⚡ electrical engineering

Multi-objective scheduling of highway PV-storage-charging microgrid clusters with EV-load cascades

This study proposes a multi-objective scheduling framework for highway PV-storage-charging microgrid clusters that integrates a spatiotemporal EV-load cascade model to account for dynamic travel behaviors, successfully generating a feasible Pareto frontier that minimizes both operating costs and carbon emissions while ensuring strict constraint compliance.

Original authors: Yanming Sun, Mohan Zhao, Pihong Gong

Published 2026-07-01

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Yanming Sun, Mohan Zhao, Pihong Gong

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a busy highway as a long river of cars. Along this river, there are three "service stations" (let's call them Station A, Station B, and Station C). Each station has electric vehicle (EV) chargers, solar panels on the roof, and a giant battery bank (like a massive power bank for the whole station).

The goal of this study is to figure out the best way to run these three stations so they save money and produce less pollution, while dealing with a tricky problem: traffic jams change where people want to charge.

Here is the story of how the researchers solved this, explained simply:

1. The "Domino Effect" of Traffic

Usually, planners think of charging demand like a fixed schedule: "Station A will have 100 cars at 5 PM." But in real life, drivers are smart. If Station A has a huge line of cars waiting (a queue), drivers might decide, "I'll skip this one and drive to Station B instead."

The researchers built a digital model that acts like a "domino effect" simulator.

The Trigger: If Station A gets too crowded, the "waiting time" gets too long.
The Reaction: Drivers check their battery. If they have enough juice to reach Station B safely, they skip A and go to B.
The Result: Station B suddenly gets a surge of cars it wasn't expecting. If Station B also gets too crowded, some might try to skip to Station C.

The study found that this "skipping" mostly happens between the first two stations (A and B). Very few cars skip all the way to Station C because they run out of battery or get too tired of driving further just to save a few minutes.

2. The Three Stations as One Big Team

Instead of managing Station A, B, and C separately, the researchers treated them as one giant team (a "microgrid cluster").

The Solar Panels: They generate free, clean energy during the day.
The Giant Battery: This is the team's "savings account." It stores extra solar power when the sun is shining and releases it when the cars arrive at night.
The Grid: This is the connection to the main city power plant.

The challenge was to tell the battery when to charge and when to discharge to keep costs low and pollution down, while making sure no station runs out of power.

3. The "No Cheating" Rules

The researchers used a smart computer algorithm (a digital optimizer) to find the best plan. But they added strict "no cheating" rules to make sure the plan actually works in the real world:

The "Empty Battery" Rule: You can't drain the battery completely by the end of the day. It must be refilled to its starting level so it's ready for tomorrow. (Imagine you can't spend your entire paycheck today if you need money for breakfast tomorrow).
The "No Export" Rule: The battery can't sell power back to the main city grid just to make a quick profit. It must only use its power to charge the cars.
The "Safety Margin" Rule: The battery can't get too full or too empty, or it might break.

4. The Result: A Balanced Plan

The computer generated 500 different "perfect" plans. The researchers picked the one that offered the best balance between saving money and saving the planet.

What happened in the winning plan?

Money Saved: The daily cost was about 13,516 CNY.
Pollution Reduced: The daily carbon emissions were about 12,068 kg of CO2.
How it worked: The battery charged up during the day when the solar panels were working hard. Then, in the evening when the cars arrived, the battery released that stored energy to charge the cars. This meant the stations didn't have to buy as much expensive, dirty power from the main grid.
The Safety Check: The plan passed every single test. The battery never ran dry, never overflowed, and the power lines never got overloaded.

The Big Takeaway

This study shows that you can't just look at a charging station in isolation. You have to look at the whole highway corridor. When a traffic jam happens at the first stop, it sends a ripple effect down the road, changing where the cars go. By understanding this "ripple" and managing the solar panels and batteries as a single team, highway stations can run cheaper, cleaner, and more reliably.

The researchers proved that if you build your plan with these "real-world" traffic behaviors and strict safety rules, you get a solution that isn't just a number on a screen, but a plan that could actually work tomorrow.

Technical Summary: Multi-objective Scheduling of Highway PV-Storage-Charging Microgrid Clusters with EV-Load Cascades

Problem Statement
Highway service areas are evolving into coupled transportation-energy nodes where electric vehicle (EV) charging demand is often modeled as a fixed, exogenous load profile. This traditional approach fails to capture the dynamic reality of long-distance travel, where queue pressure, remaining state of charge (SOC), travel delays, and station-skipping decisions cause charging demand to shift spatially along a corridor. Furthermore, existing scheduling models often treat EV demand as static, ignoring the "station-skipping" behavior where drivers bypass congested upstream stations to charge downstream. This disconnect creates a gap between traffic-side behavioral responses and energy-side microgrid operations. Additionally, multi-objective optimization for microgrids frequently relies on soft-penalty methods, which can yield solutions that appear optimal in objective space (cost/emissions) but violate critical physical constraints such as SOC bounds, transformer capacity, or terminal SOC requirements, rendering them operationally infeasible.

Methodology
The study proposes a two-layer framework linking a spatiotemporal EV-load cascade model with a corridor-level aggregated PV-storage-charging microgrid scheduling model.

EV-Load Cascade Model:
- Behavioral Logic: The model treats station-skipping as a constrained behavioral response rather than a free redistribution. It utilizes a binary logit formulation to calculate skipping probabilities based on generalized costs. These costs weigh queue waiting time, inter-station travel delay, remaining SOC, and range risk.
- Physical Constraints: A hard constraint ensures skipping is only permitted if the remaining SOC is sufficient to reach the next station while maintaining a safety margin.
- Cascade Mechanism: Skipped demand from an upstream station (e.g., Station A) is delayed by travel time and added to the arrival rate of the downstream station (e.g., Station B), creating a spatiotemporal load cascade.
Microgrid Scheduling Model:
- System Scope: Three consecutive service areas (A, B, C) are aggregated into a single corridor-level microgrid cluster to evaluate the joint operation of shared PV, Energy Storage Systems (ESS), and transformers.
- Objectives: The model minimizes two conflicting objectives: daily net operating cost (including electricity purchase, sale, and ESS degradation) and daily net carbon emissions.
- Constraints: The scheduling horizon is 24 hours (144 ten-minute intervals). Key constraints include SOC bounds ([0.1, 0.9]), transformer active-power limits, a terminal SOC requirement (final SOC $\ge$ initial SOC to prevent inter-day energy borrowing), and a prohibition on ESS-to-grid export (preventing battery-mediated arbitrage).
Optimization Workflow:
- Algorithm: The Non-dominated Sorting Genetic Algorithm II (NSGA-II) is employed.
- Feasibility Preservation: Unlike standard approaches, this workflow integrates feasibility-oriented initialization, forward constraint repair (clipping power limits, adjusting SOC sequentially), and fully feasible Pareto filtering.
- Selection: Only schedules satisfying all hard constraints (zero violations in SOC, transformer capacity, and terminal SOC) are retained. A compromise solution is selected from this "fully feasible" front using a normalized membership score with equal weights for cost and carbon.

Key Contributions

Coupled Modeling: The study explicitly links queue-induced transport responses (waiting time, skipping) with downstream microgrid load inputs, moving beyond fixed load profiles.
Corridor-Level Aggregation: It demonstrates that while demand redistribution occurs at the node level, operational scheduling must account for the aggregated resources (PV, ESS, grid capacity) of the entire corridor.
Strict Feasibility Handling: The paper introduces a workflow that enforces hard constraints (terminal SOC, no ESS-to-grid export) during the optimization process, ensuring the resulting Pareto front consists entirely of executable schedules rather than soft-penalty approximations.
Reproducibility: The study provides a fully reproducible NSGA-II workflow with fixed random seeds and specific repair mechanisms, yielding 500 fully feasible Pareto solutions.

Results
The case study focuses on a corridor with three service areas.

Load Cascade Dynamics: The results show an asymmetric load transfer. Service Area A (upstream) experiences significant queue pressure (waiting times up to 21.46 min), leading to a skipping probability average of 0.215. This demand is largely absorbed by Service Area B (downstream), which sees a load increase. Transfer to Service Area C remains limited (average skip probability 0.041) due to remaining SOC constraints and travel distance.
Optimization Outcomes: From the 500 fully feasible Pareto solutions, a compromise was selected:
- Net Operating Cost: 13,515.78 CNY/day.
- Net Carbon Emissions: 12,067.92 kg CO₂/day.
- Operational Metrics: The ESS operates primarily to shift energy within the corridor (charging during PV availability, discharging during evening peaks) rather than exporting to the grid. Total grid export, PV export, and ESS-to-grid export are all 0.00 kWh.
- Constraint Satisfaction: The solution maintains SOC within [0.2577, 0.9000], returns the terminal SOC to the initial value of 0.5000, and records zero violations for SOC bounds, transformer capacity, and ESS-to-grid export.

Significance and Claims
The paper claims that treating highway charging demand as a spatially transferable process, shaped by queue conditions and range constraints, provides a clearer interface between traffic dynamics and energy system operations. The significance of the work lies not in the novelty of the optimizer (NSGA-II) itself, but in the reproducible coupling of queue-driven load cascade modeling with feasibility-preserving corridor-level scheduling.

The authors emphasize that by enforcing terminal SOC recovery and prohibiting ESS-to-grid export, the reported cost-emission trade-off is more auditable and operationally interpretable. The selected compromise is presented not as an unconstrained numerical optimum, but as a feasible schedule that respects the physical limits of the microgrid and the behavioral realities of EV drivers. The study concludes that such a framework supports technically interpretable, constraint-feasible, and low-carbon operation of highway corridor microgrid clusters, though it acknowledges limitations regarding fixed carbon factors, single-day horizons, and the lack of AC power-flow constraints.