Source-linked AI summary
Control of Battery Storage Systems for the Simultaneous Provision of Multiple Services
Emil Namor, Fabrizio Sossan, Rachid Cherkaoui, Mario Paolone
TL;DR
Battery storage capacity can remain underused as service needs vary, motivating coordinated provision of multiple grid services. The paper proposes a two-phase scheduling and real-time control framework, then evaluates it for feeder dispatch with primary frequency regulation. Across two operating days, it demonstrates deployability, with regulating power provided alongside dispatch and low dispatch tracking error.
Problem
Battery service requirements and deployment vary, while portions of battery capacity can remain unexploited, motivating simultaneous provision of multiple services.
Source: A. Motivations · p. 1Method
A period-ahead optimization allocates power and energy budgets across services, followed by independent real-time setpoints that are superimposed.
Source: C. Paper’s contributions · p. 2Results
Across two operating days, regulating power up to 117 and 25 kW respectively was provided on top of dispatch, performed with a RMS tracking error of about 0.5 kW.
Source: §V · p. 8Takeaways & Limitations
Simulations and experiments support the framework's battery-exploitation objective within operational limits and demonstrate its practical relevance and deployability.
Source: §IV · p. 6, §V · p. 8
Abstract
from arXiv · showhide
In this paper, we propose a control framework for a battery energy storage system to provide simultaneously multiple services to the electrical grid. The objective is to maximise the battery exploitation from these services in the presence of uncertainty (load, stochastic distributed generation, grid frequency). The framework is structured in two phases. In a period-ahead phase, we solve an optimization problem that allocates the battery power and energy budgets to the different services. In the subsequent real-time phase the control set-points for the deployment of such services are calculated separately and superimposed. The control framework is first formulated in a general way and then casted in the problem of providing dispatchability of a medium voltage feeder in conjunction to primary frequency control. The performance of the proposed framework are validated by simulations and real-scale experi- ments, performed with a grid-connected 560 kWh/720 kVA Li-ion battery energy storage system.
I. INTRODUCTION
The paper motivates stacking services to use battery capacity more fully despite uncertainty, surveys multi-service scheduling and control work, and proposes a general two-phase framework with an experimental validation for feeder dispatch and primary frequency regulation.
A. Motivations
- Stacking services can match energy-intensive and power-intensive requirements while using battery capacity left idle by uncertain deployment of a primary service.The paper frames simultaneous services as a way to improve technical and economic exploitation of batteries.Source: A. Motivations · p. 1
B. Literature Survey
- Prior work studies combinations of grid services, but differs in storage technology, operational constraints, real-time control, uncertainty treatment, and battery lifetime analysis.The surveyed scheduling literature commonly aims to increase economic income through multiple revenue streams.Source: B. Literature Survey · pp. 1–2
C. Paper’s contributions
- The proposed framework allocates power and energy budgets ahead of operation, then computes and superimposes independent real-time service setpoints.Its scheduling objective is technical capacity provision rather than revenue maximization, accounting for stochastic service deployment and robust optimization against uncertainty.Source: C. Paper’s contributions · p. 2
- For a medium-voltage feeder, the battery dispatches feeder operation while using remaining capacity to provide primary frequency regulation.The paper reports experimental validation of the control toolchain and describes it as the first such validation for a multi-service BESS control scheme.Source: C. Paper’s contributions · p. 2
II. PROBLEM FORMULATION
The formulation assigns each service power and energy budgets that reflect its uncertain requirements, then optimizes their combined use subject to battery capabilities.
II. PROBLEM FORMULATION
- Each service receives energy and power budgets representing the battery capacity needed for its deployment over time window T.The budgets depend on tunable scheduling parameters and uncertain operating conditions so service requirements are covered across scenarios.Source: §II · p. 3
- The optimization maximizes the combined energy-budget widths while respecting the battery's power and energy capabilities.The combined budget represents energy required by all services within T.Source: §II · p. 3
- Figure 1 illustrates power and energy budgets for a service, while the budget trajectory is formed across the time window.The paper also defines an operation for summing budgets across services.Source: Figure 1 · p. 3, §II · p. 3
- The same framework can use alternative objectives, including economic-revenue maximization or simple operational feasibility.These variants are presented in Appendix A.Source: §II · p. 3
III. CONCURRENT DISPATCH OF A MV DISTRIBUTION
The application couples dispatch of a medium-voltage feeder with primary frequency regulation, pairing an energy-intensive task with a power-intensive service.
III. CONCURRENT DISPATCH OF A MV DISTRIBUTION
- Feeder-dispatch battery needs vary with forecast uncertainty, leaving capacity unused when prosumption uncertainty is small.In other cases, the capacity may be barely sufficient for dispatch alone.Source: §III · p. 3
- Primary frequency regulation is paired with feeder dispatch because it is power-intensive, while dispatch is energy-intensive.The paper also identifies battery ramping capability as suited to fast regulating-power demand in systems with high renewable production.Source: §III · p. 3
- Figure 2 schematizes the experimental setup, with power-flow notation referring to the real-time control.Source: Figure 2 · p. 4
A. Day-ahead problem formulation
The day-ahead problem allocates BESS power and energy budgets to feeder dispatch and primary frequency regulation over a daily planning horizon. It selects the offset profile and frequency-response coefficient while accounting for forecast and frequency uncertainty.
A. Day-ahead problem formulation
- The BESS is planned every day for the next calendar day over a 24-hour operating period, with separate power and energy budgets for dispatch and PFR.The optimization is formulated from those budgets.Source: A. Day-ahead problem formulation · p. 4
1) Power and energy budgets:
- The dispatch plan combines forecasted feeder prosumption with an offset profile F, which is computed to keep stored energy within proper limits.The battery compensates for the mismatch between aggregated prosumer flow and the five-minute-resolution dispatch plan.Source: 1) Power and energy budgets · p. 4
- Dispatch power requirements use high- and low-demand feeder scenarios together with the offset profile, with the forecast deviations obtained using a forecasting tool.The maximum positive and negative requirements are based on the corresponding scenario profiles and F.Source: 1) Power and energy budgets · p. 4
- The dispatch service's associated energy budget is defined after its power budget, for each time index k.The supplied passages specify the budget relationship but do not provide the full energy-budget expression.Source: 1) Power and energy budgets · p. 4
- PFR power is budgeted as a constant profile because instantaneous frequency deviations are difficult to forecast, using the maximum response for deviations above 200 mHz.The proportionality coefficient α is measured in kW/Hz.Source: 1) Power and energy budgets · p. 4
- The PFR energy requirement is derived from frequency data, with W_f,k representing the integrated frequency deviation and its bounds inferred from historical time-series analysis.The analysis used European grid frequency data collected over the previous two years.Source: 1) Power and energy budgets · p. 4
- The frequency-response energy budget depends on α and the frequency-deviation bounds, while α is a decision variable and those bounds and Δf_max are inputs.This separates the tunable control parameter from the quantities representing operating uncertainty.Source: 1) Power and energy budgets · p. 4
2) Decision problem formulation:
The decision problem chooses the PFR coefficient and dispatch offset profile to maximize the usable service budgets. It uses an empirical efficiency correction to impose conservative stored-energy limits.
2) Decision problem formulation:
- The decision vector is x = [α, F], while the frequency-deviation and feeder-demand bounds are treated as input quantities.The objective is expressed using these parameters and control variables.Source: 2) Decision problem formulation · pp. 4–5
- The optimization reduces to maximizing α subject to the problem's constraints.The objective is to maximize the common energy-budget share w.Source: 2) Decision problem formulation · p. 5
- Battery efficiency is incorporated through a two-stage empirical procedure that first solves with nominal energy limits and then models stored energy under non-ideal efficiency.The model uses total BESS power and efficiency to estimate stored energy across the simulation scenarios.Source: 3) Determination of Emin to include… · p. 5
- The ideal-versus-non-ideal stored-energy difference sets a conservative minimum-energy bound; the approach lets service energy budgets be designed separately and then summed.For one simulation case, the largest difference was 4% of nominal capacity and the adopted minimum was 0.05Enom; investigating less-efficient technologies is deferred.Source: 3) Determination of Emin to include… · p. 5
B. Real-time control
Real-time control calculates one-second battery setpoints by adding independent dispatch and PFR control-loop outputs. Saturation limits preserve the day-ahead allocation between the services.
B. Real-time control
- The one-second battery setpoint B_k is the algebraic sum of separate dispatch and PFR setpoints from two independent control loops.The planning phase determines α_o and the offset profile F_o for the subsequent real-time problem.Source: B. Real-time control · p. 5
- The dispatch setpoint compensates for tracking error between the five-minute dispatch target and the mean feeder-power deviation, after filtering out PFR requests.The calculation also includes a short-term load forecast for the remainder of the five-minute interval.Source: B. Real-time control · pp. 5–6
- The dispatch controller defines its setpoint to meet an energy objective over a five-minute horizon.The supplied passage states the horizon but does not provide the setpoint expression.Source: B. Real-time control · p. 6
- The PFR controller calculates a separate frequency-regulation setpoint, while both service setpoints are constrained by day-ahead saturation thresholds.The limits are ±0.2α_o for PFR and ±(Pmax − 0.2α_o) for dispatch; dispatch's five-minute average is expected to stay between L↑ + F_o and L↓ + F_o.Source: B. Real-time control · p. 6
IV. RESULTS
Simulations and real-scale experiments validate the proposed planning and control strategy on a grid-connected battery. The simulations show that selected parameters can maximize battery use within operating limits, while experiments support practical deployability.
IV. RESULTS
- Simulations show that selected α_o and F_o values maximize battery exploitation while respecting operating limits, allowing continuous operation for a month.Experiments validate control-design and simulation assumptions and demonstrate practical relevance and deployability.Source: §IV · p. 6
- The simulations and experiments use a 560 kWh/720 kVA Li-ion BESS connected to a 20 kV feeder serving five office buildings and rooftop PV.The office buildings have a combined peak demand of 300 kW, and the rooftop PV installations total 90 kWp.Source: §IV · p. 6
A. Simulations
Across a 31-day simulation, the day-ahead plan allocates battery energy between feeder dispatch and frequency regulation in response to uncertainty. It generally uses the available energy capacity while setting frequency regulation to zero on five days when demand uncertainty makes the planning constraints infeasible.
A. Simulations
- The simulated initial SOE ranged from 12% to 90%, reflecting the operation of preceding days; day one began at 35%.SOE is stored energy normalized by nominal battery capacity.Source: A. Simulations · p. 6
- The total daily energy budgets reached the battery’s operational limits on 26 of 31 days; on the other five, demand uncertainty made the planning problem infeasible and αo was set to zero.The five days were 10, 14, 16, 20, and 21; no frequency regulation was performed on those days.Source: A. Simulations · p. 6
- The average daily αo was 216.6 kW/Hz, corresponding to up to 43 kW of frequency regulation at Δfmax = 200 mHz.The simulation covered 31 days.Source: A. Simulations · p. 6
- Low feeder-prosumption uncertainty left more battery capacity for frequency regulation, while high uncertainty widened the dispatch energy budget and reduced αo.The passage gives days 5 and 17 as low-uncertainty examples and days 18–20 as high-uncertainty examples.Source: A. Simulations · p. 6
B. Experimental validation
Two days of real-system experiments tested simultaneous feeder dispatch and primary frequency regulation. The dispatch tracking error remained low, while frequency regulation used battery capacity left available after reserving energy for dispatch.
B. Experimental validation
- Day 1 used αo = 584 kW/Hz with an average offset power of 0.84 kW; Day 2 used αo = 127 kW/Hz with an average offset power of −0.56 kW.These are the day-ahead optimization values reported for the two experimental days.Source: B. Experimental validation · p. 7
- The dispatch energy budget was about 54% of nominal battery capacity on Day 1 and 10% on Day 2, with the remaining capacity allocated to frequency regulation.The regulation coefficient was computed to use the remaining capacity while respecting battery constraints.Source: B. Experimental validation · p. 7
- The dispatch tracking error’s RMS was about 0.5 kW over a feeder prosumption averaging about 130 kW.The reported metrics evaluate dispatch performance while frequency regulation is provided concurrently.Source: B. Experimental validation · p. 7, Table IV · p. 7
- The authors note that simultaneous services may reduce the likelihood of battery saturation or depletion when their uncertain processes are uncorrelated, but a limit-reaching failure would stop multiple services.They identify service-prioritization strategies as a possible contingency response.Source: B. Experimental validation · p. 8
V. CONCLUSION
The paper proposes a two-phase BESS control framework that allocates capacity to multiple grid services under uncertain requirements, then superimposes their real-time setpoints. Simulations and two days of field experiments support its use for feeder dispatch alongside primary frequency regulation.
V. CONCLUSION
- The framework schedules battery power and energy budgets for multiple services under variable, stochastic energy and power requirements.Its stated objective is to maximize battery capacity exploitation.Source: §V · p. 8
- The scheduling phase optimizes service allocations using forecast uncertainty, and the real-time phase injects the sum of the individual service setpoints.These are the two phases of the control scheme.Source: §V · p. 8
- For feeder dispatch with primary frequency regulation, daily scheduling determines the maximum regulating power and an offset profile to restore stored energy for continuous operation.The framework is first formulated generically and then applied to this service pair.Source: §V · p. 8
- Two days of operations demonstrated deployability of the proposed control scheme: regulating power up to 117 kW and 25 kW, respectively, was provided on top of dispatch, whose RMS tracking error was about 0.5 kW.Source: §V · p. 8
APPENDIX A ECONOMIC OPTIMIZATION AND FEASIBILITY PROBLEMS
Appendix A describes alternative economic and feasibility objectives for scheduling. Appendix B outlines a statistical method for estimating the energy needs of primary frequency regulation from historical frequency data.
APPENDIX A ECONOMIC OPTIMIZATION AND FEASIBILITY PROBLEMS
- An alternative scheduling objective maximizes economic benefit by optimizing service allocations against each service’s period revenue.The revenue depends on the energy and power budgets reserved for that service.Source: Appendix A · p. 8
- A separate formulation can seek an allocation that ensures feasible operation rather than maximizing energy use or revenue.The appendix presents it subject to the same stated constraints.Source: Appendix A · p. 8
APPENDIX B COMPUTATION OF BESS ENERGY NEEDS FOR PFR
- PFR energy estimates use two years of frequency-deviation data, with daily integrated profiles tested against a normal distribution and bounds derived from their means and standard deviations.The reported goodness-of-fit test did not reject normality at the 5% significance level; the stated bound targets 95% confidence.Source: Appendix B · pp. 8–9