A Lightweight DL Model for Smart Grid Power Forecasting with Feature and Resolution Mismatch

2025-10-19

AI Summary

Addressing the challenges of high noise, severe missingness, heterogeneous features, and limited contextual information in sensor data for short-term load forecasting in smart grids, this paper proposes a lightweight two-stage temporal modeling framework. First, a dual-mode imputation strategy—combining mean imputation with polynomial regression—is introduced, followed by temperature-aware fine-grained normalization. Second, a GRU-LSTM hybrid sequence-to-point model is designed to accommodate asymmetric input lengths while preserving temperature trend consistency. Evaluated on real-world high-frequency load data, the model achieves an RMSE of 601.9 W, MAE of 468.9 W, and prediction accuracy of 84.36%, with low inference latency, strong generalization capability, and practical deployability on edge devices. The key contributions are a noise-robust missing-data handling paradigm and a lightweight, edge-optimized temporal architecture tailored for resource-constrained forecasting scenarios.

Technology Category

• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Machine Learning: Learning on the Edge & Model Compression
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Energy management for devices in mobile Web and WoT environments
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

How can short-term energy consumption be accurately forecasted when sensor data is noisy, incomplete, and lacks contextual richness? This question guided our participation in the extit{2025 Competition on Electric Energy Consumption Forecast Adopting Multi-criteria Performance Metrics}, which challenged teams to predict next-day power demand using real-world high-frequency data. We proposed a robust yet lightweight Deep Learning (DL) pipeline combining hourly downsizing, dual-mode imputation (mean and polynomial regression), and comprehensive normalization, ultimately selecting Standard Scaling for optimal balance. The lightweight GRU-LSTM sequence-to-one model achieves an average RMSE of 601.9~W, MAE of 468.9~W, and 84.36% accuracy. Despite asymmetric inputs and imputed gaps, it generalized well, captured nonlinear demand patterns, and maintained low inference latency. Notably, spatiotemporal heatmap analysis reveals a strong alignment between temperature trends and predicted consumption, further reinforcing the model's reliability. These results demonstrate that targeted preprocessing paired with compact recurrent architectures can still enable fast, accurate, and deployment-ready energy forecasting in real-world conditions.

Problem

• Accurately forecasting short-term energy consumption with noisy sensor data
• Addressing feature and resolution mismatches in smart grid power forecasting
• Developing lightweight models for real-world energy prediction with incomplete inputs

Innovation

• Lightweight GRU-LSTM model for sequence-to-one prediction
• Dual-mode imputation using mean and polynomial regression
• Hourly downsizing and normalization for feature preprocessing