel consumption costs. The primary objective of this research is to establish a foundation for system design, this thesis develops a 7-step component sizing methodology that ensures proper system configuration while maintaining reliability and cost-effectiveness. The methodology begins with site evaluation using PVWatts Web application to assess solar energy potential, followed by power consumption profile analysis to determine total energy demand through evaluation of appliance ratings, quantities, and usage patterns. The approach proceeds with PV panel sizing by converting AC consumption to DC requirements while accounting for inverter efficiency losses, battery sizing incorporating industry-standard 3-day autonomy requirements with appropriate depth of discharge limitations, inverter sizing to handle peak consumption demand scenarios, charge controller sizing selecting appropriate MPPT technology with proper current handling capacity, and diesel generator sizing determining backup power capacity needed to simultaneously supply critical loads and charge battery banks during extended periods of low solar irradiance. This systematic approach ensures that each component is properly sized to work harmoniously within the overall system architecture, preventing both over-sizing that increases capital costs and under-sizing thatcompromises system reliability, while democratizing access to off-grid PV system design by eliminating specialized technical knowledge requirements. The second objective of this research is to develop an intelligent energy management system that optimizes load scheduling while considering both economic efficiency and user convenience. The study introduces a dual-group classification system that categorizeshousehold appliances into Critical Load (CL) and Uncritical Load (UL) groups based on their operational flexibility. Critical loads represent non-shiftable appliances that must operate during designated time slots to ensure essential services, while uncritical loads comprise shiftable appliances whose usage hours can be adjusted within user-acceptable scheduling durations. The proposed PSO-based algorithm formulates the load scheduling problem as a constrained optimization task, where the objective function minimizes total diesel generator fuel costs across the entire operational horizon while satisfying multiple constraints, including battery state-of-charge boundaries, generator output limits, appliance availability requirements, and user-specified acceptable waiting times. The mathematical formulation incorporates realistic consumption behavior by allowing users to specify a preferred duration for appliance operation, ensuring that optimization results remain practically implementable. The study makes significant contributions to the field of renewable energy management by introducing a constraint formulation that directly incorporates user preferences into the optimization process, addressing the limitation of existing methods that often compromise user convenience for cost optimization. The proposed approach employs an enhanced PSO implementation with dynamic inertia weight adjustment and constraint-handling mechanisms that ensure convergence to feasible solutions within acceptable computational time, while developing a comprehensive energy management framework that coordinates multiple power sources to ensure reliable power supply while minimizing operational costs. The effectiveness of the proposed approach is validated through comprehensive experimentation using multiscale load profiles and real-world data from Nanthaburi National Park, Thailand, representing authentic load profiles and solar irradiance patterns. Thecomponent sizing methodology successfully determined specifications for a 24.71 kWh/day system including 16 PV panels, 2,111 Ah battery capacity, 6.5 kW inverter, 154A charge controller, and 8.3 kW diesel generator. Multiscale accuracy validation across 14.2-61.9 kWh/day consumption profiles demonstrates that the proposed PSO-based algorithm achieves identical optimization results to exact mathematical methods (SLSQP and hand calculation), with generator power output of 6,510.5421 W and fuel cost of 11.4529 units, while dramatically reducing computation time from 2 days to 1.32 seconds, representing a 20-fold improvement in execution speed. The algorithm demonstrates superior scalability with 13-66% faster execution than SLSQP as load profiles scale up, and exhibits only 10-fold execution time growth compared to SLSQP’s 25.7-fold growth from small to large scale. Performance evaluation using a realistic load pattern from Nanthaburi National Park,Thailand. It reveals exceptional practical advantages: the proposed approach achieves 27.2% fuel cost reduction compared to no-control operation with only 0.36 seconds execution time, representing the fastest computation among all evaluated methods. Critically, incorporatinguser preferences increases fuel costs by merely 2.5% compared to traditional PSO algorithms that ignore user convenience, while achieving 7.3 times faster execution speed (0.36 seconds versus 2.633 seconds), demonstrating that user participation incurs negligible economic penalty. Comparative analysis against Genetic Algorithm (GA) approach shows the proposed method achieves 0.3% lower fuel cost with 4.4-fold faster execution (0.36s versus 1.60s), while both methods incorporate user preferences and practical scheduling constraints. A web-based platform was developed to demonstrate the practical implementation of both component sizing methodology and load scheduling optimization, providing an intuitive interface for system parameter input and visualization of optimization results. The study findings have important practical implications for off-grid PV-Battery-DGs system deployment in remote communities. The systematic 7-step methodology democratizes renewable energy system design, enabling communities to independently design reliable systems without specialized expertise. The negligible 2.5% increase in fuel costs for accommodating user preferences is highly acceptable compared to the substantial gains in system usability and user compliance, fundamentally transforming the feasibility of user-participated energy management. The exceptional computational efficiency with sub-secondexecution time makes real-time implementation feasible for resource-constrained environments, enabling dynamic rescheduling capabilities that traditional optimization methods cannot provide. This thesis demonstrates that user-participated optimization can achieve both technical efficiency and practical implementability in off-grid PV-Battery-DGs systems, providing a scalable proposed approach for renewable energy adoption in remote communities that addresses economic sustainability, computational efficiency, and usage acceptance factors crucial for long-term system viability. |
