A Non-linear Differentiable Model for Stormwater-based Irrigation of a Green Roof in Toronto

TL;DR

This paper develops a non-linear, differentiable model for stormwater flow to optimize irrigation of green roofs in Toronto, aiming to reduce potable water use and sewer overflows through predictive control strategies.

Contribution

It introduces a site-specific non-linear stormwater flow model that enables the use of model predictive control for green roof irrigation systems.

Findings

MPC outperforms reactive on/off control in simulations.

The non-linear model facilitates effective predictive control.

Potential for reducing potable water use in urban green infrastructure.

Abstract

Green infrastructure has potential to alleviate the environmental impact of rapidly growing cities. This potential has inspired laws in Toronto that require the inclusion of rooftops with large vegetation beds, called green roofs, into sufficiently sized construction projects. We study the problem of reusing stormwater to irrigate a green roof in Toronto, where potable water is the current irrigation source. The vision is that widespread reuse of stormwater runoff for irrigation of green roofs and other purposes can reduce sewer overflow volumes without over-building (with the added benefit of conserving potable water). Towards this vision, our goal is to develop and evaluate two pump controllers for transporting stormwater to the green roof of interest in simulation. A key contribution is our development of a site-specific non-linear model for stormwater flow using smoothing techniques that permits linearization and a standard model predictive controller (MPC). We compare the efficacy of the MPC, which anticipates the weather, and an on/off controller, which is reactive rather than anticipative, for the site in simulation. With further study, we are hopeful that this research will advance control systems technology to improve the performance of green and stormwater infrastructure in growing urban areas.