Yuan Zhou

Title: Yuan Zhou and Hybrid Reinforcement Learning for Water Distribution Networks.

Introduction

Yuan Zhou is an inventor based in Waterloo, CA, and works at Autodesk, Inc. His patent work applies hybrid reinforcement learning to the control of water distribution networks. He is associated with 1 patent.

Latest Patents

Zhou’s patent, “Hybrid reinforcement learning (RL) to control a water distribution network,” describes a method and system for managing network operations. The system maintains a database of prior states using residential water demand, tank level, and energy tariff information, then determines the network’s current state.

It evaluates rewards that account for tank-level constraints, energy costs, and the number of control toggles. A query-based model identifies control points for a prior state, and a reinforcement-learning agent is trained using prior states and rewards. The agent selects pump-speed setpoints intended to maintain tank levels, reduce energy costs, and comply with toggle limits. It also selects time slots and generates hybrid setpoints for controlling the network during those periods.

Career Highlights

Zhou’s listed invention brings reinforcement learning and query-based modeling together to address operational decisions in water distribution. Its stated objectives include maintaining tank levels, managing energy costs, and limiting control toggles.

Collaborations

The listed coworkers associated with Zhou are Shu Wang and Harsh Bakulchandra Patel.

Conclusion

Yuan Zhou’s patent describes a data-driven approach to water distribution control that uses reinforcement learning to balance tank-level requirements, energy costs, and operational constraints.