TL;DR
This paper presents a multi-objective optimization framework for scheduling grid-connected microgrids that integrate solar power, energy storage, and flexible loads. By leveraging the NSGA-II algorithm, the authors demonstrate how to simultaneously slash operational costs, reduce carbon footprints, and stabilize the power interface between the microgrid and the main distribution network.
What Problem Does This Paper Address?
Renewable energy (RE) sources like solar and wind are essential for a sustainable future, but they are notoriously "fickle"—their output is intermittent and unpredictable. When a microgrid heavily reliant on RE connects to the main power grid, it can cause significant stability issues due to rapid power fluctuations.
Microgrid operators face a difficult "trilemma":
- Economy: How to keep operational costs low.
- Environment: How to minimize the carbon footprint (especially when the main grid relies on coal).
- Stability: How to minimize "tie-line" power fluctuations to ensure the main distribution network remains safe and stable.
While previous research often focused on one or two of these goals, this paper treats energy storage and "demand-side management" (asking users to shift or cut power) as active, dispatchable resources to solve all three simultaneously.
Methods
The authors established a multi-objective optimization model with three primary goals:
- $f_1$ (Operation Cost): Includes electricity purchase costs, battery degradation/operation costs, and the compensation paid to users for interrupting or transferring their loads.
- $f_2$ (Environmental Cost): Calculated based on the carbon emissions associated with purchasing power from the coal-heavy main grid.
- $f_3$ (Tie-line Fluctuation): A metric measuring the "jerkiness" of power flow between the microgrid and the distribution network.
To solve this non-linear, multi-constraint problem, they employed the Non-dominated Sorting Genetic Algorithm-II (NSGA-II). Because these three goals often conflict (e.g., the cheapest power might be the "dirtiest"), the algorithm produces a Pareto frontier—a set of optimal trade-off solutions. Finally, a fuzzy membership function is used to help decision-makers select the single best "compromise" solution from that frontier.
Datasets
The study utilized the EPRI (Electric Power Research Institute) microgrid structure as its benchmark model. The simulation was conducted using a joint platform of MATLAB and OpenDSS (an open-source electricity distribution system simulator).
- Inputs: 24-hour profiles for photovoltaic (PV) output, total load demand, and Time-of-Use (TOU) electricity pricing.
- Load Data: Specific parameters for three classes of interruptible and transferable loads, including compensation costs and allowable durations.
Key Insights and Ideas
- Load as a "Virtual Battery": One of the core ideas is treating demand-side management not just as a consumer behavior, but as a dispatchable resource. By compensating users to shift (transferable) or cut (interruptible) power, the microgrid gains flexibility similar to physical battery storage.
- The "Good Neighbor" Policy: By including "tie-line fluctuation" as a primary objective, the model ensures the microgrid doesn't just look out for its own bottom line, but actively works to prevent technical headaches for the larger distribution network.
- Cooperative Optimization: The results highlight that "source-storage-load" coordination is superior to optimizing any single component. The battery charges when load is low/PV is high and discharges during peaks, while flexible loads are shifted to "fill the valleys" of the demand curve.
Key Results
The simulation proved that the optimized schedule significantly outperformed the baseline (pre-optimization) scenario across all metrics:
- Operation Cost: Reduced by ~5.8% ($14,309 → $13,474).
- Environmental Cost: Reduced by ~1.4% ($12,649 → $12,468).
- Tie-line Fluctuation: The most dramatic improvement, dropping by 31.7% (2,314 kW → 1,580 kW).
Qualitatively, the probability density curves in the paper show a much "tighter" distribution of power changes after optimization, meaning the microgrid's impact on the main grid became much more predictable and stable.
Limitations and Future Work
- CAPEX Ignored: The authors explicitly state they did not consider the initial investment costs (Capital Expenditure) for PV and energy storage equipment, focusing strictly on daily operational optimization.
- Deterministic Assumptions: The model appears to use deterministic forecasts for PV and load. In reality, weather is unpredictable; future iterations could incorporate stochastic optimization to handle real-time uncertainty.
- Scale: The study focuses on a single microgrid. Future research could explore the "multi-microgrid" scenario, where several microgrids coordinate with each other and the main grid simultaneously.
Summary updated