A Natural Gas Consumption Forecasting System for Continual Learning Scenarios based on Hoeffding Trees with Change Point Detection Mechanism

TL;DR

This paper tackles nonstationary natural gas consumption forecasting under take-or-pay contracts by introducing a modular continual-learning pipeline that couples data-stream processing with change point detection. It leverages a Hoeffding Tree Regressor as the online forecasting core and uses the Pruned Exact Linear Time (PELT) algorithm to partition time series into segments, enabling CP-divided model collections and two selection schemes (Pure and Mixed). Empirical results on eight years of Prague energy data show that CP-aware HT approaches, especially HT-PCPDMC with fewer change points, outperform baselines and deep learning models, illustrating strong robustness for continual learning in dynamic energy environments. The work demonstrates the practicality of change-point driven model aggregation for reliable multistep forecasting in real-world, nonstationary settings with potential applications beyond natural gas to other energy domains.

Abstract

Forecasting natural gas consumption, considering seasonality and trends, is crucial in planning its supply and consumption and optimizing the cost of obtaining it, mainly by industrial entities. However, in times of threats to its supply, it is also a critical element that guarantees the supply of this raw material to meet individual consumers' needs, ensuring society's energy security. This article introduces a novel multistep ahead forecasting of natural gas consumption with change point detection integration for model collection selection with continual learning capabilities using data stream processing. The performance of the forecasting models based on the proposed approach is evaluated in a complex real-world use case of natural gas consumption forecasting. We employed Hoeffding tree predictors as forecasting models and the Pruned Exact Linear Time (PELT) algorithm for the change point detection procedure. The change point detection integration enables selecting a different model collection for successive time frames. Thus, three model collection selection procedures (with and without an error feedback loop) are defined and evaluated for forecasting scenarios with various densities of detected change points. These models were compared with change point agnostic baseline approaches. Our experiments show that fewer change points result in a lower forecasting error regardless of the model collection selection procedure employed. Also, simpler model collection selection procedures omitting forecasting error feedback leads to more robust forecasting models suitable for continual learning tasks.

Figures (7)

• Figure 1: The proposed multistep forecasting pipeline with CL support.
• Figure 2: Natural gas consumption (endogenous variable) from January 1, 2013, to December 31, 2020 (years are divided by colors).
• Figure 3: Electricity load (endogenous variable) from January 1, 2013, to December 31, 2020 (years are divided by colors).
• Figure 4: Detected change points by PELT algorithm with Low settings in selected years of the natural gas consumption data.
• Figure 5: Detected change points by PELT algorithm with Low settings in selected years of the electricity load data.