As renewable energy integration into distribution networks increases, challenges such as intermittency and unpredictability arise, affecting grid stability. In this context, energy storage systems, particularly vanadium redox flow battery (VRB) energy storage cells, offer a promising solution to enhance renewable energy utilization and mitigate adverse impacts. This article explores an optimal configuration strategy for VRB energy storage cells in distribution networks, considering uncertainties in wind power, photovoltaic generation, and load demand. I will analyze the charging and discharging characteristics of VRB energy storage cells, establish a comprehensive model, and propose an optimization approach using genetic algorithms. The strategy is validated through simulations on an IEEE 33-node system, demonstrating improved renewable energy consumption and reduced reliance on traditional fuel-based power from the main grid.
Renewable energy sources like wind and solar are inherently variable, leading to mismatches between supply and demand in distribution networks. Energy storage cells, especially VRB energy storage cells, can store excess energy during periods of high generation and discharge it when needed, thus smoothing out fluctuations. VRB energy storage cells are advantageous due to their scalable capacity, long lifespan, high efficiency, and safety. In this work, I focus on optimizing the configuration of VRB energy storage cells to maximize economic benefits while promoting renewable energy integration. The optimization considers the unique properties of VRB energy storage cells, including their state-of-charge (SOC) dynamics and efficiency curves.
The VRB energy storage cell operates based on redox reactions of vanadium ions in electrolyte solutions. The power output of a VRB energy storage cell depends on its SOC and stack voltage. The charging and discharging power in per-unit values can be expressed as:
$$ P_{ab} = \begin{cases}
(a_c^{ab} U_{ab} + b_c^{ab}) \text{SOC} + c_c^{ab} U_{ab} + d_c^{ab} & \text{for charging} \\
(a_d^{ab} U_{ab} + b_d^{ab}) \text{SOC} + c_d^{ab} U_{ab} + d_d^{ab} & \text{for discharging}
\end{cases} $$
where \( U_{ab} \) is the stack voltage in per-unit, and \( a_c^{ab}, b_c^{ab}, c_c^{ab}, d_c^{ab}, a_d^{ab}, b_d^{ab}, c_d^{ab}, d_d^{ab} \) are constants specific to the energy storage cell. The efficiency of the VRB energy storage cell varies with SOC and power, as shown in the following equations for charging efficiency \( \eta_c^{\text{VRB}} \) and discharging efficiency \( \eta_d^{\text{VRB}} \):
$$ \eta_c^{\text{VRB}} = \frac{(a_c^p \text{SOC} + b_c^p) P_{\text{VRB}} + c_c^p \text{SOC} + d_c^p}{P_{\text{VRB}}} $$
$$ \eta_d^{\text{VRB}} = \frac{P_{\text{VRB}}}{a_d^p P_{\text{VRB}} + b_d^p \text{SOC}(\text{SOC} – 1) + c_d^p} $$
Here, \( P_{\text{VRB}} \) is the output power in per-unit, and other parameters are constants derived from the energy storage cell’s characteristics. The SOC of the VRB energy storage cell is updated based on charging and discharging actions:
$$ \text{SOC}_k = \begin{cases}
\text{SOC}_{k-1} – \frac{P_{\text{VRB},k} \Delta t}{\eta_d^{\text{VRB}} E_{\text{rated}}^{\text{VRB}}} & \text{for discharging} \\
\text{SOC}_{k-1} – \frac{P_{\text{VRB},k} \eta_c^{\text{VRB}} \Delta t}{E_{\text{rated}}^{\text{VRB}}} & \text{for charging}
\end{cases} $$
where \( P_{\text{VRB},k} \) is the power at time interval \( k \), \( \Delta t \) is the duration, and \( E_{\text{rated}}^{\text{VRB}} \) is the maximum energy capacity of the energy storage cell. To summarize these relationships, I present a table of key parameters for the VRB energy storage cell model.
| Parameter | Symbol | Typical Value | Description |
|---|---|---|---|
| Rated Power | \( P_{\text{rated}}^{\text{VRB}} \) | 300-900 kW | Maximum power output of the energy storage cell |
| Maximum Energy Capacity | \( E_{\text{rated}}^{\text{VRB}} \) | 2000-8000 kWh | Total storable energy of the energy storage cell |
| Charging Efficiency Coefficient | \( a_c^p \) | 0.85 | Constant for charging efficiency model |
| Discharging Efficiency Coefficient | \( a_d^p \) | 0.88 | Constant for discharging efficiency model |
| SOC Limits | SOC | 0.2-0.8 | Operational range for the energy storage cell |
Uncertainties in renewable generation and load demand are critical for configuring the energy storage cell. Wind power \( P_{\text{wind}}(t) \) and photovoltaic power \( P_{\text{pv}}(t) \) include forecast errors modeled as normal distributions:
$$ P_{\text{wind}}(t) = P_{\text{wind}}^s(t) + P_{\text{wind}}^\epsilon(t), \quad P_{\text{wind}}^\epsilon(t) \sim \mathcal{N}(0, \sigma_t^w) $$
$$ P_{\text{pv}}(t) = P_{\text{pv}}^s(t) + P_{\text{pv}}^\epsilon(t), \quad P_{\text{pv}}^\epsilon(t) \sim \mathcal{N}(0, \sigma_t^{pv}) $$
with standard deviations \( \sigma_t^w = (\alpha_w + \beta_w \times P_{\text{wind}}^s(t)) S_w \) and \( \sigma_t^{pv} = (\alpha_{pv} + \beta_{pv} \times P_{\text{pv}}^s(t)) S_{pv} \), where \( \alpha_w, \beta_w, \alpha_{pv}, \beta_{pv} \) are site-specific parameters, and \( S_w, S_{pv} \) are installed capacities. Load demand \( P_L(t) \) is similarly modeled with forecast error \( \epsilon \sim \mathcal{N}(\mu, \sigma) \). The optimization strategy for the VRB energy storage cell is based on real-time balancing of generation and demand. The rules are:
- If \( P_{\text{wind}} + P_{\text{pv}} > P_{\text{load}} \) and SOC < 0.8, the energy storage cell charges.
- If \( P_{\text{wind}} + P_{\text{pv}} > P_{\text{load}} \) and SOC = 0.8, the energy storage cell remains idle.
- If \( P_{\text{wind}} + P_{\text{pv}} < P_{\text{load}} \) and SOC > 0.2, the energy storage cell discharges.
- If \( P_{\text{wind}} + P_{\text{pv}} < P_{\text{load}} \) and SOC = 0.2, the energy storage cell remains idle.
The power output of the energy storage cell is constrained by its physical limits and efficiency:
$$ P_{\text{VRB}}(t) = \begin{cases}
P_{ab} & \text{if } |P_{\text{load}} – P_{\text{wind}} – P_{\text{pv}} – P_{\text{grid}}| > P_{ab} \\
\eta_{\text{VRB}} (P_{\text{load}} – P_{\text{wind}} – P_{\text{pv}} – P_{\text{grid}}) & \text{if } |P_{\text{load}} – P_{\text{wind}} – P_{\text{pv}} – P_{\text{grid}}| \leq P_{ab}
\end{cases} $$
where \( P_{\text{grid}} \) is power from the main grid, and \( \eta_{\text{VRB}} \) is the efficiency. The economic model for the VRB energy storage cell includes daily costs and revenues. The total daily cost \( K_c \) comprises investment, maintenance, and conversion costs:
$$ K_c = k_1 P_{\text{rated}}^{\text{VRB}} + k_2 E_{\text{rated}}^{\text{VRB}} + a_0 $$
with constants \( k_1 = 2.28 \) ¥/kWh, \( k_2 = 0.14 \) ¥/kWh, and \( a_0 = 200 \) ¥. Revenue \( K_b \) comes from selling stored energy and reducing grid purchases:
$$ K_b = \int_{t_1}^{t_2} P_{\text{VRB}}(t) \eta p_e dt – \int_{0}^{24} \frac{P_{\text{grid}}}{\eta_g} p_f dt $$
where \( p_e = 0.45 \) ¥/kWh is the electricity price, \( \eta_g = 0.75 \) is grid generation efficiency, and \( p_f = 0.25 \) ¥/kWh is the fuel price. A penalty cost \( P_c \) for curtailed renewable energy is included:
$$ P_c = \int_{0}^{24} (P_{\text{wind}} + P_{\text{pv}} + P_{\text{grid}} + P_{\text{VRB}} – P_{\text{load}}) c_a dt $$
with \( c_a = 0.18 \) ¥/kWh. The objective function minimizes the total cost:
$$ \min f(P_{\text{rated}}^{\text{VRB}}, E_{\text{rated}}^{\text{VRB}}) = K_c – K_b + P_c $$
subject to constraints on power, energy, SOC, and grid balance. For instance, the energy storage cell’s power must satisfy \( |P_{\text{VRB},k}| \leq |P_{ab,k}(\text{SOC}_{\text{VRB},k}, V_{ab,k})| \), and SOC is bounded between 0.2 and 0.8. Grid power balance at each node \( i \) is enforced through:
$$ P_i + P_{\text{wind}} + P_{\text{pv}} + P_{\text{VRB}} = P_{Li} + U_i \sum_{j=1}^N U_j (G_{ij} \cos \theta_{ij} + B_{ij} \sin \theta_{ij}) $$
$$ Q_i + Q_{\text{wind}} + Q_{\text{pv}} + Q_{\text{VRB}} = Q_{Li} + U_i \sum_{j=1}^N U_j (G_{ij} \sin \theta_{ij} + B_{ij} \cos \theta_{ij}) $$
with voltage limits \( U_{i,\min} \leq U_i \leq U_{i,\max} \). To solve this optimization, I employ a genetic algorithm (GA). The GA encodes \( P_{\text{rated}}^{\text{VRB}} \) and \( E_{\text{rated}}^{\text{VRB}} \) as binary strings, with a population size of 50 and maximum generations of 100. The fitness function is the objective function \( f \), and operators include roulette wheel selection, segmented crossover, and uniform mutation. The algorithm iteratively converges to optimal values for the energy storage cell parameters.
For simulation, I use an IEEE 33-node distribution system with a 1 MW photovoltaic unit at node 17, a 3.6 MW wind turbine at node 32, and the VRB energy storage cell at node 1. Load demand varies between 2.3 MW and 2.8 MW. Parameters for uncertainties are set as \( \alpha_w = 0.02 \), \( \beta_w = 0.16 \), \( \alpha_{pv} = 0.015 \), \( \beta_{pv} = 0.15 \), and load forecast error has \( \mu = 0 \), \( \sigma = 0.01 \). The GA yields optimal values of \( P_{\text{rated}}^{\text{VRB}} = 616 \) kW and \( E_{\text{rated}}^{\text{VRB}} = 5636 \) kWh, with a minimum daily cost of 696.755 ¥. The convergence of the GA is shown in the iteration curve, where the fitness improves over 68 generations. Below is a table summarizing the simulation inputs and results for the energy storage cell configuration.
| Component | Parameter | Value |
|---|---|---|
| Photovoltaic System | Installed Capacity | 1 MW |
| Wind Turbine | Installed Capacity | 3.6 MW |
| Load Demand | Range | 2.3-2.8 MW |
| VRB Energy Storage Cell | Optimal Rated Power | 616 kW |
| VRB Energy Storage Cell | Optimal Maximum Energy | 5636 kWh |
| Economic Outcome | Minimum Daily Cost | 696.755 ¥ |
With the optimized energy storage cell, the power output varies throughout the day, charging during excess renewable generation and discharging during deficits. The SOC curve shows nonlinear changes, peaking at 14:30 and decreasing to 0.45 by 24:00, indicating partial utilization of stored energy. The efficiency of the energy storage cell fluctuates, reaching up to 0.9 at times, but often lower due to low power outputs. Compared to the case without the energy storage cell, renewable energy curtailment is significantly reduced. For example, abandoned wind power decreases from peaks during early hours to minimal levels. Additionally, power from the main grid is largely reduced, with only minor inputs around 20:00 when demand exceeds available energy from renewables and the energy storage cell. This demonstrates the effectiveness of the VRB energy storage cell in enhancing renewable energy consumption and grid economics.
In conclusion, the proposed optimal configuration strategy for VRB energy storage cells effectively addresses uncertainties in distribution networks. By modeling the energy storage cell’s characteristics and economic factors, and using genetic algorithms for optimization, the strategy improves renewable energy utilization and reduces traditional fuel-based power intake. Future work could explore real-time pricing and advanced control strategies to further enhance the performance of energy storage cells. Overall, VRB energy storage cells represent a viable solution for sustainable energy systems, and this approach provides a framework for their deployment in modern grids.