Efficient mid-term forecasting of hourly electricity load using generalized additive models

Overview

• Accurate mid-term (weeks to one year) hourly electricity load forecasts are essential for strategic decision-making in power plant operation, ensuring supply security and grid stability, and energy trading.
• While numerous models effectively predict short-term (hours to a few days) hourly load, mid-term forecasting solutions remain scarce.
• Modeling challenges in mid-term load forecasting include capturing daily, weekly, and annual seasonal and autoregressive effects, as well as weather, holiday, and socio-economic non-stationarities in the data.
• To address these challenges, the authors propose a novel forecasting method using Generalized Additive Models (GAMs) built from interpretable P-splines and enhanced with autoregressive post-processing.

Plain English Explanation

Predicting how much electricity people will use in the coming weeks and months is critical for power companies. They need these forecasts to make decisions about running power plants, maintaining a stable electricity grid, and trading energy on the market. While there are good models for short-term (day-to-day) electricity load forecasting, models for medium-term (weeks to a year) forecasting are still limited.

Forecasting medium-term electricity use is challenging because it involves capturing complex patterns in the data, like daily, weekly, and yearly cycles, as well as the effects of weather, holidays, and changes in the economy over time. To tackle these challenges, the researchers developed a new forecasting method using a type of statistical model called Generalized Additive Models (GAMs). Their model uses smooth temperature data, seasonal trends, and holiday effects to make accurate medium-term electricity load predictions.

The researchers tested their model on load data from 24 European countries and found that it significantly outperformed other state-of-the-art methods. Importantly, the model is also highly interpretable, meaning they can clearly see how much each component (like weather or holidays) is influencing the predicted electricity load. This makes it a powerful tool for power companies to use in their decision-making.

Technical Explanation

The proposed forecasting method uses Generalized Additive Models (GAMs) built from interpretable P-splines and enhanced with autoregressive post-processing. This model takes as input smoothed temperatures, Error-Trend-Seasonal (ETS) modeled non-stationary states, a nuanced representation of holiday effects with weekday variations, and seasonal information.

The model was evaluated on load data from 24 European countries. The analysis demonstrated that the proposed model not only has significantly enhanced forecasting accuracy compared to state-of-the-art methods, but also offers valuable insights into the influence of individual components on predicted load, given its full interpretability.

The model's ability to achieve performance akin to day-ahead Transmission System Operator (TSO) forecasts in fast computation times of a few seconds for several years of hourly data underscores its potential for practical application in the power system industry.

Critical Analysis

The paper provides a comprehensive evaluation of the proposed forecasting method, including comparisons to state-of-the-art approaches and analysis of the interpretability of the model. However, the authors do not discuss any potential limitations or caveats of their approach.

For example, the model's performance may be sensitive to the quality and availability of the input data, such as weather and holiday information. Additionally, the authors do not explore how the model would perform in scenarios with significant structural changes in the electricity market or consumer behavior, which could impact the stationarity assumptions.

Further research could investigate the model's robustness to such changes, as well as explore the potential for incorporating probabilistic forecasting to provide a more complete picture of the uncertainty in the medium-term electricity load projections.

Conclusion

The proposed Generalized Additive Model with autoregressive post-processing represents a significant advancement in medium-term electricity load forecasting. By accurately capturing the complex seasonal, weather, and holiday effects influencing electricity demand, while maintaining high interpretability, this model offers power system operators and energy traders a valuable tool for strategic decision-making. The authors' rigorous evaluation and demonstration of the model's practical applicability underscore its potential to improve the reliability and efficiency of power systems worldwide.