Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Learning-Augmented Online Control for Decarbonizing Water Infrastructures

Jianyi Yang, Pengfei Li, Tongxin Li, Adam Wierman, Shaolei Ren

TL;DR

A learning-augmented online control algorithm, termed LAOC, designed to dynamically schedule the activation and/or speed of water pumps, that can effectively reduce environmental and energy costs while guaranteeing safety constraints is proposed.

Abstract

Water infrastructures are essential for drinking water supply, irrigation, fire protection, and other critical applications. However, water pumping systems, which are key to transporting water to the point of use, consume significant amounts of energy and emit millions of tons of greenhouse gases annually. With the wide deployment of digital water meters and sensors in these infrastructures, Machine Learning (ML) has the potential to optimize water supply control and reduce greenhouse gas emissions. Nevertheless, the inherent vulnerability of ML methods in terms of worst-case performance raises safety concerns when deployed in critical water infrastructures. To address this challenge, we propose a learning-augmented online control algorithm, termed LAOC, designed to dynamically schedule the activation and/or speed of water pumps. To ensure safety, we introduce a novel design of safe action sets for online control problems. By leveraging these safe action sets, LAOC can provably guarantee safety constraints while utilizing ML predictions to reduce energy and environmental costs. Our analysis reveals the tradeoff between safety requirements and average energy/environmental cost performance. Additionally, we conduct an experimental study on a building water supply system to demonstrate the empirical performance of LAOC. The results indicate that LAOC can effectively reduce environmental and energy costs while guaranteeing safety constraints.

Learning-Augmented Online Control for Decarbonizing Water Infrastructures

TL;DR

A learning-augmented online control algorithm, termed LAOC, designed to dynamically schedule the activation and/or speed of water pumps, that can effectively reduce environmental and energy costs while guaranteeing safety constraints is proposed.

Abstract

Water infrastructures are essential for drinking water supply, irrigation, fire protection, and other critical applications. However, water pumping systems, which are key to transporting water to the point of use, consume significant amounts of energy and emit millions of tons of greenhouse gases annually. With the wide deployment of digital water meters and sensors in these infrastructures, Machine Learning (ML) has the potential to optimize water supply control and reduce greenhouse gas emissions. Nevertheless, the inherent vulnerability of ML methods in terms of worst-case performance raises safety concerns when deployed in critical water infrastructures. To address this challenge, we propose a learning-augmented online control algorithm, termed LAOC, designed to dynamically schedule the activation and/or speed of water pumps. To ensure safety, we introduce a novel design of safe action sets for online control problems. By leveraging these safe action sets, LAOC can provably guarantee safety constraints while utilizing ML predictions to reduce energy and environmental costs. Our analysis reveals the tradeoff between safety requirements and average energy/environmental cost performance. Additionally, we conduct an experimental study on a building water supply system to demonstrate the empirical performance of LAOC. The results indicate that LAOC can effectively reduce environmental and energy costs while guaranteeing safety constraints.
Paper Structure (38 sections, 9 theorems, 22 equations, 6 figures, 2 tables, 1 algorithm)

This paper contains 38 sections, 9 theorems, 22 equations, 6 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Problem Formulation
  4. Safe Online Control for Water Supply
  5. Broader Applications
  6. Assumptions
  7. Learning-Augmented Online Control (LAOC)
  8. Challenges Due to Safety Requirements
  9. Algorithm Design
  10. Performance Bounds
  11. Safety constraint satisfaction
  12. Average performance
  13. Generalization performance of safety-aware finetuning
  14. Case Study
  15. Setup
  16. ...and 23 more sections

Key Result

Proposition 4.1

Define the quality of pure ML as the normalized difference between the ML advice and the offline optimal action ${\|\tilde{u}-u^*\|^2}/{J_H^*}$. If the pure ML have an arbitrarily low quality (i.e., ${\|\tilde{u}-u^*\|^2}/{J_H^*}\to\infty$), Lin with $\rho\in (0,1]$ cannot guarantee $(1+\lambda)-$ s

Figures (6)

  • Figure 1: Water Supply Infrastructure with ML predictions.
  • Figure 2: Safety constraint violation and average costs. By default, OGD is the control prior for LAOC. ML (Ep. $N$) is the ML model at the $N$th epoch. LAOC (ML (Ep. $N$)) is LAOC using the purely-trained ML model at the $N$th epoch.
  • Figure 3: Average testing loss and the maximum risk ratio. By default, OGD is the control prior for LAOC. ML (Ep. $N$) is the ML model at the $N$th epoch. LAOC (ML (Ep. $N$)) is LAOC using the purely-trained ML model at the $N$th epoch. LAOC-F is LAOC with safety-aware finetuning \ref{['eqn:training']}.
  • Figure 4: Average loss, carbon cost and energy cost for different LAOC algorithms and control priors. LAOC algorithms use purely-trained ML model at Epoch 400. Here, MPC represents MPC-0.03 with a prediction error of 0.03.
  • Figure 5: Safety Violation Probability Under OOD Setting.
  • ...and 1 more figures

Theorems & Definitions (9)

  • Proposition 4.1
  • Proposition 4.2
  • Theorem 4.3
  • Theorem 4.4
  • Theorem 4.5
  • Lemma C.1: convex_course
  • Lemma D.1
  • Lemma D.2
  • Lemma D.3