(A) multi-objective operation strategy for a PV-integrated hybrid UPS-ESS using predictive heuristics and receding-horizon control

국문 초록 (Abstract)

최근 산업계에서는 기후 변화 대응의 방법으로 태양광(PV)을 활용한 재생 에너지 도입이 촉진되고 있다. 정밀 제조 시설에서는 순간적인 전압 강하 및 정전에 대응하기 위해 무정전 전원 공급 장치(UPS)가 요구되며, PV의 간헐적인 출력을 보정하기 위해서는 에너지 저장 시스템(ESS)이 필수적이다. 이에 본 연구는 PV가 통합된 하이브리드 UPS-ESS(HUE) 시스템을 위한 다목적 제어 전략을 제안한다.
부하 및 PV 발전량에 대한 단기 예측은 장기 단기 기억(LSTM) 모델을 사용하여 수행한다. 이러한 예측을 바탕으로, UPS 예비 전력을 확보하기 위해 필요한 최소 충전 상태(SOC)와 시스템 운영을 위한 목표 SOC가 각 시간 단계에서 계산된다. 이후 다중 기준 의사 결정(MCDM)을 통해 각 목적 함수의 가중치를 도출하고, 예측형 휴리스틱과 이동구간제어(RHC)를 결합하여 실시간 운영 전략을 구축한다.
예측형 휴리스틱은 운영 제약에서 충전 및 방전 동작을 결정하고, RHC는 예측값과 실제값을 모두 반영하여 각 시간 단계마다 제어 결정을 업데이트한다. 제안된 제어는 UPS 대응 준비를 위한 적절한 SOC 여유 공간 확보, 전력 비용 절감을 위한 전력망 피크 억제, 배터리 열화를 가속화하는 불필요한 SOC 변동 제한을 목표로 한다.
시뮬레이션 결과는 예측형 휴리스틱과 RHC를 결합한 제안된 다목적 운영 전략이 SOC 안정성, 전력망 운영 신뢰성, 비용 효율성 측면에서 비교 시나리오보다 우수한 성능을 보유함을 입증한다.

다국어 초록 (Multilingual Abstract)

In recent industry, the adoption of renewable energy using photovoltaics (PV) has been promoted as part of efforts to address climate change. In precision manufacturing facilities, an uninterruptible power supply (UPS) is required to respond to momentary voltage sags and outages, and an energy storage system (ESS) is essential to compensate for the intermittent output of PV. Accordingly, this study investigates a multi-objective control strategy for a hybrid UPS–ESS (HUE) system integrated with PV.
Short-term forecasts of load and PV generation are performed using a Long Short-Term Memory (LSTM) model. Based on these forecasts, the minimum state of charge (SOC) required to secure HUE reserve power and the target SOC for system operation are calculated at each time step. Subsequently, the weights of each objective function are derived using Multi-Criteria Decision Making (MCDM), and a real-time operation strategy is constructed by combining predictive heuristics and Receding-Horizon Control (RHC).
The predictive heuristic determines charge and discharge actions under operational constraints, while RHC updates the control decision at each time step by incorporating both forecasted and actual values. The proposed control aims to maintain an adequate SOC headroom for UPS readiness, suppress grid peaks to reduce electricity costs, and limit unnecessary SOC fluctuations that accelerate battery degradation.
Simulation results demonstrate that the proposed multi-objective operation strategy combining predictive heuristics and RHC outperforms the comparison scenarios in terms of SOC stability, grid operation reliability, and cost efficiency.

목차 (Table of Contents)

• 1. Introduction 1
• 1.1. Research Background 1
• 1.2. Research Objective 2
• 1.3. Research Contribution 3
• 1.4. Structure of Thesis 5
• 2. PV-Integrated Hybrid UPS-ESS System 7
• 2.1. Energy Storage System (ESS) 7
• 2.1.1. ESS Configuration and Control System 8
• 2.1.2. SOC Calculation Formula and Constraints 9
• 2.1.3. Predictive-based ESS control case 10
• 2.1.4. ESS Control Technology by Application 12
• 2.2. Uninterruptible Power Supply (UPS) 13
• 2.2.1. Offline UPS 14
• 2.2.2. Line-Interactive UPS 15
• 2.2.3. Online UPS 16
• 2.2.4. Control Strategies for UPS Systems 17
• 2.3. Hybrid PV-UPS-ESS System Overview 18
• 2.3.1. Related Work of Hybrid UPS-ESS Systems 20
• 2.4. Research Trends and Limitations 21
• 3. Predictive Heuristic Operation Strategy 24
• 3.1. LSTM-Based PV and Load Forecasting 24
• 3.1.1. Data Preprocessing and Inputs 24
• 3.1.2. Network Architecture and Training 26
• 3.1.3. Forecasting Performance 29
• 3.2. Predictive based Heuristic operation strategy 31
• 3.2.1. System Model and State Variables 32
• 3.2.2. Minimum SOC for UPS Backup 33
• 3.2.3. Peak-Shaving Control Rules 34
• 3.2.4. Dynamic SOC and ESS Control Logic 35
• 4. Receding-Horizon Control (RHC) 37
• 4.1. Lookahead SOC target for RHC 37
• 4.2. Charge/Discharge Decision Rules 38
• 4.3. Multi-Objective Functions & Metrics 39
• 4.3.1. J1 – Safety-Headroom Deviation 40
• 4.3.2. J2 – Electricity Cost 41
• 4.3.3. J3 – Cycling Severity (EFC) 42
• 4.3.4. J – Integrated objective function 43
• 4.4. MCDM for Horizon Selection 43
• 4.4.1. Pareto Front & Knee Point 44
• 4.4.2. TOPSIS (Equal/Entropy Weights) 45
• 4.4.3. Weight Back-Fitting 47
• 4.4.4. Final Weight 48
• 5. Simulation and Discussion 49
• 5.1. Simulation Configuration 49
• 5.1.1. UPS Backup Capacity and SOC Lower Limits 50
• 5.2. Heuristic and Peak-Limiting Simulation Results 52
• 5.3. RHC Simulation Results 54
• 5.3.1. SOC target for RHC 54
• 5.3.2. Lookahead Horizon of the Objective Function 57
• 5.3.3. Weights of the objective function 59
• 5.3.4. SOC & Grid of RHC 62
• 5.4. Performance Analysis 64
• 6. Conclusions 67
• References 69
• Korean Abstract 74