PaperlayerView exact revision on arXiv

Source-linked AI summary

Economic Optimal Operation of Community Energy Storage Systems in Competitive Energy Markets

Reza Arghandeh, Jeremy Woyak, Ahmet Onen, Jaesung Jung, Robert P. Broadwater

arXiv:1407.0433v2physics.soc-phq-fin.GN

TL;DR

Peak-load, outage, and power-quality benefits from energy storage arise only during infrequent events, motivating ways to obtain value between them. The paper proposes market-informed hierarchical CES scheduling and analyzes its operating tradeoffs. Its conclusion emphasizes that economic benefits depend strongly on LMP and load forecast accuracy.

  • Problem

    Energy storage benefits such as peak-load reduction and standby power arise during infrequent peak-load or outage events, motivating additional economic value from fluctuating energy costs.

    Source: §1 · p. 2

  • Method

    The paper proposes hierarchical CES charge/discharge scheduling using GHO, combining market prices and forecasts with real-time measurements and operating constraints.

    Source: §1 · p. 4

  • Results

    Economic benefits are highly dependent on the accuracy of LMP and load forecasts.

    Source: §7 · p. 16

  • Takeaways & Limitations

    Dynamic reserve capacity trades profit potential against outage support duration, giving utilities a basis to align reliability and savings with performance-based rates.

    Source: §6.2 · p. 12

    The study models a 25 kW lithium-ion CES unit and its operational limits, but does not model battery chemical reactions in detail.

    Source: §1 · p. 3

Abstract

from arXiv · show

Distributed, controllable energy storage devices offer several benefits to electric power system operation. Three such benefits include reducing peak load, providing standby power, and enhancing power quality. These benefits, however, are only realized during peak load or during an outage, events that are infrequent. This paper presents a means of realizing additional benefits by taking advantage of the fluctuating costs of energy in competitive energy markets. An algorithm for optimal charge/discharge scheduling of community energy storage (CES) devices as well as an analysis of several of the key drivers of the optimization are discussed.

Symbols

The symbols define the CES scheduling quantities, operating limits, financial terms, and acronyms used in the paper.

Symbols

  • Cmin and Cmax bound stored energy, while Ct denotes battery capacity at time t and ΔCt denotes the change in stored energy.CRsv_t is reserve capacity; HR is outage support duration.Source: Symbols · pp. 1–2
  • PmaxCh and PmaxDCh are maximum charge and discharge rates; PtCESout is CES output power and PtLoad is load at time t.Other power terms cover primary-issue limits, transformer ratings, and feeder-loss reduction.Source: Symbols · pp. 1–2
  • LMPt is the locational marginal price at time t, while RtCh and RtDch denote charging and discharging revenue or cost.Schprofit and Schprofitopt denote scheduled and optimal scheduling profit.Source: Symbols · pp. 1–2
  • CES means Community Energy Storage System, GHO means Gradient-based Heuristic Optimization, and LMP means Locational marginal Price.The acronym list also defines control, network, and energy-resource terms used throughout the paper.Source: Symbols · p. 2

1. Introduction

The paper proposes utility-operated CES scheduling that uses fluctuating market prices while retaining reliability and network-service priorities. It combines hierarchical control and multiple operating considerations to address gaps in distributed-storage optimization.

1. Introduction

  • The strategy seeks additional economic value from fluctuating electricity costs while preserving occasional reliability, efficiency, and peak-feeder-loading benefits.The paper frames those benefits as infrequent, while price-based operation may yield year-round economic benefits.Source: §1 · p. 2
  • Prior work often studied utility-scale storage or centralized substation batteries; distributed-storage studies addressed narrower objectives or omitted day-ahead behavior.The cited studies include feeder-loss reduction and electricity-cost minimization without system day-ahead behavior.Source: §1 · p. 3
  • The study models a utility-owned 25 kW lithium-ion CES unit and considers operational limits, but not detailed battery chemical reactions.CES refers here to utility-owned distributed storage installed on residential distribution networks.Source: §1 · p. 3
  • The paper proposes hierarchical CES control, combining centralized and distributed approaches for system-wide DER operation.The group controller makes substation-level decisions, while each CES controller schedules its battery locally and reports operating conditions and capability.Source: §1 · p. 4
  • The proposed real-time control scheme uses GHO to schedule CES charging and discharging against transformer loading, feeder loss, and LMP prediction.It combines day-ahead load and price forecasts with real-time circuit measurements and market prices.Source: §1 · p. 4
  • The optimization uses a detailed physical distribution-network model with single-phase, multiphase, and unbalanced loads.The paper identifies the lack of physical-based detailed models in prior distributed-resource control and optimization literature.Source: §1 · p. 4

2. Overview of the Community Energy Storage (CES) System

The CES system places utility-operated batteries at distribution transformers and coordinates local units through a hierarchical control structure. Its network model and constraints address both transformer-level and feeder-level operation.

2. Overview of the Community Energy Storage (CES) System

  • CES units connect to the secondary side of distribution transformers, serving small groups of houses and supporting 120/240-volt circuits.The paper notes that this placement can continue serving customers even if the primary system is compromised.Source: §2 · p. 4
  • The system assumes utility ownership and operation, allowing CES units to be aggregated for system-wide optimal operation.The units are considered in the context of residential distribution networks.Source: §2 · p. 4
  • Local CCUs manage voltage, loads, and unit-level issues, while the GCU sends charging or discharging commands and can defer to local control during islanding or communication loss.The GCU algorithm resides in the DCC, where LMP data, forecasts, and operational alerts are available; CCUs report local information upstream.Source: §2 · pp. 4–5
  • CES scheduling incorporates secondary-level transformer constraints, while feeder-level primary constraints are handled through coordinated or aggregated operation.The physical-based DEW-ISM model includes geography, component characteristics, load measurements, and supply measurements.Source: §2 · p. 5

3. CES Reserve Capacity Requirements

The reserve policy uses a dynamic energy reserve tied to expected outage-restoration duration and load. This can reduce storage underutilization, but reserve requirements may conflict with immediate load management.

3. CES Reserve Capacity Requirements

  • A static reserve kept for lengthy outages can underuse CES units when such outages are uncommon.The paper prioritizes reliability benefits in CES control.Source: §3 · p. 5
  • Dynamic reserve capacity keeps enough stored energy to serve the load for a specified outage duration, HR, rather than maintaining a fixed storage level.For a two-hour restoration interval, required reserve may equal full capacity at peak load but only a fraction during light load.Source: §3 · p. 5
  • When preventing a present overload conflicts with the dynamic reserve requirement, the paper prioritizes preventing the current outage over preparing for a potential future one.Storm information may modify reserve capacity in anticipation of longer restoration times.Source: §3 · pp. 5–6

4. CES Control and Optimization Problem Description

CES scheduling maximizes operational savings using energy prices and transformer load forecasts, while respecting battery and network constraints. A greedy gradient-based heuristic builds the schedule from profitable charge/discharge pairs.

4.1. Objective Function for the Optimal CES Scheduling

  • The schedule maximizes savings from CES energy and, when counted, feeder-loss changes; LMP predictions and load forecasts are its primary drivers.Day-ahead LMPs or utility forecasts supply future energy prices, while transformer load estimates cover the next 24 hours.Source: §4.1 · p. 6
  • Feeder-loss changes are estimated by interpolating a quadratic polynomial fitted to three power-flow points, avoiding repeated feeder-wide calculations during scheduling.The coefficients are calculated in advance, requiring three power-flow runs per battery.Source: §4.1 · p. 7

4.2. Equality and Inequality Constraints for the Optimal CES Scheduling

  • Constraints cover battery efficiency, capacity, state of charge, charge/discharge rates, transformer loading, voltage, and reliability requirements.The minimum stored-energy limit prevents full discharge, and transformer-loading bounds use local measurements and load forecasts.Source: §4.2 · pp. 7–8
  • Reserve capacity may be fixed or calculated dynamically from transformer loading and outage-support duration, and stored energy is constrained to preserve that reserve.CES units first satisfy local constraints such as transformer overload prevention while also contributing to fleet-level primary constraints.Source: §4.2 · pp. 7–8

4.3. Gradient-based Heuristic Optimization Solution

  • The heuristic starts from a minimally charging/discharging feasible schedule, then iteratively adds equal energy amounts to the most profitable charge/discharge pair.Charging and discharging occur in different hours, maintaining a consistent time-integrated stored-energy amount; battery and inverter losses can make external input and output differ slightly.Source: §4.3 · pp. 8–9
  • The iteration step is an energy increment whose marginal cost defines the heuristic’s gradient; its maximum size reflects the desired accuracy and available execution time.A time slot stops participating once it reaches a constraint, and the algorithm stops when no remaining charge/discharge pair exceeds the minimum profit set using battery life-cycle information.Source: §4.3 · pp. 8–9
  • The algorithm converges to an optimum as step size decreases when CES and feeder-loss functions are convex in all dimensions.The paper states this convergence condition for the objective as a whole.Source: §4.3 · p. 9

5. Assumptions for the Optimization Solution

The optimization is intended to be rerun as market prices or demand measurements change. The reported operational-profit scope excludes capital and installation costs.

5. Assumptions for the Optimization Solution

  • The algorithm is intended to rerun as LMP prices or demand measurements change; reported average computation time in the DEW environment is under 5 minutes.The paper describes revising operation every five to ten minutes and says neither nearly 24-hour-old nor 15-minute-old information is feasible for operation.Source: §5 · p. 10
  • The simulation uses 50 kWh lithium-ion CES units with a maximum discharge rate of 25 kW and neglects battery self-discharge.The paper notes that the general algorithm can be used with varied storage technologies, outputs, capacities, and losses.Source: §5 · p. 10
  • The simulations use load-research data, customer billing information, and primary-system measurements to model time-varying conditions.Source: §5 · p. 11
  • Reported operational profit means CES operating cost savings and excludes capital and installation costs.Source: §5 · p. 11

6. Simulation Results and Trade-Off Analysis

The simulations show that CES economic benefits depend on LMP forecast accuracy, transformer loading, reserve policy, feeder-loss treatment, and optimization step size. A 20-unit fleet charges during low-price hours and discharges at peak prices.

6.1. LMP Prediction Accuracy

  • CES operation profit was 35% with day-ahead LMP forecasts and 65% with ideal forecasts; the representative transformer’s energy cost without CES was $284.The $284 baseline uses a 0.12 $/kWh tariff in the Detroit, Michigan area.Source: §6.1 · p. 11, Table 1 · p. 11

6.2. Transformer Loading and Reserve Capacity

  • Lightly loaded transformers offer greater market profit, while dynamic reserve decisions trade savings against outage-support duration and depend on accurate load forecasts.Utilities can use the profit–reliability tradeoff to choose reserve duration; fixed reserve capacity is an alternative when load forecasting is poor or loads are constant. Figure 6 also frames fixed reserve capacity as a tradeoff between storage size and cost savings.Source: §6.2 · pp. 12–13, Figure 3 · p. 12, Figure 4 · p. 12
  • Static reserve capacity delivers 39% less benefit than dynamic reserve capacity at 80% transformer loading.The reported variation between the methods depends on customer load shape.Source: §6.2 · p. 13, Figure 5 · p. 13
  • With 30% PEV adoption, CES charging and discharging are constrained to prevent transformer overload during the four-hour PEV charging period.The PEV load creates an overload from 5 PM to 9 PM; the figure distinguishes nominal transformer capacity, PEV load, and PEV load with CES.Source: §6.2 · pp. 13–14, Figure 7 · p. 14

6.3. Impact of the Feeder Losses

  • Accounting for feeder-loss reduction has little effect on attainable savings, while calculating feeder losses takes almost three times the total objective-maximization time.The comparison uses a single CES unit on a moderately loaded feeder and ideal LMP prediction.Source: §6.3 · p. 14, Table 2 · p. 14

6.4. Impact of the Optimization Step Size

  • Changing the optimization step size has relatively little effect on savings but significantly changes computational time.The effective step can be capped by the battery’s remaining stored energy; for example, a nominal 5 kWh step becomes 2 kWh when only 2 kWh remains.Source: §6.4 · p. 14, Table 3 · p. 15

6.5. Aggregated Response of the CES Fleet

  • In a 20-unit fleet, CES units charge during low-LMP hours and discharge at peak prices, reducing peak feeder loading.The case study places 20 units of 50 kWh each on a 13.2 kV residential feeder; the units have 1 MWh total capacity.Source: §6.5 · p. 15, Figure 9 · p. 15

7. Conclusions

The paper presents real-time CES control to increase revenue while retaining occasional reliability, efficiency, and peak-feeder-loading benefits, and finds economic benefits depend strongly on forecast accuracy.

7. Conclusions

  • The Gradient-based, Heuristic Optimization algorithm calculates an optimal charge/discharge schedule for real-time CES control without sacrificing occasional reliability, efficiency, and peak-feeder-loading benefits.The strategy further increases the value of CES systems that support power-system reliability and capacity requirements.Source: §7 · p. 16
  • Economic benefits are highly dependent on the accuracy of LMP and load forecasts.The analysis also considers transformer loading, reserve capacity, and feeder losses.Source: §7 · p. 16
  • Including feeder losses in the objective function has a rather small impact on total profit but a rather large impact on computational time.The authors frame this as a trade-off between accuracy and computational time.Source: §7 · p. 16
Loading 1407.0433v2…