Automated Energy-Aware Time-Series Model Deployment on Embedded FPGAs for Resilient Combined Sewer Overflow Management

TL;DR

This paper presents an energy-efficient, hardware-aware framework for deploying accurate time-series forecasting models directly on embedded FPGAs to improve sewer overflow management during extreme weather events.

Contribution

It introduces a novel automated deployment pipeline that optimizes lightweight Transformer and LSTM models for energy and accuracy on FPGA edge devices.

Findings

The 8-bit Transformer achieves high accuracy with low energy (0.370 mJ per inference).

The 8-bit LSTM consumes over 40x less energy but with slightly reduced accuracy.

The framework enables resilient, local forecasting for sewer systems during communication outages.

Abstract

Extreme weather events, intensified by climate change, increasingly challenge aging combined sewer systems, raising the risk of untreated wastewater overflow. Accurate forecasting of sewer overflow basin filling levels can provide actionable insights for early intervention, helping mitigating uncontrolled discharge. In recent years, AI-based forecasting methods have offered scalable alternatives to traditional physics-based models, but their reliance on cloud computing limits their reliability during communication outages. To address this, we propose an end-to-end forecasting framework that enables energy-efficient inference directly on edge devices. Our solution integrates lightweight Transformer and Long Short-Term Memory (LSTM) models, compressed via integer-only quantization for efficient on-device execution. Moreover, an automated hardware-aware deployment pipeline is used to search for optimal model configurations by jointly minimizing prediction error and energy consumption on an AMD Spartan-7 XC7S15 FPGA. Evaluated on real-world sewer data, the selected 8-bit Transformer model, trained on 24 hours of historical measurements, achieves high accuracy (MSE 0.0376) at an energy cost of 0.370 mJ per inference. In contrast, the optimal 8-bit LSTM model requires significantly less energy (0.009 mJ, over 40x lower) but yields 14.89% worse accuracy (MSE 0.0432) and much longer training time. This trade-off highlights the need to align model selection with deployment priorities, favoring LSTM for ultra-low energy consumption or Transformer for higher predictive accuracy. In general, our work enables local, energy-efficient forecasting, contributing to more resilient combined sewer systems. All code can be found in the GitHub Repository (https://github.com/tianheng-ling/EdgeOverflowForecast).