TL;DR
This paper tackles the critical issue of load shedding in developing countries, where insufficient power generation leads to frequent, often unfair, electricity cutoffs. The authors propose and evaluate four novel heuristic algorithms designed to allocate electricity more equitably among households, aiming to minimize discomfort and maximize overall welfare.
What Problem Does This Paper Address?
Developing countries often face a significant gap between electricity demand and generation capacity. To prevent grid collapse, utilities resort to "load shedding," which involves systematically disconnecting parts of the power grid. While this maintains grid stability, it can lead to prolonged and unevenly distributed power outages for households, causing inconvenience and hindering development. Current load shedding methods often prioritize grid stability over fairness, leading to situations where some communities are left without power for extended periods. This paper addresses the need for fair and optimized load shedding strategies that consider the needs and comfort of individual households.
Methods
The core of this paper lies in the development and evaluation of four heuristic algorithms for household-level load shedding. These algorithms aim to achieve fairness by considering various criteria:
- Grouper Algorithm: This baseline algorithm groups agents (households) whose combined consumption can offset the deficit and then selects the group with the minimum total number of disconnections. It's a more traditional approach focusing on group-level optimization.
- Consumption-Sorter Algorithm: This heuristic prioritizes fairness by ensuring agents are disconnected in a way that keeps the number of disconnections per agent as similar as possible. It sorts agents by their consumption levels and selects them sequentially.
- Random-Selector Algorithm: Similar to the Consumption-Sorter, this algorithm aims for fairness in disconnection frequency but selects agents randomly, avoiding bias based on consumption.
- Cost-Sorter Algorithm: This algorithm introduces comfort costs (derived from historical consumption patterns) into the fairness equation. It prioritizes agents with lower comfort costs for disconnection, while still aiming for parity in disconnection frequency.
The algorithms are designed to operate at the household level, allowing for more granular control and potentially maximizing revenue for electricity providers by shedding only the necessary amount of load.
Datasets
The paper utilizes a simulated dataset for developing countries, specifically Nigeria, by adapting data from the USA.
- Pecan Street Dataport: Disaggregated electricity consumption data from over 700 households in the USA was used as a basis.
- Adaptation for Nigeria: The authors identified similarities in appliance usage between the USA and Nigeria and focused on common appliances like air conditioners, washing machines, light bulbs, and refrigerators. Usage data was collected over 13 weeks of summer in Austin, Texas, to simulate weather conditions relevant to Nigeria. This adapted dataset forms a novel benchmark for simulating developing country energy consumption for load shedding research.
Key Insights and Ideas
- Household-level shedding is key: Shedding load at the household level, rather than at the substation level, allows for a much closer match to the deficit, potentially increasing revenue for suppliers and reducing wasted electricity.
- Fairness is multi-faceted: The paper highlights that fairness in load shedding can be measured by various metrics, including the number of disconnections, individual comfort costs, and the overall system's social welfare (utilitarian and egalitarian objectives).
- Comfort costs matter: Incorporating historical consumption patterns to estimate "comfort costs" provides a more nuanced understanding of the impact of disconnections on individual households.
- No single "best" algorithm: Different heuristics excel under different fairness objectives, suggesting that the optimal approach depends on the specific priorities of the utility and the context of the developing country.
Key Results
The paper presents comparative analyses of the four heuristics across different fairness objectives:
- Comfort Costs:
- Utilitarian: The Grouper algorithm performed best, indicating the lowest aggregated discomfort.
- Egalitarian: The Consumption-Sorter algorithm achieved the lowest highest individual comfort cost.
- Envy-freeness: The Consumption-Sorter algorithm also showed the lowest envy between agents.
- Number of Times Agents are Disconnected:
- Utilitarian: The Grouper algorithm again showed the lowest aggregated number of disconnections.
- Egalitarian & Envy-freeness: The Consumption-Sorter and Random-Selector algorithms performed best, indicating a more even distribution of disconnections.
- Other Performance Considerations:
- The paper demonstrates that household-level shedding leads to a closer match between the deficit and the shed load, suggesting efficiency.
- Figure 3 shows that the Grouper algorithm provides the best ratio of agents disconnected per kWh shed, while Consumption-Sorter and Cost-Sorter show more variation.
Limitations and Future Work
- Data Simulation: The reliance on simulated data adapted from the USA, while a necessary step due to the lack of granular data in Nigeria, is a limitation. Future work could involve collecting and utilizing actual disaggregated consumption data from developing countries.
- Model Complexity: The current models are heuristic-based. Future work could explore more complex optimization techniques, such as goal programming, to incorporate social welfare objectives as constraints.
- Combining Heuristics: The authors suggest that combining components of different heuristics could lead to even better outcomes, tailored to specific environmental and system characteristics.
- Resource Allocation: The paper notes that the developed heuristics could serve as a benchmark for allocating other scarce resources, such as water, suggesting a broader applicability of the fair allocation principles.
Summary updated