search

IBM / rl-testbed-for-energyplus

Reinforcement Learning Testbed for Power Consumption Optimization using EnergyPlus
MIT License
191 stars 77 forks source link

Problem caused by shorter timestep #25

Closed OnedgeLee closed 5 years ago

OnedgeLee commented 5 years ago

Energyplus goes shorter timestep if zone temperature timestep difference is bigger than 0.3C. When I receive minutes from Energyplus environment through named pipe, (SET tmp_val1 = @ExtCtrlObs 10 Minute,) I could find shorter timestep comes after normal timestep. (example, 0:15:00 -> 0:30:00 -> 0:17:00 -> 0:20:00 -> 0:24:00 -> 0:27:00 -> 0:30:00) AfterPredictorAfterHVACManagers EMS calling point calls EMS program every timesteps, even shorter timesteps. Rollout sequence goes wrong because of this shorter timesteps.

Rollout should be in sequence 0:15:00 -> 0:30:00 -> 0:45:00 -> 1:00:00 ... but goes under 0:15:00 -> 0:30:00 -> 0:17:00 -> 0:20:00 ...

Shorter timesteps should be ignored, do actions cloned from normal timesteps, and shoud be omitted from rollout sequence I think.

antoine-galataud commented 5 years ago

Hi @OnedgeLee. I think your issue is the same as the one I reported in https://github.com/IBM/rl-testbed-for-energyplus/issues/9 Basically, EnergyPlus has system timesteps that are not meant to be used and should be skipped during rollout or trajectory sampling.

OnedgeLee commented 5 years ago

Hi @OnedgeLee. I think your issue is the same as the one I reported in #9 Basically, EnergyPlus has system timesteps that are not meant to be used and should be skipped during rollout or trajectory sampling.

Thank you, you are right.