Assessing Modeling Fidelity for Long-Term Battery Energy Storage Planning: Operation, Degradation, and Temporal Representation

Background

• Capacity loss, efficiency, and self-discharge affect dispatchable energy and energy balance, shaping lifecycle cost and adequacy over long project horizons.Battery condition also feeds back into dispatch, degradation, available capacity, operating losses, and replacement timing.Source: §1 · p. 3

Modeling challenge

• Full hourly chronology preserves operating sequence, while representative periods reduce model size but can disrupt storage-state continuity or omit adequacy-critical conditions.A reduced model may reproduce annual energy or lifecycle cost yet miss short-duration adequacy events.Source: §1 · p. 3

Study objective

• The study fixes capacities sized by a degradation-naive model, then tests alternative battery and temporal representations against cost, SOH, replacement timing, and energy adequacy.The goal is to identify fidelity consequential to the specific lifecycle outcome, rather than maximize model detail.Source: §1 · p. 4

2.1 Battery Degradation and Operational Representation

• Linear, throughput-based, rainflow, and replacement formulations trade physical detail against computational tractability, with impacts depending on operating duty and degradation representation.Operational simplifications also matter: capacity fade affects available energy and replacement, while efficiency and self-discharge affect operating energy balance.Source: §2.1 · pp. 4–5

2.2 Temporal Representation in Long-Term Planning

• Representative periods reduce computational burden, but independent periods can misrepresent storage when interperiod energy transfer matters, and annual energy totals alone may miss adequacy-critical sequences.The highest net-load hour need not fall on the day with the largest sustained energy deficit, so adequacy metrics complement economic and degradation outcomes.Source: §2.2 · p. 5

2.3 Research Gaps and Contributions

• Because prior studies vary system designs and representations, the paper proposes a controlled comparison on a fixed portfolio spanning battery mechanisms, health updates, temporal models, and lifecycle metrics.Its contributions include a composite battery model with targeted ablations, comparisons of two linear surrogates, and tests of interactions among simplifications.Source: §2.3 · pp. 5–6

Planning and lifecycle framework

• A degradation-naive planning model first determines PV and BESS capacities over a 20-year horizon, using fixed BOL efficiencies and self-discharge without capacity fade or replacement.The four-stage assessment framework is shown in Fig. 2.Source: §3.1 · p. 6, Figure 2 · p. 6
• The optimized capacities are fixed during lifecycle experiments, where sequential operation determines degradation and updates SOH, available energy, efficiency, and self-discharge.When the replacement threshold is reached, replacement is recorded and degradation state resets.Source: §3.1 · p. 6
• Cases compare lifecycle trajectories under alternative battery and temporal representations, isolating fidelity effects from changes in optimized system design.The comparisons use economic, degradation, replacement, and adequacy metrics.Source: §3.1 · pp. 6–7

Planning objective

• The planning objective minimizes PV and BESS investment plus discounted operating costs, including net grid transactions and PV fixed O&M.The planning horizon spans years with hourly operating periods.Source: §3.2 · p. 7

Reliability enforcement

• The reliability-enforcement coefficient discourages avoidable load shedding while preserving feasibility, and is not interpreted as a value of lost load.It is introduced for numerical enforcement.Source: §3.2 · p. 7

Operating constraints

• Grid import and export are each bounded by interconnection capacity, with mutually exclusive import/export and charge/discharge operation enforced by binary variables.Annual load growth and declining PV availability are represented in the planning model.Source: §3.2 · p. 7

Battery representation

• The planning model uses rated BESS energy, fixed BOL charging and discharging efficiencies, and fixed self-discharge; the optimized capacities are fixed in lifecycle tests.Chronological energy continuity is maintained from an initial 50% SOC, with a 50% terminal SOC condition each year and no daily or intra-year cyclic closure.Source: §3.2 · p. 7

3.3 Battery Degradation and Replacement

• Calendar and cycle degradation are evaluated from the executed BESS trajectory, with replacement resetting SOH and degradation histories when the threshold is reached.The degradation equations are evaluated after operation rather than embedded in the dispatch optimization.Source: §3.3 · p. 8, §3.3.2 · p. 9

3.3.1 Calendar Aging

• Calendar capacity loss depends on temperature, SOC, and exposure time; equivalent exposure preserves accumulated degradation as SOC stress changes.Temperature is fixed at 25°C in this study, while SOC follows the executed hourly trajectory.Source: §3.3.1 · p. 8

3.3.2 Cycle Aging and Rainflow Counting

• Cycle depths are extracted from the executed SOC trajectory using rainflow counting, and each cycle contributes FEC weighted by cycle count and depth.The effective C-rate uses active charging or discharging periods, excluding idle periods; active periods are normalized by available energy across health-update intervals.Source: §3.3.2 · p. 9
• Accumulated cycle loss is mapped to equivalent exposure under current cycle stress, and rainflow residuals persist across health updates until battery replacement.The square-root exposure relationship means equal FEC increments need not produce equal incremental capacity loss.Source: §3.3.2 · p. 9

3.4 Operational Battery Performance and Linearization

• Lifecycle validation updates charging and discharging efficiency and self-discharge with SOH, so available capacity and operating losses evolve with battery condition.Separate one-way efficiency curves preserve charging/discharging asymmetry, and their product is set to 0.85 RTE at 1C.Source: §3.4 · p. 10, §3.4.1 · p. 10, Figure 3 · p. 10

3.4.2 Self-Discharge

• At the 80% SOH replacement threshold, the reference relationship reduces 1C system RTE from 0.850 to approximately 0.844.The assumed monthly self-discharge rises linearly from 3% at BOL to 6% at the threshold; this is a study-specific sensitivity, not a generally validated LFP degradation law.Source: §3.4.1 · p. 11, §3.4.2 · p. 11

3.4.3 Piecewise Linear Implementation of Dynamic Efficiency

• Fixed power-to-energy-flow maps and SOS2 interpolation preserve continuous operation between efficiency breakpoints while remaining compatible with mixed-integer linear optimization.The optimization interpolates charging power to stored-energy gain and discharging power to energy withdrawal directly, rather than multiplying power by interpolated efficiency.Source: §3.4.3 · pp. 11–12

3.5 Sequential Health Update and Lifecycle Validation

• At each health-update boundary, the inherited physical state is fixed, dispatch is executed through the current interval, and degradation and performance are updated before reoptimization.The process continues through the 20-year horizon, replacing the battery when the SOH criterion is reached.Source: §3.5 · p. 12
• The full-chronology simulation starts at 50% SOC, preserves SOC fraction as capacity changes, and transfers pre-replacement SOC fraction to the new battery.No terminal SOC constraint is imposed at health-update boundaries; rolling solves retain the 50% year-end condition.Source: §3.5 · p. 12
• NPC reports operating and replacement costs for the fixed portfolio, while adequacy is assessed separately using ENS, LOLH, shortfall days, and peak shortfall.The reliability-enforcement term is excluded from reported economic NPC; shortfall measures distinguish unmet-energy magnitude from event duration, frequency, and severity.Source: §3.5 · pp. 12–13

4.1 System Data and Reference Assumptions

• The 20-year Houston microgrid study fixes 29.515 MW PV and 52.196 MWh BESS, repeating one hourly chronology with deterministic load growth and PV degradation.The reference assumes a 3% real discount rate, 25°C battery temperature, and replacement at 80% SOH.Source: §4.1 · p. 13

4.2 Battery Fidelity, Ablation, and Sensitivity Cases

• B1–B3 progressively add cycle-specific C-rate and C-rate-dependent efficiency, then SOH-dependent efficiency and self-discharge; B3 is the reference representation.The battery hierarchy retains the installed PV and BESS capacities across cases.Source: §4.2 · pp. 13–14, Table 2 · p. 14
• Two linear surrogates are distinguished: one omits calendar aging, while the other retains calendar and cycling losses with coefficients calibrated to B3 at its first replacement boundary.The cycle-only throughput surrogate also linearizes cycle degradation, confounding mechanism omission with linearization.Source: §4.2 · p. 14

4.3 Health Update and Temporal Representation Tests

• Health-update intervals compare the three-month B3 reference with six- and twelve-month cases, holding SOH, available energy, efficiency, and self-discharge fixed longer as updates become less frequent.Replacement is evaluated at the end of each health interval.Source: §4.3 · p. 15
• Temporal cases include full chronology, monthly and annual averages, and three 12-day clusters; clustered cases use 11 medoids plus an extreme day and impose 50% daily SOC closure.Daily closure removes original inter-day state continuity, so the reduced cases test complete temporal representations rather than clustering alone.Source: §4.3 · p. 15

4.4 Combined Simplification, Implementation, and Verification

• Lifecycle implementation uses Python/Pyomo and Gurobi, with hourly balance, battery-state, degradation, replacement, ENS, and representative-day-weight checks.The full 8,760-h B3 case is the reference; reported deviations are from this representation, not physical ground truth.Source: §4.4 · p. 16

5.1 Reference Lifecycle Behavior and Battery Fidelity Hierarchy

• B3 reached the 80% SOH replacement threshold in years 9 and 18, ended at 90.59% SOH, and had $111.255 million lifecycle NPC versus $72.763 million planned.B3 produced 24.012 MWh cumulative ENS, equivalent to 0.002588% of lifecycle demand.Source: §5.1 · p. 17, Table 4 · p. 17
• B2 closely reproduced replacement timing and lifecycle cost with ENS near B3; the calibrated calendar-plus-throughput surrogate matched replacement years and NPC within −0.20% but undercounted ENS.At the first B3 replacement boundary, calendar aging contributed 12.61 percentage points, or 62.2% of capacity loss, versus 7.66 points from cycling.Source: §5.1 · p. 17, Table 4 · p. 17

5.2 Mechanism-Specific Effects and Adequacy-Metric Dependence

• Removing calendar aging eliminated replacements and reduced NPC to $72.181 million; the calibrated surrogate matched replacement timing but not the full SOH trajectory, while small C-rate effects reflected low operating duty.B3 mean active C-rate was approximately 0.098C; the calibrated surrogate ended near 95.5% SOH versus 90.6% under B3.Source: §5.2 · p. 18, Figure 4 · p. 18
• The efficiency controls produced different ENS, LOLH, shortfall-day, and peak outcomes; stronger SOH-dependent efficiency deterioration raised ENS about 36% while changing NPC by less than 0.5%.Replacement timing remained unchanged in the stronger-deterioration cases.Source: §5.2 · pp. 18–19, Table 5 · p. 19, Table 6 · p. 19

5.3 Effect of the Health Update Interval

• All tested health-update intervals preserved replacement timing and near-identical NPC, but 12-month updates missed the year-9 shortfall and reduced cumulative ENS by about 18.8%.Annual updating reduced runtime by 58.0%; the longer interval also changes health feedback and redispatch frequency.Source: §5.3 · pp. 19–20, Table 7 · p. 20

5.4 Temporal Representation and Adequacy

• Temporal reductions produced zero ENS in monthly, annual, natural-weight, and peak-informed cases, versus 24.012 MWh under B3; the energy-deficit-informed case produced 365.99 MWh.Day 352 had the largest daily energy deficit but was absent from the peak-informed set; daily SOC closure in the energy-deficit case distorted its severity.Source: §5.4 · pp. 20–21, Table 8 · p. 20
• The temporal experiments repeat one annual chronology and therefore omit inter-annual variation in load, solar availability, and electricity prices.The study also holds configuration and assumptions—including PCC capacity, BESS duration, battery temperature, and deterministic inputs—fixed during lifecycle validation.Source: §5.4 · p. 21, §5.5 · p. 22

5.5 Combined Simplification and Practical Fidelity Implications

• Years 9 and 18 replacements and near-reference lifecycle NPC persisted under combined simplification, but ENS fell from 24.01 to 21.27 MWh and first shortfall shifted from year 9 to 14.Runtime decreased by approximately 40%; the study cautions that sensitivities may vary with system design and operating assumptions.Source: §5.5 · pp. 21–22, Table 9 · p. 22

• Removing calendar aging eliminated both replacements and reduced NPC by 35.1%, whereas a calibrated linear surrogate retained both replacement years and limited NPC deviation to 0.20%.Source: §6 · p. 22
• The linear surrogate’s ENS remained 31.8% below B3, and it did not reproduce the nonlinear SOH trajectory.Source: §6 · p. 22
• The peak-informed 12-day case preserved replacement timing and reported zero ENS; substituting the maximum-deficit day increased ENS to 365.99 MWh while other outcomes remained nearly unchanged.Source: §6 · pp. 22–23
• B3 produced replacements in years 9 and 18, $111.255 million NPC, and 24.012 MWh cumulative ENS.Source: §6 · p. 22
• Full-chronology diagnostics linked the increased ENS to differences in battery state surrounding the critical event.Source: §6 · p. 23
• No single fidelity level preserves every lifecycle metric; select modeling fidelity according to the lifecycle outcome the analysis intends to preserve.Source: §6 · p. 23