How beardhealth/combined-heat-and-power-system-economic-dispatch works
This repository presents deep-reinforcement-learning approaches to combined heat-and-power economic dispatch. A run selects an algorithm, interacts with a dispatch environment to obtain states and rewards, and either trains a policy or evaluates it and records dispatch outcomes. The sampled files show TD3 with a separate environment module and asynchronous PPO with a CHP environment. The README describes the product broadly but gives little implementation detail; the test-system environment’s internals were not sampled.
Main entry points: DPPO_CHP.py PPO runner · DRL.py TD3 runner · CHP_MODEL.py CHP environment · test_system4.py Test-system environment
Components 13
TD3 dispatch
- TD3 runner
DRL.py - TD3 agent
DRL.py - Replay buffer
DRL.py - TD3 checkpoints
DRL.py
PPO dispatch
- PPO runner
DPPO_CHP.py - PPO agent
DPPO_CHP.py - Rollout worker
DPPO_CHP.py - Rollout queue
DPPO_CHP.py - PPO results
DPPO_CHP.py
Dispatch simulation
- Test-system environment
test_system4.py - CHP environment
CHP_MODEL.py
Other components
- Researcher
- TensorBoard logs
Connections 15
- Researcher → TD3 runner: starts run
DRL.py - TD3 runner → TD3 agent: selects actions
DRL.py - TD3 runner → Test-system environment: steps environment
DRL.py - TD3 agent → Replay buffer: stores samples
DRL.py - TD3 agent → TensorBoard logs: writes metrics
DRL.py - TD3 runner → TD3 checkpoints: saves or loads
DRL.py - Researcher → PPO runner: starts run
DPPO_CHP.py - PPO runner → PPO agent: updates policy
DPPO_CHP.py - PPO runner → Rollout worker: starts workers
DPPO_CHP.py - Rollout worker → PPO agent: chooses actions
DPPO_CHP.py - Rollout worker → CHP environment: steps environment
DPPO_CHP.py - Rollout worker → Rollout queue: queues rollouts
DPPO_CHP.py - PPO agent → Rollout queue: consumes rollouts
DPPO_CHP.py - PPO runner → CHP environment: evaluates policy
DPPO_CHP.py - PPO runner → PPO results: saves outputs
DPPO_CHP.py