Electrical Behavior Association Mining for Household ShortTerm Energy Consumption Forecasting

TL;DR

This work tackles day-ahead household short-term energy forecasting by combining probabilistic association mining of pairwise appliance behavior with spectral clustering to form appliance clusters, followed by a CNN-GRU forecasting model that leverages cluster-based inputs. The probabilistic matrix $Q$ captures pairwise appliance relationships within a target window $T_e$, enabling clustering that reduces model complexity while preserving predictive power. Using the AMPds2 dataset, the approach yields improved RMSE and MAE over direct total-load forecasts and provides insights into feature correlations and the limits of day-ahead predictability for intermittently used appliances. The method offers a scalable, interpretable pathway to enhance home energy management systems and informs future work on intraday forecasting and adaptive association updates.

Abstract

Accurate household short-term energy consumption forecasting (STECF) is crucial for home energy management, but it is technically challenging, due to highly random behaviors of individual residential users. To improve the accuracy of STECF on a day-ahead scale, this paper proposes an novel STECF methodology that leverages association mining in electrical behaviors. First, a probabilistic association quantifying and discovering method is proposed to model the pairwise behaviors association and generate associated clusters. Then, a convolutional neural network-gated recurrent unit (CNN-GRU) based forecasting is provided to explore the temporal correlation and enhance accuracy. The testing results demonstrate that this methodology yields a significant enhancement in the STECF.

Figures (3)

• Figure 1: Model structure of household STECF
• Figure 2: The flowchart of the proposed method
• Figure 3: Adjacency matrix (assocation probability matrix) of pairwise appliances