1. Introduction
With the global transition toward renewable energy and carbon neutrality goals, the integration of intermittent energy sources like wind and solar into power systems has accelerated. However, the inherent variability of renewables and fluctuating load demands pose significant challenges to energy management, particularly in industrial parks. Energy storage batteries play a pivotal role in stabilizing grids, reducing peak-valley differences, and enhancing economic efficiency. Yet, frequent charge-discharge cycles and improper operational strategies accelerate battery degradation, undermining long-term system sustainability.
This paper proposes a double-layer optimization framework to coordinate source-grid-load-storage (SGLS) systems while minimizing electricity costs and extending energy storage battery lifespan. The upper layer optimizes energy flow scheduling using time-of-use (TOU) pricing, while the lower layer employs genetic algorithms to manage battery charge-discharge behaviors. Experimental results demonstrate superior performance in cost reduction, renewable utilization, and battery longevity compared to conventional methods.
2. System Description
The SGLS system in an industrial park comprises photovoltaic (PV) arrays, wind turbines (WT), grid connections, loads, and energy storage systems (ESS). Renewable generation prioritizes load supply, with surplus energy stored in ESS during low-price periods. ESS discharges during peak hours to minimize grid purchases. The ESS consists of multiple independently controlled energy storage batteries, whose collective charge/discharge power must align with upper-layer scheduling decisions.
2.1 Wind Power Generation
Wind turbine output power PWTPWT depends on wind speed vv, air density ρρ, and rotor radius RR:PWT=12ρπR2v3CpPWT=21ρπR2v3Cp
where CpCp is the wind power coefficient. Wind variability directly impacts grid purchase decisions and ESS discharge frequency.
2.2 Photovoltaic Generation
PV output PPVPPV is influenced by solar irradiance GACGAC and panel temperature TcTc:PPV=PSTC⋅GACGSTC⋅[1+δ(Tc−TSTC)]PPV=PSTC⋅GSTCGAC⋅[1+δ(Tc−TSTC)]Tc=Ta+30⋅GAC1000Tc=Ta+100030⋅GAC
where PSTCPSTC, GSTCGSTC, and TSTCTSTC denote standard test conditions, and δδ is the temperature coefficient (-0.47%/°C).
2.3 Energy Storage Battery Dynamics
ESS charge/discharge power PESSPESS must satisfy:−PmaxESS≤PESS≤PmaxESS−PmaxESS≤PESS≤PmaxESS
State of charge (SOC) is updated as:SOCt={SOCt−1−α⋅PESS⋅ηd⋅ΔtEESS,PESS≥0 (discharging)SOCt−1−(1−α)⋅PESS⋅ΔtEESS⋅ηc,PESS<0 (charging)SOCt={SOCt−1−EESSα⋅PESS⋅ηd⋅Δt,SOCt−1−EESS⋅ηc(1−α)⋅PESS⋅Δt,PESS≥0 (discharging)PESS<0 (charging)
where ηdηd, ηcηc, and EESSEESS represent discharge efficiency, charge efficiency, and ESS capacity.
Battery lifespan degradation ΔStΔSt depends on depth of discharge (DOD), SOC fluctuations, and operational switching:ΔSt={1×10−4⋅DOD50%+Mt,normal operation1×10−4⋅DOD50%⋅(1+5%)+Mt,high-stress operationΔSt={1×10−4⋅50%DOD+Mt,1×10−4⋅50%DOD⋅(1+5%)+Mt,normal operationhigh-stress operationMt={1×10−6,charge/discharge mode switching0,otherwiseMt={1×10−6,0,charge/discharge mode switchingotherwise
Cumulative lifespan after nn cycles:Sn=1−∑t=1nΔStSn=1−t=1∑nΔSt
3. Double-Layer Optimization Model
3.1 Upper Layer: Economic Cost Minimization
The upper layer employs mixed-integer linear programming (MILP) to minimize electricity costs:min∑t=1TCtgrid⋅(Ptgrid,L+Ptgrid,ESS)mint=1∑TCtgrid⋅(Ptgrid,L+Ptgrid,ESS)
Decision Variables:
- PtPV,LPtPV,L, PtWT,LPtWT,L: Renewable power to load
- PtPV,ESSPtPV,ESS, PtWT,ESSPtWT,ESS: Renewable power to ESS
- PtESS,LPtESS,L: ESS discharge to load
- Ptgrid,LPtgrid,L, Ptgrid,ESSPtgrid,ESS: Grid power to load/ESS
Constraints:
- Power Balance:
PtPV,L+PtWT,L+PtESS,L+Ptgrid,L=PtLPtPV,L+PtWT,L+PtESS,L+Ptgrid,L=PtL
- Renewable Generation Limits:
PtPV,L+PtPV,ESS≤PtPVPtPV,L+PtPV,ESS≤PtPVPtWT,L+PtWT,ESS≤PtWTPtWT,L+PtWT,ESS≤PtWT
- ESS Operational Limits:
−PmaxESS≤PtESS≤PmaxESS−PmaxESS≤PtESS≤PmaxESSSOCmin≤SOCt≤SOCmaxSOCmin≤SOCt≤SOCmax
3.2 Lower Layer: Battery Lifespan Optimization
The lower layer uses a genetic algorithm to schedule individual energy storage battery operations, minimizing lifespan degradation:min∑t=1T∑j=1NΔSt,jbatterymint=1∑Tj=1∑NΔSt,jbattery
Decision Variables:
- Pj,tbatteryPj,tbattery: Charge/discharge power of battery jj at time tt.
Constraints:
- Aggregate Power Matching:
∑j=1NPj,tbattery={PtESS,L,dischargingPtPV,ESS+PtWT,ESS+Ptgrid,ESS,chargingj=1∑NPj,tbattery={PtESS,L,PtPV,ESS+PtWT,ESS+Ptgrid,ESS,dischargingcharging
- Individual Battery Limits:
−Pmaxbattery≤Pj,tbattery≤Pmaxbattery−Pmaxbattery≤Pj,tbattery≤PmaxbatterySOCminbattery≤SOCj,t≤SOCmaxbatterySOCminbattery≤SOCj,t≤SOCmaxbattery
4. Case Study and Results
4.1 Experimental Setup
A case study was conducted in a northeastern industrial park with 2 MW wind capacity, 1000 m² PV panels, and 120 energy storage batteries. Key parameters include:
| Parameter | Value |
|---|---|
| ESS Capacity (EESSEESS) | 500 kWh |
| Max Charge/Discharge Rate | ±200 kW |
| TOU Pricing (Peak/Off-Peak) | 0.15/0.15/0.08 per kWh |
| Initial Battery Lifespan | 1 (per battery) |
4.2 Performance Comparison
Economic Efficiency:
| Metric | Double-Layer Optimization | Single-Layer Optimization |
|---|---|---|
| Total Cost (30 days) | $12,340 | $13,420 |
| Cost Reduction | 8.12% | – |
Energy Storage Battery Lifespan:
| Metric | Double-Layer Optimization | Single-Layer Optimization |
|---|---|---|
| Total Lifespan Loss | 0.72 | 1.22 |
| Remaining Lifespan | 119.28 | 118.78 |
Renewable Utilization:
| Source | Double-Layer Optimization | Single-Layer Optimization |
|---|---|---|
| Wind/PV Contribution | 46.34% | 40.85% |
| Grid Reliance | 24.07% | 28.06% |
The double-layer approach reduces grid dependency by 13.7% and enhances renewable utilization by 5.49%.
5. Conclusion
This paper presents a novel double-layer optimization framework for SGLS systems, addressing both economic efficiency and energy storage battery longevity. By decoupling grid cost minimization and battery lifespan optimization, the strategy achieves superior performance in real-world scenarios. Future work will explore dynamic weight adaptation for multi-objective optimization and scalability for larger grids.