In my recent work, I have focused on developing a comprehensive planning method for battery energy storage systems (BESS) that accounts for the entire lifecycle of the energy storage cell. The increasing penetration of renewable energy sources such as wind and solar has made grid stability a critical challenge. Battery energy storage cells offer remarkable flexibility, providing services ranging from fast frequency regulation to bulk energy arbitrage. However, the economic viability of an energy storage cell is highly dependent on how it is utilized over its lifetime. Different applications impose vastly different stress levels on the battery, leading to varying rates of degradation. My objective is to maximize the total net profit by optimally allocating the capacity of the energy storage cell to different services across its lifecycle.
To address this, I propose a two-stage planning framework. In the first stage of the energy storage cell’s life, it is dedicated to providing frequency regulation in the ancillary services market. After its capacity degrades to a certain threshold, the cell transitions to the second stage, where it participates in the energy arbitrage market, performing load shifting. This sequential utilization concept is novel because it acknowledges that a fresh energy storage cell is best suited for high-power, short-duration tasks like frequency support, while an aged cell can still be profitably used for slower, energy-intensive applications. The planning method determines the optimal sizing and location of each energy storage cell in a distribution system, considering both technical constraints and economic returns over the entire lifecycle.
Multi-Service Application Scenarios for Energy Storage Cells
Ancillary Services Market: Primary Frequency Regulation
The fast response capability of energy storage cells makes them ideal for primary frequency regulation. Unlike conventional generators, an energy storage cell can inject or absorb active power within milliseconds. In my model, the energy storage cell follows a droop characteristic to support grid frequency. A dead band of ±0.033 Hz around the nominal frequency (50 Hz) is defined to avoid unnecessary activation. When the frequency deviation exceeds this band, the energy storage cell’s output power is a linear function of the deviation, as described by the following piecewise function:
$$P_{i,t}^{\xi} = \pm P_{i,\text{bat,max}} \cdot \frac{|\Delta f(t)| / f_{\text{rated}}}{\sigma}$$
where σ is the droop coefficient (set to 5%), Pi, bat, max is the reserved capacity of the energy storage cell i for frequency regulation, and Δf(t) is the frequency deviation at time t. The reserve capacity is the amount of power the energy storage cell commits to have available at all times. In the ancillary services market, the system operator pays a fixed annual price for this reserved capacity, regardless of actual energy delivered. The profit from frequency regulation in the first lifecycle stage is:
$$ \text{Profit}_1 = C_{\text{reserve}} \times C_{i,\text{bat,max}} $$
where Creserve is the annual payment (e.g., 120 USD/kW/year) and Ci, bat, max is the energy capacity (in MW) of the energy storage cell. It is important to note that the actual cycling of the energy storage cell during frequency events accelerates degradation; this is accounted for in the cost model.
Energy Arbitrage Market: Load Shifting
After the energy storage cell’s capacity has degraded, it is better suited for energy arbitrage, which involves charging during low-price periods and discharging during high-price periods. I model the day-ahead electricity price with two periods: off-peak (e.g., 02:00–07:00) and peak (e.g., 17:00–22:00). The energy storage cell charges at λoffpeak and discharges at λpeak. The daily profit is:
$$ \text{Profit}_2 = \sum_h \left( \lambda_{\text{peak}}(h) P_{i,h}^{\xi,\text{dis}} – \lambda_{\text{offpeak}}(h) P_{i,h}^{\xi,\text{ch}} \right) $$
where Pi,h denotes the power exchanged by the energy storage cell in scenario ξ at hour h. The cycle depth is limited to preserve the remaining life of the energy storage cell.
Whole-Life-Cycle Planning Methodology
The core idea of my planning method is to partition the entire lifecycle of an energy storage cell into two distinct periods. Let Y1st and Y2nd be the durations (in years) of the first and second lifecycle stages, respectively. The decision variables include the location and initial capacity of each energy storage cell, as well as the transition point (typically when state-of-charge capacity falls to 20% of initial). The objective is to maximize the net present value of total profit considering investment, operation, maintenance, and degradation costs.
$$ \max \left\{ \sum_{y=1}^{Y_{1st}} \left( \text{Profit}_1 – \frac{365}{S} \sum_{\xi=1}^S \sum_{t=1}^T \left( C_{\text{opm}} + C_{\text{age}} + C_{\text{penalty}} \right) \right) + \sum_{y=Y_{1st}+1}^{Y_{2nd}} \left( \frac{365}{S’} \sum_{\xi=1}^{S’} \sum_{h=1}^H \left( \text{Profit}_2 – C’_{\text{opm}} – C’_{\text{age}} \right) \right) – C_{\text{inv}} \right\} $$
where:
- Copm = Pi, bat, max × CO,BESS (annual O&M cost proportional to capacity)
- Cage = (Cinv / ΓR) × ∑ξ∑t Γeff(t) (degradation cost based on effective cycle life)
- Cpenalty = (365/S) ∑ξ ρ × |ΔPdev(t)| × Δt (penalty for failing to provide the reserved frequency capacity)
- C’opm and C’age are analogous for the second stage.
- Cinv = d(1+d)(Y1st+Y2nd) / [(1+d)(Y1st+Y2nd)+1 – 1] × ∑CE Ci, bat, max (annualized investment cost).
In the first stage, the energy storage cell is dispatched every 15 seconds based on simulated frequency deviations from multiple typical scenarios. The S scenarios represent different wind and load profiles. The penalty term ensures that the energy storage cell does not over-commit capacity that it cannot deliver due to degradation. In the second stage, the energy storage cell follows a daily charge/discharge schedule over H hours, with S’ scenarios capturing price variations. The degradation model uses a linear damage accumulation based on effective full cycles, where ΓR is the rated total cycles (e.g., 4580 for lithium‑ion) and Γeff is the fraction of a full cycle consumed in each operation.
Case Study on the IEEE 33‑Node Distribution System
To validate the proposed method, I applied it to the modified IEEE 33‑node distribution system, as shown in the test network. Two wind farms are connected at nodes 19 and 32. The technical and economic parameters of the lithium‑ion energy storage cell are listed in Table 1.
| Parameter | Value |
|---|---|
| State-of-Charge (SOC) limits | 80% (max), 20% (min) |
| Initial SOC | 50% |
| Charge/discharge efficiency | 75% |
| Rated cycle life | 4580 cycles |
| Self-discharge (per day) | 0.24% |
| Energy capacity cost (CE) | 325 USD/kWh |
| Annual O&M cost (CO,BESS) | 20 USD/kW/year |
Table 2 provides the planning scenario parameters.
| Parameter | Value |
|---|---|
| CAP investment budget | 1,500,000 USD |
| BESS project lifetime | 15 years |
| Annual interest rate | 5% |
| Reserve price (Creserve) | 120 USD/kW/year |
| Penalty factor (ρ) | 20 USD/kW |
| Peak electricity price | 0.28 USD/kWh |
| Off-peak electricity price | 0.04 USD/kWh |
I compared four cases to demonstrate the advantage of the proposed lifecycle planning:
- Case 1: Energy storage cells used only for load shifting (energy arbitrage).
- Case 2: Energy storage cells used only for frequency regulation.
- Case 3: Energy storage cells used simultaneously for both services (mixed operation).
- Case 4: The proposed sequential lifecycle planning (frequency regulation first, then load shifting).
The optimal locations and capacities determined by the planning model are summarized in Table 3. The total net profit over the entire lifecycle (in USD) and the service life of each energy storage cell are also shown.
| Item | Case 1 | Case 2 | Case 3 | Case 4 |
|---|---|---|---|---|
| Load shifting | ✓ | ✗ | ✓ | ✓ |
| Frequency regulation | ✗ | ✓ | ✓ | ✓ |
| Node locations | 31, 18 | 29, 14 | 30, 13 | 30, 14 |
| BESS capacity (MWh) | 3.15 / 1.46 | 3.09 / 1.52 | 3.26 / 1.32 | 2.77 / 1.84 |
| BESS lifetime (years) | 12.5 / 12.5 | 4.35 / 4.4 | 2.16 / 2.1 | 10.66 / 10.8 |
| Investment cost (USD) | 813,330 | 1,210,299 | 1,339,614 | 889,049 |
| Maintenance cost (USD) | 28,904 | 10,191 | 4,946 | 19,958 |
| Frequency regulation revenue (USD) | — | 2,563,531 | 1,246,161 | 589,407 |
| Load shifting revenue (USD) | 2,300,088 | — | 202,407 | 2,438,818 |
| Total net profit (USD) | 1,380,864 | 1,343,040 | 1,784,347 | 2,119,217 |
From the table, several insights emerge:
- Case 1 (only load shifting) yields a long lifetime (12.5 years) but relatively low total profit because the revenue from daily cycles is modest.
- Case 2 (only frequency regulation) generates high short‑term revenue but severely degrades the energy storage cell, reducing its life to only 4.35 years. The net profit is slightly lower than Case 1.
- Case 3 (simultaneous provision of both services) attempts to capture the high revenues of frequency regulation while also getting some arbitrage income. However, the frequent cycling from both services causes extreme degradation, leading to a lifetime of only 2.16 years. Although the total net profit improves over Cases 1 and 2, it is still suboptimal.
- Case 4 (proposed sequential lifecycle planning) achieves the highest total profit (2,119,217 USD), which is 34.8% higher than Case 1, 36.6% higher than Case 2, and 15.8% higher than Case 3. The service life of each energy storage cell is about 10.7 years, much longer than Cases 2 and 3, and only slightly shorter than Case 1. This demonstrates that by reserving the fresh energy storage cell for high‑value frequency regulation and then transitioning to load shifting after degradation, the overall economic return is maximized while maintaining a reasonable lifetime.
The results also show that the optimal location of the energy storage cells differs among cases. In Case 4, the two energy storage cells are placed at nodes 30 and 14, which are associated with higher wind variability and load peaks, allowing them to provide effective frequency support and later benefit from price arbitrage. The capacity allocation ratio (2.77 MWh vs 1.84 MWh) reflects the balance between the two services.
To further illustrate the advantage, I performed sensitivity analysis on the transition threshold. The baseline uses a 20% SOC capacity threshold (i.e., when the usable capacity of the energy storage cell degrades to 20% of its initial value, it moves to the second stage). Changing this threshold affects the duration of each stage. For example, a higher threshold (e.g., 30%) would shift the energy storage cell earlier to load shifting, reducing the total frequency regulation revenue but potentially extending the overall life. My experiments indicate that the 20% threshold yields near‑optimal results for the given market prices and degradation parameters.
Conclusion
In this work, I have introduced a whole‑life‑cycle planning method for energy storage cells that considers multiple application scenarios in a sequential manner. The key innovation is that a fresh energy storage cell is first used for high‑frequency, high‑revenue ancillary services, and after its capacity degrades, it is repurposed for energy arbitrage. This approach aligns the usage pattern with the natural aging profile of the energy storage cell, thereby maximizing total profit while prolonging the useful life. The case study on a standard 33‑node distribution system confirms that the proposed method outperforms both single‑service and simultaneous‑service strategies in terms of net present value. The method is scalable and can be readily adapted to other battery chemistries and market structures. Future work could incorporate more sophisticated battery degradation models and consider uncertainties in electricity prices and frequency deviations explicitly.