@@ -31,7 +31,71 @@ Change Log
3131- [???] "asynch" multienv
3232- [???] properly model interconnecting powerlines
3333
34- [1.10.1] - 2024-03-18
34+ [1.10.2] - 2024-05-27
35+ -------------------------
36+ - [BREAKING] the `runner.run_one_episode ` now returns an extra first argument:
37+ `chron_id, chron_name, cum_reward, timestep, max_ts = runner.run_one_episode() ` which
38+ is consistant with `runner.run(...) ` (previously it returned only
39+ `chron_name, cum_reward, timestep, max_ts = runner.run_one_episode() `)
40+ - [BREAKING] the runner now has no `chronics_handler ` attribute (`runner.chronics_handler `
41+ is not defined)
42+ - [BREAKING] now grid2op forces everything to be connected at busbar 1 if
43+ `param.IGNORE_INITIAL_STATE_TIME_SERIE == True ` (**NOT ** the default) and
44+ no initial state is provided in `env.reset(..., options={"init state": ...}) `
45+ - [ADDED] it is now possible to call `change_reward ` directly from
46+ an observation (no need to do it from the Observation Space)
47+ - [ADDED] method to change the reward from the observation (observation_space
48+ is not needed anymore): you can use `obs.change_reward `
49+ - [ADDED] a way to automatically set the `experimental_read_from_local_dir ` flags
50+ (with automatic class creation). For now it is disable by default, but you can
51+ activate it transparently (see doc)
52+ - [ADDED] possibility to set the grid to an initial state (using an action) when using the
53+ "time series" classes. The supported classes are `GridStateFromFile ` - and all its derivative,
54+ `FromOneEpisodeData `, `FromMultiEpisodeData `, `FromNPY ` and `FromHandlers `. The classes `ChangeNothing `
55+ and `FromChronix2grid ` are not supported at the moment.
56+ - [ADDED] an "Handler" (`JSONInitStateHandler `) that can set the grid to an initial state (so as to make
57+ compatible the `FromHandlers ` time series class with this new feature)
58+ - [ADDED] some more type hints in the `GridObject ` class
59+ - [ADDED] Possibility to deactive the support of shunts if subclassing `PandaPowerBackend `
60+ (and add some basic tests)
61+ - [ADDED] a parameters (`param.IGNORE_INITIAL_STATE_TIME_SERIE `) which defaults to
62+ `False ` that tells the environment whether it should ignore the
63+ initial state of the grid provided in the time series.
64+ By default it is NOT ignored, it is taken into account
65+ (for the environment that supports this feature)
66+ - [FIXED] a small issue that could lead to having
67+ "redispatching_unit_commitment_availble" flag set even if the redispatching
68+ data was not loaded correctly
69+ - [FIXED] EducPandaPowerBackend now properly sends numpy array in the class attributes
70+ (instead of pandas series)
71+ - [FIXED] an issue when loading back data (with `EpisodeData `): when there were no storage units
72+ on the grid it did not set properly the "storage relevant" class attributes
73+ - [FIXED] a bug in the "gridobj.generate_classes()" function which crashes when no
74+ grid layout was set
75+ - [FIXED] notebook 5 on loading back data with `EpisodeData `.
76+ - [FIXED] converter between backends (could not handle more than 2 busbars)
77+ - [FIXED] a bug in `BaseMultiProcessEnvironment `: set_filter had no impact
78+ - [FIXED] an issue in the `Runner ` (`self.chronics_handler ` was sometimes used, sometimes not
79+ and most of the time incorrectly)
80+ - [FIXED] on `RemoteEnv ` class (impact all multi process environment): the kwargs used to build then backend
81+ where not used which could lead to"wrong" backends being used in the sub processes.
82+ - [FIXED] a bug when the name of the times series and the names of the elements in the backend were
83+ different: it was not possible to set `names_chronics_to_grid ` correctly when calling `env.make `
84+ - [IMPROVED] documentation about `obs.simulate ` to make it clearer the
85+ difference between env.step and obs.simulate on some cases
86+ - [IMPROVED] type hints on some methods of `GridObjects `
87+ - [IMPROVED] replace `np.nonzero(arr) ` calls with `arr.nonzero() ` which could
88+ save up a bit of computation time.
89+ - [IMPROVED] force class attributes to be numpy arrays of proper types when the
90+ classes are initialized from the backend.
91+ - [IMPROVED] some (slight) speed improvments when comparing actions or deep copying objects
92+ - [IMPROVED] the way the "grid2op compat" mode is handled
93+ - [IMPROVED] the coverage of the tests in the "test_basic_env_ls.py" to test more in depth lightsim2grid
94+ (creation of multiple environments, grid2op compatibility mode)
95+ - [IMPROVED] the function to test the backend interface in case when shunts are not supported
96+ (improved test `AAATestBackendAPI.test_01load_grid `)
97+
98+ [1.10.1] - 2024-03-xx
3599----------------------
36100- [FIXED] issue https://github.com/rte-france/Grid2Op/issues/593
37101- [FIXED] backward compatibility issues with "oldest" lightsim2grid versions
0 commit comments