Explaining Deep Learning-based Anomaly Detection in Energy Consumption Data by Focusing on Contextually Relevant Data

TL;DR

This work tackles the explainability gap in deep learning–based anomaly detection for energy consumption by introducing a context-aware explanation framework. By selecting background data for SHAP calculations through a weighted cosine similarity that emphasizes contextually relevant features, the approach stabilizes explanations while reducing computational burden. Across five datasets, ten models, and five XAI methods, the method lowers explanation variability by about $38\%$ on average, with robust statistical support. The contribution advances trust and practical deployment of DL-based anomaly detection in energy systems by delivering consistent, interpretable insights into the drivers of anomalies.

Abstract

Detecting anomalies in energy consumption data is crucial for identifying energy waste, equipment malfunction, and overall, for ensuring efficient energy management. Machine learning, and specifically deep learning approaches, have been greatly successful in anomaly detection; however, they are black-box approaches that do not provide transparency or explanations. SHAP and its variants have been proposed to explain these models, but they suffer from high computational complexity (SHAP) or instability and inconsistency (e.g., Kernel SHAP). To address these challenges, this paper proposes an explainability approach for anomalies in energy consumption data that focuses on context-relevant information. The proposed approach leverages existing explainability techniques, focusing on SHAP variants, together with global feature importance and weighted cosine similarity to select background dataset based on the context of each anomaly point. By focusing on the context and most relevant features, this approach mitigates the instability of explainability algorithms. Experimental results across 10 different machine learning models, five datasets, and five XAI techniques, demonstrate that our method reduces the variability of explanations providing consistent explanations. Statistical analyses confirm the robustness of our approach, showing an average reduction in variability of approximately 38% across multiple datasets.

Figures (8)

• Figure 1: Methodology overview
• Figure 2: Illustration of the sliding window approach for sequence-to-sequence model training with 48 Input sequence and 24 output sequence
• Figure 3: Illustration of the anomaly detection process using IQR
• Figure 4: Prediction results for the LSTM model, showing 10% of the test residential dataset.
• Figure 5: Global Feature Importance Line Plot of Sequence Windows for LSTM Model on residential dataset