AUdaltsova commited on
Commit
8379a1f
·
verified ·
1 Parent(s): 79182b6

Upload models - 7s0axgkr

Browse files
Files changed (4) hide show
  1. README.md +9 -8
  2. data_config.yaml +67 -12
  3. model_config.yaml +14 -5
  4. model_weights.safetensors +2 -2
README.md CHANGED
@@ -8,7 +8,7 @@ Do not remove elements like the above surrounded by two curly braces and do not
8
  -->
9
 
10
  <!-- Title - e.g. PVNet2, WindNet, PVNet India -->
11
- # TEMPLATE
12
 
13
  <!-- Provide a longer summary of what this model is/does. -->
14
  ## Model Description
@@ -17,6 +17,8 @@ Do not remove elements like the above surrounded by two curly braces and do not
17
  This model class uses satellite data, and numerical weather predictions to forecast the near-term (up to 8 hours ahead) PV power output at all Grid Service Points (GSPs) in Great Britain. More information can be found in the model repo [1]. The model repo also includes links to our workshop paper on this model and some experimental notes.
18
  -->
19
 
 
 
20
  - **Developed by:** openclimatefix
21
  - **Model type:** Fusion model
22
  - **Language(s) (NLP):** en
@@ -31,6 +33,7 @@ The model is trained on data from 2019-2022 and validated on data from 2022-2023
31
 
32
  See the data_config.yaml file for more information on the channels and window-size used for each input data source.
33
  -->
 
34
 
35
  <!-- The preprocessing section is not strictly nessessary but perhaps nice to have -->
36
  ### Preprocessing
@@ -39,11 +42,13 @@ See the data_config.yaml file for more information on the channels and window-si
39
  Data is prepared with the `ocf_data_sampler/torch_datasets/datasets/pvnet_uk` Dataset [2].
40
  -->
41
 
 
 
42
  ## Results
43
 
44
  <!-- Do not remove the lines below -->
45
  The training logs for this model commit can be found here:
46
- - [https://wandb.ai/openclimatefix/NL-Solar/runs/rvdfwb7o](https://wandb.ai/openclimatefix/NL-Solar/runs/rvdfwb7o)
47
 
48
 
49
  <!-- The hardware section is also just nice to have -->
@@ -60,9 +65,5 @@ This model was trained using the following Open Climate Fix packages:
60
 
61
  <!-- Especially do not change the two lines below -->
62
  The versions of these packages can be found below:
63
- - pvnet==5.3.0.post6+git.f4136853.dirty
64
- - ocf-data-sampler==1.0.1.post1+git.b7c40f80.dirty
65
-
66
-
67
- ---
68
- **Migration Note**: This model was migrated on 2025-12-18 to pvnet version 5.3.0.post6+git.f4136853.dirty
 
8
  -->
9
 
10
  <!-- Title - e.g. PVNet2, WindNet, PVNet India -->
11
+ # PVNet Netherlands
12
 
13
  <!-- Provide a longer summary of what this model is/does. -->
14
  ## Model Description
 
17
  This model class uses satellite data, and numerical weather predictions to forecast the near-term (up to 8 hours ahead) PV power output at all Grid Service Points (GSPs) in Great Britain. More information can be found in the model repo [1]. The model repo also includes links to our workshop paper on this model and some experimental notes.
18
  -->
19
 
20
+ This model uses NWP and Satellite data to predict regional production of solar power in Netherlands. Predictions are done with 15 minutely resolution from 15 minutes to 36 hours ahead.
21
+
22
  - **Developed by:** openclimatefix
23
  - **Model type:** Fusion model
24
  - **Language(s) (NLP):** en
 
33
 
34
  See the data_config.yaml file for more information on the channels and window-size used for each input data source.
35
  -->
36
+ This model was trained on data from 2019-2023 and validated on data from 2024. It uses NWP data from ECMWF IFS model and satellite data from the EUMETSAT MSG SEVIRI instrument. See the data_config.yaml file for more information on the channels and window-size used for each input data source.
37
 
38
  <!-- The preprocessing section is not strictly nessessary but perhaps nice to have -->
39
  ### Preprocessing
 
42
  Data is prepared with the `ocf_data_sampler/torch_datasets/datasets/pvnet_uk` Dataset [2].
43
  -->
44
 
45
+ Data is prepared with the `ocf_data_sampler/torch_datasets/datasets/pvnet_uk` Dataset [2].
46
+
47
  ## Results
48
 
49
  <!-- Do not remove the lines below -->
50
  The training logs for this model commit can be found here:
51
+ - [https://wandb.ai/openclimatefix/NL-Solar/runs/7s0axgkr](https://wandb.ai/openclimatefix/NL-Solar/runs/7s0axgkr)
52
 
53
 
54
  <!-- The hardware section is also just nice to have -->
 
65
 
66
  <!-- Especially do not change the two lines below -->
67
  The versions of these packages can be found below:
68
+ - pvnet==5.3.0.post0+git.489723d6.dirty
69
+ - ocf-data-sampler==1.0.6
 
 
 
 
data_config.yaml CHANGED
@@ -1,7 +1,16 @@
1
  general:
2
- description: Initial config for NL model
3
- name: Netherlands_pvnet_config
4
  input_data:
 
 
 
 
 
 
 
 
 
5
  nwp:
6
  ecmwf:
7
  provider: ecmwf
@@ -101,16 +110,62 @@ input_data:
101
  diff_direct_shortwave_radiation_flux_gl:
102
  mean: 469169.5
103
  std: 818950.6875
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
104
  solar_position:
105
- interval_start_minutes: -2880
106
  interval_end_minutes: 2160
107
  time_resolution_minutes: 15
108
- generation:
109
- time_resolution_minutes: 15
110
- interval_start_minutes: -2880
111
- interval_end_minutes: 2160
112
- dropout_timedeltas_minutes:
113
- - -15
114
- dropout_fraction: 1.0
115
- dropout_value: -1
116
- zarr_path: /home/alex/NL/NL_regional_generation_kw.nc
 
1
  general:
2
+ description: Configuration for a Regional Netherlands model
3
+ name: NL_regional_config
4
  input_data:
5
+ generation:
6
+ zarr_path: PLACEHOLDER.zarr
7
+ interval_start_minutes: -60
8
+ interval_end_minutes: 2160
9
+ time_resolution_minutes: 15
10
+ dropout_timedeltas_minutes:
11
+ - -15
12
+ dropout_fraction: 1.0
13
+ dropout_value: -1
14
  nwp:
15
  ecmwf:
16
  provider: ecmwf
 
110
  diff_direct_shortwave_radiation_flux_gl:
111
  mean: 469169.5
112
  std: 818950.6875
113
+ satellite:
114
+ zarr_path: PLACEHOLDER.zarr
115
+ interval_start_minutes: -60
116
+ interval_end_minutes: -5
117
+ time_resolution_minutes: 5
118
+ channels:
119
+ - IR_016
120
+ - IR_039
121
+ - IR_087
122
+ - IR_097
123
+ - IR_108
124
+ - IR_120
125
+ - IR_134
126
+ - VIS006
127
+ - VIS008
128
+ - WV_062
129
+ - WV_073
130
+ image_size_pixels_height: 20
131
+ image_size_pixels_width: 20
132
+ dropout_timedeltas_minutes: []
133
+ dropout_fraction: 0
134
+ normalisation_constants:
135
+ IR_016:
136
+ mean: 0.17594202
137
+ std: 0.21462157
138
+ IR_039:
139
+ mean: 0.86167645
140
+ std: 0.04618041
141
+ IR_087:
142
+ mean: 0.7719318
143
+ std: 0.06687243
144
+ IR_097:
145
+ mean: 0.8014212
146
+ std: 0.0468558
147
+ IR_108:
148
+ mean: 0.71254843
149
+ std: 0.17482725
150
+ IR_120:
151
+ mean: 0.89058584
152
+ std: 0.06115861
153
+ IR_134:
154
+ mean: 0.944365
155
+ std: 0.04492306
156
+ VIS006:
157
+ mean: 0.09633306
158
+ std: 0.12184761
159
+ VIS008:
160
+ mean: 0.11426069
161
+ std: 0.13090034
162
+ WV_062:
163
+ mean: 0.7359355
164
+ std: 0.16111417
165
+ WV_073:
166
+ mean: 0.62479186
167
+ std: 0.12924142
168
  solar_position:
169
+ interval_start_minutes: -60
170
  interval_end_minutes: 2160
171
  time_resolution_minutes: 15
 
 
 
 
 
 
 
 
 
model_config.yaml CHANGED
@@ -17,6 +17,15 @@ nwp_encoders_dict:
17
  number_of_conv3d_layers: 4
18
  conv3d_channels: 32
19
  image_size_pixels: 10
 
 
 
 
 
 
 
 
 
20
  output_network:
21
  _target_: pvnet.models.late_fusion.linear_networks.networks.ResFCNet
22
  _partial_: true
@@ -25,7 +34,7 @@ output_network:
25
  res_block_layers: 2
26
  dropout_frac: 0.2
27
  include_sun: true
28
- interval_minutes: 15
29
  location_id_mapping:
30
  1: 1
31
  2: 2
@@ -40,14 +49,14 @@ location_id_mapping:
40
  11: 11
41
  12: 12
42
  forecast_minutes: 2160
43
- history_minutes: 2880
 
44
  min_sat_delay_minutes: 5
45
- sat_history_minutes: 65
46
- pv_history_minutes: 2880
47
  nwp_history_minutes:
48
  ecmwf: 120
49
  nwp_forecast_minutes:
50
  ecmwf: 2220
51
  nwp_interval_minutes:
52
  ecmwf: 60
53
- include_generation_history: true
 
17
  number_of_conv3d_layers: 4
18
  conv3d_channels: 32
19
  image_size_pixels: 10
20
+ sat_encoder:
21
+ _target_: pvnet.models.late_fusion.encoders.encoders3d.DefaultPVNet
22
+ _partial_: true
23
+ in_channels: 11
24
+ out_features: 256
25
+ number_of_conv3d_layers: 6
26
+ conv3d_channels: 32
27
+ image_size_pixels: 20
28
+ add_image_embedding_channel: false
29
  output_network:
30
  _target_: pvnet.models.late_fusion.linear_networks.networks.ResFCNet
31
  _partial_: true
 
34
  res_block_layers: 2
35
  dropout_frac: 0.2
36
  include_sun: true
37
+ include_generation_history: false
38
  location_id_mapping:
39
  1: 1
40
  2: 2
 
49
  11: 11
50
  12: 12
51
  forecast_minutes: 2160
52
+ history_minutes: 60
53
+ interval_minutes: 15
54
  min_sat_delay_minutes: 5
55
+ sat_history_minutes: 60
56
+ pv_history_minutes: 60
57
  nwp_history_minutes:
58
  ecmwf: 120
59
  nwp_forecast_minutes:
60
  ecmwf: 2220
61
  nwp_interval_minutes:
62
  ecmwf: 60
 
model_weights.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:7e321baa57c46809d450510fa0bf0c7d096a87ebefabed601aad38414d4423ec
3
- size 2581760
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9b5540fd5e150b356103f603525426622f01603b42fd4b0b038a293a00424723
3
+ size 15898832