Introducing Learning for Storages

assume/strategies/__init__.py

@@ -51,9 +51,11 @@
     from assume.strategies.learning_advanced_orders import (
         RLAdvancedOrderStrategy,
     )
-    from assume.strategies.learning_strategies import RLStrategy
+    from assume.strategies.learning_strategies import RLStrategy, BatteryRLStrategy, BatteryRLStrategy_Curve
 
     bidding_strategies["pp_learning"] = RLStrategy
+    bidding_strategies["storage_learning"] = BatteryRLStrategy
+    bidding_strategies["storage_learning_curve"] = BatteryRLStrategy_Curve
     bidding_strategies["learning_advanced_orders"] = RLAdvancedOrderStrategy
 
 except ImportError:

examples/examples.py

@@ -73,6 +73,8 @@
     "small_learning_2_lstm": {"scenario": "example_02b", "study_case": "base_lstm"},
     # Further DRL example simulation showcasing learning features
     "learning_with_complex_bids": {"scenario": "example_02d", "study_case": "dam"},
+    "small_learning_storage": {"scenario": "example_02e", "study_case": "base"},
+
     #
     # full year examples to show real-world scenarios
     "large_2019_eom": {"scenario": "example_03", "study_case": "base_case_2019"},
@@ -100,8 +102,8 @@
     - local_db: without database and grafana
     - timescale: with database and grafana (note: you need docker installed)
     """
-    data_format = "local_db"  # "local_db" or "timescale"
-    example = "small"
+    data_format = "timescale"  # "local_db" or "timescale"
+    example = "small_learning_storage"
 
     if data_format == "local_db":
         db_uri = f"sqlite:///./examples/local_db/assume_db_{example}.db"

examples/inputs/example_02e/config.yaml

@@ -0,0 +1,93 @@
+# SPDX-FileCopyrightText: ASSUME Developers
+#
+# SPDX-License-Identifier: AGPL-3.0-or-later
+
+tiny:
+  start_date: 2019-01-01 00:00
+  end_date: 2019-01-05 00:00
+  time_step: 1h
+  save_frequency_hours: null
+  learning_mode: True
+
+  learning_config:
+    continue_learning: False
+    trained_policies_save_path: null
+    max_bid_price: 100
+    algorithm: matd3
+    actor_architecture: mlp
+    learning_rate: 0.001
+    training_episodes: 10
+    episodes_collecting_initial_experience: 3
+    train_freq: 24h
+    gradient_steps: -1
+    batch_size: 64
+    gamma: 0.99
+    device: cpu
+    noise_sigma: 0.1
+    noise_scale: 1
+    noise_dt: 1
+    validation_episodes_interval: 5
+
+  markets_config:
+    EOM:
+      operator: EOM_operator
+      product_type: energy
+      products:
+        - duration: 1h
+          count: 1
+          first_delivery: 1h
+      opening_frequency: 1h
+      opening_duration: 1h
+      volume_unit: MWh
+      maximum_bid_volume: 100000
+      maximum_bid_price: 3000
+      minimum_bid_price: -500
+      price_unit: EUR/MWh
+      market_mechanism: pay_as_clear
+
+
+base:
+  start_date: 2019-03-01 00:00
+  end_date: 2019-03-31 00:00
+  time_step: 1h
+  save_frequency_hours: null
+  learning_mode: True
+
+  learning_config:
+    continue_learning: False
+    trained_policies_save_path: null
+    max_bid_price: 100
+    algorithm: matd3
+    actor_architecture: mlp
+    learning_rate: 0.001
+    training_episodes: 50
+    episodes_collecting_initial_experience: 5
+    train_freq: 24h
+    gradient_steps: -1
+    batch_size: 256
+    gamma: 0.99
+    device: cpu
+    noise_sigma: 0.1
+    noise_scale: 1
+    noise_dt: 1
+    validation_episodes_interval: 5
+    early_stopping_steps: 10
+    early_stopping_threshold: 0.05
+
+  markets_config:
+    EOM:
+      operator: EOM_operator
+      product_type: energy
+      products:
+        - duration: 1h
+          count: 1
+          first_delivery: 1h
+      opening_frequency: 1h
+      opening_duration: 1h
+      volume_unit: MWh
+      maximum_bid_volume: 100000
+      maximum_bid_price: 3000
+      minimum_bid_price: -500
+      price_unit: EUR/MWh
+      market_mechanism: pay_as_clear
+