def task_configuration():
es_history_config = EsConfig(host="51.144.39.71:9200", index_name="wf_scada_hist", document_type="wf_scada_hist")
es_probability_config = EsConfig(host="51.144.39.71:9200", index_name="wf_model_predictions", document_type="wf_model_prediction")
es_rul_config = EsConfig(host="51.144.39.71:9200", index_name="wf_model_ruls", document_type="wf_model_rul")
return {
"company_id": 1,
"turbine_id": 1,
"end_date": datetime.datetime(2016, 2, 29),
"window_delta": 1,
"es_history_config": es_history_config,
"es_probability_config": es_probability_config,
"es_rul_config": es_rul_config,
"components_definitions": []
}
def test_residual_time():
es_config = EsConfig(host="51.144.39.71:9200", index_name="wf_model_ruls", document_type="wf_model_rul")
turbine_id = 1
end_date = datetime(2016, 2, 29)
window_delta = timedelta(days=10)
ruls = get_rul(turbine_id, end_date, window_delta, es_config)
res_time = residual_time(ruls['global_model'])
np.testing.assert_almost_equal(res_time, 10.662488246015524, decimal=5, verbose=True)
def test_last_rul_healthscore():
es_config = EsConfig(host="51.144.39.71:9200", index_name="wf_model_ruls", document_type="wf_model_rul")
turbine_id = 1
end_date = datetime(2016, 2, 29)
window_delta = timedelta(hours=1)
ruls = get_rul(turbine_id, end_date, window_delta, es_config)
global_model_rul_healthscore = last_rul_healthscore(ruls['global_model'])
assert 0 <= global_model_rul_healthscore <= 100
def test_get_rul():
es_config = EsConfig(host="51.144.39.71:9200", index_name="wf_model_ruls", document_type="wf_model_rul")
turbine_id = 1
end_date = datetime(2016, 2, 29)
window_delta = timedelta(days=10)
ruls = get_rul(turbine_id, end_date, window_delta, es_config)
assert isinstance(ruls, pd.DataFrame)
indexes = ruls.index
assert indexes[0] == pd.Timestamp(end_date - window_delta)
assert indexes[-1] == end_date
def test_last_probability_healthscore():
es_config = EsConfig(host="51.144.39.71:9200", index_name="wf_model_predictions",
document_type="wf_model_prediction")
turbine_id = 1
end_date = datetime(2016, 2, 29)
window_delta = timedelta(days=10)
probabilities = get_probabilities(turbine_id, end_date, window_delta, es_config)
global_model_score = last_probability_healthscore(probabilities['global_model'])
np.testing.assert_almost_equal(global_model_score, 88.11894166872555, decimal=5, verbose=True)
if __name__ == '__main__':
args = parser.parse_args()
logging.info("Running prediction")
configuration = parse_config_yaml(
os.path.join(args.data_dir, 'configuration.yaml'))
end_date = args.day
start_date = end_date - timedelta(hours=int(args.window_delta))
turbines = args.turbines # tuple: (1,)
method = args.method
transformer_data = pickle.load(
open(os.path.join(args.data_dir, transfomer_data), 'rb'))
es_history_config = EsConfig(**configuration['elasticsearch']['history'])
es_probability_config = EsConfig(
**configuration['elasticsearch']['probability'])
es_rul_config = EsConfig(**configuration['elasticsearch']['rul'])
for model_name, configuration in configuration['models'].items():
configuration['datasource_params']['host'] = es_history_config.host
configuration['datasource_params'][
'index_name'] = es_history_config.index_name
configuration['datasource_params']['start_date'] = start_date
configuration['datasource_params']['end_date'] = end_date
configuration['transformer_params']['X_min'] = transformer_data[
'X_min']
configuration['transformer_params']['X_max'] = transformer_data[
'X_max']