English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Buildings sector is one of the major consumers of energy in the UnitedStates. The buildings HVAC (Heating, Ventilation, and Air Conditioning)systems, whose functionality is to maintain thermal comfort and indoor airquality (IAQ), account for almost half of the energy consumed by the buildings.Thus, intelligent scheduling of the building HVAC system has the potential fortremendous energy and cost savings while ensuring that the control objectives(thermal comfort, air quality) are satisfied. Recently, several works havefocused on model-free deep reinforcement learning based techniques such as DeepQ-Network (DQN). Such methods require extensive interactions with theenvironment. Thus, they are impractical to implement in real systems due to lowsample efficiency. Safety-aware exploration is another challenge in realsystems since certain actions at particular states may result in catastrophicoutcomes. To address these issues and challenges, we propose a model-basedreinforcement learning approach that learns the system dynamics using a neuralnetwork. Then, we adopt Model Predictive Control (MPC) using the learned systemdynamics to perform control with random-sampling shooting method. To ensuresafe exploration, we limit the actions within safe range and the maximumabsolute change of actions according to prior knowledge. We evaluate our ideasthrough simulation using widely adopted EnergyPlus tool on a case studyconsisting of a two zone data-center. Experiments show that the averagedeviation of the trajectories sampled from the learned dynamics and the groundtruth is below $20\%$. Compared with baseline approaches, we reduce the totalenergy consumption by $17.1\% \sim 21.8\%$. Compared with model-freereinforcement learning approach, we reduce the required number of trainingsteps to converge by 10x.

Building HVAC Scheduling Using Reinforcement Learning via Neural Network Based Model Approximation