Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

EcoNet: Multiagent Planning and Control Of Household Energy Resources Using Active Inference

John C. Boik, Kobus Esterhuysen, Jacqueline B. Hynes, Axel Constant, Ines Hipolito, Mahault Albarracin, Alex B. Kiefer, Karl Friston

TL;DR

EcoNet presents a Bayesian active-inference approach to coordinating household energy resources under uncertainty. By modeling two agents (Thermostat and Battery) within a generative framework, it demonstrates planning ahead and uncertainty-aware decision-making for HVAC, solar generation, and storage. The study shows deterministic and learning-based scenarios where agents balance costs, emissions, and comfort, and it demonstrates learning of a missing room-temperature model over time. The work highlights the potential of multi-agent active inference for DER coordination and sets the stage for scaling to neighborhoods and dynamic, probabilistic forecasts.

Abstract

Advances in automated systems afford new opportunities for intelligent management of energy at household, local area, and utility scales. Home Energy Management Systems (HEMS) can play a role by optimizing the schedule and use of household energy devices and resources. One challenge is that the goals of a household can be complex and conflicting. For example, a household might wish to reduce energy costs and grid-associated greenhouse gas emissions, yet keep room temperatures comfortable. Another challenge is that an intelligent HEMS agent must make decisions under uncertainty. An agent must plan actions into the future, but weather and solar generation forecasts, for example, provide inherently uncertain estimates of future conditions. This paper introduces EcoNet, a Bayesian approach to household and neighborhood energy management that is based on active inference. The aim is to improve energy management and coordination, while accommodating uncertainties and taking into account potentially conditional and conflicting goals and preferences. Simulation results are presented and discussed.

EcoNet: Multiagent Planning and Control Of Household Energy Resources Using Active Inference

TL;DR

EcoNet presents a Bayesian active-inference approach to coordinating household energy resources under uncertainty. By modeling two agents (Thermostat and Battery) within a generative framework, it demonstrates planning ahead and uncertainty-aware decision-making for HVAC, solar generation, and storage. The study shows deterministic and learning-based scenarios where agents balance costs, emissions, and comfort, and it demonstrates learning of a missing room-temperature model over time. The work highlights the potential of multi-agent active inference for DER coordination and sets the stage for scaling to neighborhoods and dynamic, probabilistic forecasts.

Abstract

Advances in automated systems afford new opportunities for intelligent management of energy at household, local area, and utility scales. Home Energy Management Systems (HEMS) can play a role by optimizing the schedule and use of household energy devices and resources. One challenge is that the goals of a household can be complex and conflicting. For example, a household might wish to reduce energy costs and grid-associated greenhouse gas emissions, yet keep room temperatures comfortable. Another challenge is that an intelligent HEMS agent must make decisions under uncertainty. An agent must plan actions into the future, but weather and solar generation forecasts, for example, provide inherently uncertain estimates of future conditions. This paper introduces EcoNet, a Bayesian approach to household and neighborhood energy management that is based on active inference. The aim is to improve energy management and coordination, while accommodating uncertainties and taking into account potentially conditional and conflicting goals and preferences. Simulation results are presented and discussed.
Paper Structure (16 sections, 9 equations, 9 figures)

This paper contains 16 sections, 9 equations, 9 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. RL and MPC for Home Energy Management Systems
  4. Active Inference for Home Energy Management Systems
  5. Active Inference
  6. Model Description
  7. Results
  8. Deterministic Scenario
  9. Parameter Learning Scenario
  10. Discussion
  11. Declarations
  12. Availability of Data and Material
  13. Competing Interests
  14. Funding
  15. Authors' Contributions
  16. ...and 1 more sections

Figures (9)

  • Figure 1: Pseudo-factor graph for two-agent EcoNet model. See text for meaning of state and observation variables and for information shared between agents. Squares labeled $A$ and $B$ refer to $A$ (likelihood) and $B$ (prior transition) matrices in the pymdp implementation.
  • Figure 2: Daily outdoor temperature and baseline energy use as examples of input data (top and bottom panels). Shading indicates occupancy status. Target temperatures for occupied and unoccupied status are prior preferences set by the user. The middle panel shows cumulative absolute difference between outdoor temperature and targets. This is the worst case scenario for indoor room temperature, for a house without a HVAC system.
  • Figure 3: Distribution of negative EFE over policies and simulation steps.
  • Figure 4: Observed outdoor and room temperatures (top panel). Difference between observed room temperature and targets (middle panel). The selected thermostat actions (bottom panel).
  • Figure 5: Agent beliefs about battery state of charge. Beliefs about ToU rates are shaded.
  • ...and 4 more figures