Generalized reinforcement learning for building control using Behavioral Cloning

نویسندگان

چکیده

Advanced building control methods such as model predictive (MPC) offer significant benefits to both consumers and grid operators, but high computational requirements have acted barriers more widespread adoption. Local computation requires installation of expensive hardware, while cloud computing introduces data security privacy concerns. In this paper, we drastically reduce the local advanced through a reinforcement learning (RL)-based approach called Behavioral Cloning, which represents MPC policy neural network that can be locally implemented quickly computed on low-cost programmable logic controller. While previous RL approximate must specifically trained for each building, our key improvement is proposed controller generalize many buildings, electricity rates, thermostat setpoint schedules without additional, effort-intensive retraining. To provide versatility, adapted traditional Cloning two innovations: (1) constraint-informed parameter grouping (CIPG) method provides efficient representation training (2) new deep model-structure reverse-time recurrent networks (RT-RNN) allows future information flow backward in time effectively interpret temporal disturbance predictions. The result an easy-to-deploy, generalized behavioral clone little building-specific tuning, reducing effort costs associated with implementing smart residential heat pump control.

برای دانلود رایگان متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

15 Reinforcement Learning for Building Environmental Control

During the last few decades, significant changes have been made in the area of building construction and management, especially regarding climate control and energy conservation. A significant turning point was reached in the early 70s with the oil crisis driving a movement for airtight buildings to minimize energy consumption. Unfortunately this turn resulted in a significant deterioration of ...

متن کامل

Neuroevolutionary reinforcement learning for generalized control of simulated helicopters

This article presents an extended case study in the application of neuroevolution to generalized simulated helicopter hovering, an important challenge problem for reinforcement learning. While neuroevolution is well suited to coping with the domain's complex transition dynamics and high-dimensional state and action spaces, the need to explore efficiently and learn on-line poses unusual challeng...

متن کامل

Reinforcement Learning for Control

Reinforcement learning (RL) offers a principled way to control nonlinear stochastic systems with partly or even fully unknown dynamics. Recent advances in areas such as deep learning and adaptive dynamic programming (ADP) have led to significant inroads in applications from robotics, automotive systems, smart grids, game playing, traffic control, etc. This open track provides a forum of interac...

متن کامل

Instructional control of reinforcement learning: a behavioral and neurocomputational investigation.

Humans learn how to behave directly through environmental experience and indirectly through rules and instructions. Behavior analytic research has shown that instructions can control behavior, even when such behavior leads to sub-optimal outcomes (Hayes, S. (Ed.). 1989. Rule-governed behavior: cognition, contingencies, and instructional control. Plenum Press.). Here we examine the control of be...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

ژورنال

عنوان ژورنال: Applied Energy

سال: 2021

ISSN: ['0306-2619', '1872-9118']

DOI: https://doi.org/10.1016/j.apenergy.2021.117602