An EV fleet operator and public charging-point operator face two coupled decisions: which vehicle to charge when at what power level; where to deploy new charging points and at what capacity. Classical EV OR problems: EV Charging Scheduling and Charging-Station Location Planning.
In plain words
Sound familiar?
- We have a 50-500 EV fleet (courier, delivery, taxi, rental); the evening charging sequence at the depot is decided by the planner's intuition.
- 30-100 charging points are installed at the depot; when all open simultaneously the transformer power spikes to peak, and the distribution utility charges a peak-draw penalty.
- Time-of-use tariff makes night rate half the day rate; but the operational schedule does not exploit this — some vehicles are plugged in early and charged at expensive hours.
- We have 10-30 public charging points; some sites are full 8-12 hours a day while others are used 1-2 hours — site selection was historically intuitive.
- We are evaluating a new charging-station investment; where to site, slow or fast, how many sockets — without a concrete demand forecast or cost model.
- Our DC fast-charge points (50-150 kW) show high utilisation but battery aging complaints come in; we cannot balance pricing against battery OPEX.
- Distribution grid connection capacity for new investment zones means transformer + line build can take 6-18 months; our joint investment planning is weak.
Why it matters
How it's solved
Technical depth
How it's solved
Technical depthIn one sentence: Spread the window between the start of the night tariff and the deadline by which each vehicle must be ready across the fleet without breaching the transformer capacity, and use fast-charging only for the urgent cases — minimising electricity cost and battery-replacement cost together.
The problem is studied in Operations Research (the discipline that uses math and computers to solve business decisions) as two coupled problems: EV Charging Scheduling (operational — which vehicle to charge when, at what power level) and Charging-Station Location Planning (tactical — where to place public charging points, at what capacity). A third emerging layer: Vehicle-to-Grid (V2G — selling vehicle battery back to the grid). Solution in three stages:
1. Modeling. For operational scheduling — input data: vehicle fleet (per vehicle: battery capacity kWh, current state-of-charge SoC, minimum SoC required next morning — deadline constraint, allowed charging power levels), depot charging points (count, maximum kW per point), transformer/grid capacity (instantaneous total kW peak limit), time-of-use tariff (hourly electricity price curve — day/peak/night). Constraints: each vehicle ready by deadline, simultaneous total draw ≤ transformer capacity, battery SoC lower-upper limits, battery aging cost (modern literature adds this to the objective). Objective: minimum total electricity cost + battery aging cost + peak-draw penalty. For siting — input data: regional EV penetration forecast (5-10 years ahead), candidate locations (car parks, malls, highway stops, fuel-station renewals), per candidate the land cost + grid-connection cost + expected daily demand, AC slow vs DC fast mix. Constraints: investment budget, regional coverage (every EV user within X km of a point), grid-connection feasibility.
2. Solver-driven decision. Operational scheduling: small-medium fleet (50-200 EVs) by MILP (Mixed-Integer Linear Programming — optimisation with some 0/1 variables and some continuous) — hourly charging decision is binary (on/off in 1-hour slots) or continuous (kW power level), 24-48 hour horizon. Solves in minutes-hours on a single machine. Large fleet (200+ EVs) by Lagrangian relaxation or decomposition (splitting a big problem into parallel pieces — per-vehicle sub-problems solved in parallel, grid capacity is the global constraint). Dynamic vehicle arrival (taxi, rental fleet — arrival times uncertain): online policy (priority rule deciding at arrival time) or rolling-horizon MILP (re-solved every half hour). Siting: classical approach — p-median or maximum-coverage formulation (maximises user coverage among candidate locations under investment budget). For EV-penetration uncertainty, two-stage stochastic programming (first stage siting decision, second stage scenario-based demand realisation). V2G: selling battery back to the grid involves an arbitrage opportunity (charge cheap at night, discharge expensive in the day) traded off against the battery aging cost; solved by MILP or stochastic MPC (Model Predictive Control — rolling-horizon predict-and-control).
3. Field integration. Output in three layers: daily schedule (per vehicle: which charge point, which hour, which power level — Gantt panel in operations centre), charge-point control commands (over OCPP — Open Charge Point Protocol, the industry-standard open protocol for opening/closing a charge point and setting power), monthly siting report (per public site: utilisation, demand trend, capacity expansion recommendation). The Fleet Management System (FMS) feeds telematics (per-vehicle live SoC, location, planned next-day duty) into the schedule. Quarterly committee: actual vs planned peak draw, electricity-cost variance, battery-aging tracking, public-site utilisation curve, investment decision (open/close site, raise capacity).
Alternatives
Manual + spreadsheet + heuristic ordering
FreeZero license
Who it fits: Small fleet (<30 EVs) or 1-3 public sites
- + Zero software cost
- + Planner experience leads
- + Simple rule (start at night rate) yields 5-10% saving
- − Above 30 EVs the peak-draw control mentally saturates the planner
- − Battery aging cost is not accounted for
- − As demand grows site selection is done without a numerical model
- − V2G arbitrage opportunity missed
Local charge-management platform
Enterprise200K-800K TRY license + 80K-250K TRY/year maintenance (TR market observation)
Who it fits: Medium fleet (30-150 EVs) or 5-15 public-site operator
- + OCPP integration standard
- + Local-language UI, local support
- + Billing + payment integration included
- + Basic time-of-use tariff optimisation rules
- − MILP-based scheduling module typically absent — heuristic priority rules
- − Battery aging cost modelling weak
- − Siting-planning module limited
International charge-management + optimisation package
Enterprise400K-2M EUR license + 100K-500K EUR/year maintenance
Who it fits: Large fleet (150+ EVs) or multi-site (20+) charging-network operator
- + MILP or rolling-horizon MPC module mature
- + Battery aging model integrated
- + Siting analysis with GIS + demand forecast two-way
- + V2G module available (in advanced markets)
- − High license + 9-18 month deployment
- − Local-tariff customisation extends project
- − Wide training programme for operations team
Open-source solver + custom module
Open SourceFree license; in-house build 16-32 weeks or 600K-1.5M TRY consulting
Who it fits: Operator with a tech team, integration with existing FMS / charge-management platform
- + No license fee
- + Electric-vehicle charging scheduling methods are well defined for open-source solvers
- + Site-location modelling under demand uncertainty is available in the open literature
- + Open Charge Point Protocol (OCPP) — charge-point hardware independence
- − Needs in-house OR specialist + power engineer + integration team
- − Moving from academic prototype to field system 6-12 months
- − Maintenance burden stays with the operator
Recommendation
Ask in the meeting
- Does the charge-scheduling module use exact mathematical programming, rolling-horizon predictive control (re-planning every 15-30 minutes), or only a heuristic priority rule?
- Is battery aging cost part of the objective function? Which modelling approach (calendar aging + cycle-aging curve, depth-of-discharge based) is used?
- Is the hourly variable tariff structure (hourly price, day/peak/night windows, seasonal variation) configurable, or hard-coded?
- Is the peak-draw (kW) constraint dynamic (time-varying grid capacity) or fixed? Are multiple transformers / multiple depots supported?
- Which approach does the public-site location module use — average-distance-minimising location, maximum-user-coverage location, or scenario-based location under demand uncertainty? How is electric-vehicle-penetration uncertainty modelled?
- Which version of OCPP (Open Charge Point Protocol) is supported? Has multi-brand charge-point integration been tested?
- Is a V2G (Vehicle-to-Grid) module available? How is the arbitrage-revenue vs battery-aging-cost trade-off optimised?
- If we end the contract, in which format can we export charging history, battery SoC data, pricing rules and siting-analysis reports?
Technical details
Editor’s note
In plain speech this problem is called “EV charging plan”, “charging-station rollout” or “fleet charging management”. In the academic literature it carries two distinct but linked names: the operational daily decision is EV Charging Scheduling, the infrastructure investment is Charging-Station Location Planning. A third emerging layer is Vehicle-to-Grid (V2G) — sale of vehicle battery back to the grid. Without telling the three names apart, when a charge-management platform firm pitches a module you cannot test which decision layer is truly being optimised — “charge-point tracking” is one thing, “MILP-based scheduling” another.
Distinguish these three problems from other energy OR problems: Unit commitment + economic dispatch (#029) is the generation side — plant on/off and MW output. Demand response + peak shaving (#020) is consumer-side general load shaping — when flexible loads run. EV charging scheduling (#052) sits between the two: both demand shaping (which vehicle to charge when) and infrastructure planning (how many charging points where). EV has a distinctive feature: the deadline constraint of each vehicle (ready by the next morning) — sets it apart from demand response. And battery aging (cycle aging) — an operational cost item absent in the other energy problems.
Most-skipped point in the sector: battery aging cost. Field calculations usually contain only electricity price and peak-draw penalty; they skip that the battery ages faster under fast charge (DC fast 50-350 kW), with 15-30% shorter cycle life. Battery replacement cost is 25-40% of a new vehicle’s value; when the total cost of ownership (TCO — Total Cost of Ownership) does not include it, a “use cheap fast charging always” policy looks optimal in the short run but a battery-replacement bill hits the operator in 3-5 years. Modern literature adds battery aging cost to the objective; a so-called “optimal” schedule built without it turns into a long-term loss in the field. Second skipped point: grid-connection feasibility. After a decision is made for a new charging site, distribution-utility transformer + line build can take 6-18 months; if the siting plan does not model time, the station is built but “no electricity arrives”. A stochastic siting model must have grid-connection lead time as a scenario variable.
Step-by-step path — for the SMB
Stage 1 — Measure first, plan after. At least 6 months of vehicle + charging data: per vehicle, daily usage profile (trip duration, distance, depot in-out, average SoC cycle); per charge point, daily utilisation curve (hourly on/off, kW draw profile, OCPP event logs); monthly electricity bill and tariff structure (time-of-use windows, peak-draw calculation). Grid inventory: connected transformer power, instantaneous peak-draw limit, contractual penalty curve. Battery inventory: per vehicle, battery capacity, year of manufacture, warranty cycle count, current health.
Stage 2 — Extract the knowledge capital. Vehicle availability by hour (fleet type: courier leaves 8am back 10pm; taxi runs 24/7; municipal service does 100-200 km/day). Time-of-use tariff windows and hourly price curve. Transformer-capacity limit and seasonal variation. Battery aging curve (cycle-life loss in fast vs slow charge — manufacturer + literature). For public-site operators: customer-arrival distribution, mean dwell time, balking rate (saw the queue and left).
Stage 3 — Pilot. 8-12 weeks. For a subset (30-50% of fleet, or one depot, or 3-5 public sites) run the MILP schedule output in parallel with the current heuristic allocation. The decision stays with the operations centre; MILP gives a recommendation. Success criterion set in advance: monthly electricity cost -10% minimum, peak-draw penalty -50% minimum, battery-aging tracking included.
Stage 4 — Rollout. 9-15 months to full fleet + all public sites + siting planning + V2G module integration. Multi-brand charge-point compatibility via OCPP. Quarterly committee: schedule performance, battery-aging trend, utilisation curve, investment decisions (open/close sites, capacity uplift), regulatory updates (TR’s EV-incentive policy, local-production support rules).
Risks — what can go wrong
- Battery aging cost absent from the objective. Optimising only electricity price + peak-draw penalty makes the “optimal” plan that heavily uses fast charge look great in the short run; the battery-replacement bill upends TCO in 3-5 years. Modern models include it; the module’s coverage of this curve must be tested before procurement.
- Grid-connection delay. For new public sites the distribution-utility transformer + line build can take 6-18 months; a siting plan that does not model time leaves the station built but with no power. Stochastic siting models must have grid-connection lead time as a scenario variable.
- Regulatory change. TR’s EV incentives (local-production support rule, tax exemptions, station-permit conditions) are updated annually; EU side has the carbon border adjustment mechanism (CBAM) and EV-penetration targets at the 2030/2035 horizon. Investment horizon is 5-10 years; a single-scenario plan is brittle. Scenario-based analysis and phased investment are essential.
- Single-supplier charge-management-system lock-in. Although OCPP is an open protocol, without a contract clause for export of billing rules + tariff parameters + battery-aging curves leaving the system means losing the operator’s operational memory.
- V2G arbitrage revenue overstated. Marketing claims X-Y% V2G revenue; net arbitrage gain after battery aging cost is typically 2-3x lower. Net accounting must be scenario-based.
Solution method — a technical view
| Approach | Typical scale | Solve time | Guaranteed optimum? |
|---|---|---|---|
| Heuristic priority rule (start-at-night, SoC-threshold) | Small fleet, <30 EVs | instant | No, 50-70% optimum |
| MILP — operational scheduling | Mid fleet, 30-200 EVs, 24-48h horizon | minutes-hours | Yes (within bound) |
| Lagrangian relaxation / decomposition | Large fleet, 200-1000 EVs | hours | Good upper-lower bound |
| Rolling-horizon MPC | Dynamic (taxi, rental), hourly re-plan | minutes per hour | Practical quality |
| Online policy (insertion heuristic) | Public site — live arrivals | seconds | No, transitional quality |
| p-median / maximum-coverage (siting) | Regional infrastructure, 5-50 candidates | minutes-hours | Yes (within bound) |
| Stochastic siting (two-stage) | EV-penetration uncertainty | hours | Yes (per scenario) |
| V2G stochastic MPC | Arbitrage + battery aging | hours | Practical quality |
Objective function choice:
- Objective 1 — Total electricity cost minimum: Simplest, but ignores battery aging.
- Objective 2 — Electricity cost + battery aging cost minimum: TCO-aligned, accurate but data-heavy.
- Objective 3 — Peak-draw penalty minimum (suppress kW peak): For fleets close to the contractual kW limit.
- Objective 4 — Utilisation maximum (public site): Operator revenue-oriented siting.
- Objective 5 — Net V2G arbitrage revenue minus battery aging maximum: Advanced operation.
Multi-objective: weighted sum or hierarchical (first deadline + peak-draw, then electricity cost, then battery aging).
Academic sources
Listed in the page frontmatter under sources.
Sources
- Sortomme, E., Hindi, M. M., MacPherson, S. D. J. and Venkata, S. S. (2011). Coordinated charging of plug-in hybrid electric vehicles to minimize distribution system losses. IEEE Transactions on Smart Grid, 2(1), 198-205. Foundational coordinated-EV-charging paper.
- Sweda, T. M. and Klabjan, D. (2012). An agent-based decision support system for electric vehicle charging infrastructure deployment. IEEE Vehicle Power and Propulsion Conference (VPPC). Foundational reference for charging-station siting.
- Mukherjee, J. C. and Gupta, A. (2015). A review of charge scheduling of electric vehicles in smart grid. IEEE Systems Journal, 9(4), 1541-1553. EV charging scheduling survey.
- Rahman, I., Vasant, P. M., Singh, B. S. M., Abdullah-Al-Wadud, M. and Adnan, N. (2016). Review of recent trends in optimization techniques for plug-in hybrid, and electric vehicle charging infrastructures. Renewable and Sustainable Energy Reviews, 58, 1039-1047. Optimization techniques survey.
- Ko, J. and Shang, J. S. (2010). Locating refueling stations for alternative-fuel vehicles. Transportation Research Part C, 18(5), 700-710. Alternative-fuel station siting.
- YÖK Thesis Center — keywords: ’elektrikli araç şarj’, ‘EV şarj çizelgeleme’ or ‘şarj istasyonu yerleşim’ — 15+ theses (a growing area). tez.yok.gov.tr
Glossary
- EV Charging Scheduling
- Operational OR problem of deciding when and at what power level to charge each EV in a fleet or queue, subject to deadline, grid power limits, battery SoC bounds, time-of-use pricing and battery-aging cost.
- Charging-Station Location Planning
- Strategic OR problem of selecting locations, types (AC slow vs DC fast) and counts of public EV charging stations to deploy in a region, balancing capital and operational cost against expected user coverage and waiting time.
- MIP
- An optimization model where some decision variables are forced to be whole numbers (e.g. number of trucks, number of shifts).
- Demand Response
- An energy-management approach that optimizes when loads run on the consumer side, based on the tariff, system conditions, or incentives.