Detect passing valves

detect_passing_valves/README.md

@@ -0,0 +1,5 @@
+# Detect passing valves in HVAC equipment
+
+This app detects valves in HVAC equipment that do not close fully even when actuated to a fully closed position, also known as “passing valves”. The application compares fluid temperatures upstream and downstream from the valve and calculates the expected long-term difference between the two fluid streams when the valve is currently closed, and has been closed for some time. The app then analyzes the expected trends with the actual data to determine if the valve is in good operating condition or malfunctioning.
+
+This app produces a plot for each analyzed valve with file name formatted as `<site name>-<equipment name>-<valve name>.png` when run. Each plot shows a solid green horizontal line that represents the average temperature difference between the downstream and upstream fluids when valve is commanded closed, a dashed purple line shows the expected correct operating behavior trend of the valve (based on the green points), a solid pink horizontal line shows the average temperature difference when the application detected a possible  passing valve (based on the red points). The ‘Bad ratio’ value is a ratio of the number of bad operating point values to goodoperating point values. Each file is either allocated in a 'good' or 'bad' folder depending if the analyzed valve is operating correctly (good folder) or malfunctioning (bad folder).

detect_passing_valves/aggregate_results.py

@@ -0,0 +1,141 @@
+from os.path import join
+import os
+import pandas as pd
+import numpy as np
+
+def parse_dict_list_file(line):
+
+    dictionary = dict()
+    pairs = line.strip().strip(",").strip('{}').split(', ')
+    for pr in pairs:
+        pair = pr.split(': ')
+        dictionary[pair[0].strip('\'\'\"\"')] = pair[1].strip('\'\'\"\"')
+
+    return dictionary
+
+def separate_heat_transfer_col(df, col):
+    heat_tfr_pwer = []
+    heat_tfr_enrg = []
+    for hl in df[col]:
+        if pd.notnull(hl):
+            hl = hl.strip('()').split(", ")
+            heat_tfr_pwer.append(float(hl[0]))
+            heat_tfr_enrg.append(float(hl[1]))
+        else:
+            heat_tfr_pwer.append(np.nan)
+            heat_tfr_enrg.append(np.nan)
+
+    return heat_tfr_pwer, heat_tfr_enrg
+
+if __name__ == '__main__':
+    dat_folder = join('with_airflow_checks_year_start', 'csv_data')
+    project_folder = join('./', 'external_analysis')
+
+    mortar_dat_file = join(project_folder, 'MORTAR', 'lg_4hr_shrt_1hr_test_no_aflw_req_10C_threshold')
+    bear_dat_file = join(project_folder, 'bldg_trc_rs', 'lg_4hr_shrt_1hr_test_no_aflw_req_10C_threshold')
+    lion_dat_file = join(project_folder, 'bldg_gt_pr', 'lg_4hr_shrt_1hr_test_no_aflw_req_10C_threshold')
+
+    all_datasets = []
+    for dataset in [mortar_dat_file, bear_dat_file, lion_dat_file]:
+        # Perform additional analysis
+        f = open(join(dataset, 'minimum_airflow_values.txt'), 'r')
+        lines = f.readlines()
+        f.close()
+
+        vav_results = []
+        for line in lines:
+            vav_results.append(parse_dict_list_file(line))
+
+        vav_results = pd.DataFrame.from_records(vav_results)
+        all_datasets.append(vav_results)
+
+    all_datasets = pd.concat(all_datasets).reset_index(drop=True)
+
+    # clean dataframe
+    numeric_cols = ['minimum_air_flow_cutoff', 'long_t', 'long_tbad', 'bad_ratio', 'long_to']
+    avail_cols = list(set(all_datasets.columns).intersection(set(numeric_cols)))
+    all_datasets[avail_cols] = all_datasets[avail_cols].apply(pd.to_numeric, errors='coerce')
+
+    all_datasets = all_datasets.dropna(subset=['folder'])
+
+    all_datasets['folder_short'] = all_datasets['folder'].apply(os.path.basename)
+
+
+    # add rest of info including heat loss due to passing valves
+    all_passing_valve_results = []
+    for dataset in [mortar_dat_file, bear_dat_file, lion_dat_file]:
+        final_df = pd.read_csv(join(dataset, 'passing_valve_results.csv'), index_col=0)
+        all_passing_valve_results.append(final_df)
+
+    all_passing_valve_results = pd.concat(all_passing_valve_results).reset_index(drop=True)
+
+    all_datasets = pd.merge(all_datasets, all_passing_valve_results, how='left', on=['vlv', 'site', 'equip'])
+
+    # separate heat rate loss from energy loss
+    heat_loss_pwer, heat_loss_enrg = separate_heat_transfer_col(df=all_datasets, col='heat_loss_pwr-avg_nrgy-sum')
+
+    all_datasets['heat_loss_pwr'] = heat_loss_pwer
+    all_datasets['heat_loss_enrg'] = heat_loss_enrg
+
+    # separate intentional heat rate
+    heat_intend_pwer, heat_intend_enrg = separate_heat_transfer_col(df=all_datasets, col='heat_intentional_pwr-avg_nrgy-sum')
+
+    all_datasets['heat_intend_pwr'] = heat_intend_pwer
+    all_datasets['heat_intend_enrg'] = heat_intend_enrg
+
+
+    # summary statistics on long term difference for each site
+    site_grp = all_datasets.groupby(['site'])
+    folder_grp = all_datasets.groupby(['folder_short'])
+    vav_results_grp = all_datasets.groupby(['site', 'folder_short'])
+    grp_dat_stats = vav_results_grp['long_t'].describe()
+    folder_grp['long_t'].describe()
+
+    # summary statistics on long term difference for all data
+    agg_dat_grp = all_datasets.groupby(['folder_short'])
+    agg_dat_stats = agg_dat_grp['long_t'].describe()
+
+
+    # summary statistics on heat loss due to passing valves
+    no_sensor_faults = all_datasets.loc[all_datasets["folder_short"] != "sensor_fault"]
+    folder_grp_nsf = no_sensor_faults.groupby(['folder_short'])
+    heat_loss = folder_grp_nsf["heat_loss_enrg"].sum().sum()
+    heat_intent = folder_grp_nsf["heat_intend_enrg"].sum().sum()
+
+    heat_loss_ratio = heat_loss/heat_intent
+
+    # on aggregate
+    no_sensor_faults["long_term_fail_num_times_detected"].describe()
+    no_sensor_faults["long_term_fail_avg_minutes"].describe()
+
+    no_sensor_faults["short_term_fail_num_times_detected"].describe()
+    no_sensor_faults["short_term_fail_avg_minutes"].describe()
+
+    no_sensor_faults["heat_loss_pwr"].describe()
+    no_sensor_faults["heat_loss_enrg"].sum()/1000
+
+    # by site
+    from_btuhr_to_watts = 0.293071
+    site_grp_nsf = no_sensor_faults.groupby(['site'])
+
+    site_heat_loss = site_grp_nsf["heat_loss_enrg"].sum()
+    site_heat_intend = site_grp_nsf["heat_intend_enrg"].sum()
+
+    site_heat_loss/site_heat_intend
+
+    # by fault category
+    site_grp_nsf["heat_loss_pwr"].describe()*from_btuhr_to_watts
+
+    # all
+    all_datasets.loc[all_datasets["folder_short"] == "bad_valves", "heat_loss_enrg"]
+    all_datasets.loc[:,"heat_intend_enrg"]
+
+    all_datasets["heat_loss_pwr"].describe()*from_btuhr_to_watts
+
+    subset_lion = np.logical_and(all_datasets["site"] == "lion", all_datasets["folder_short"] == "bad_valves")
+    df_bad_lion = all_datasets.loc[subset_lion]
+
+    df_bad_lion["loss_pct"] = df_bad_lion["heat_loss_enrg"]/df_bad_lion["heat_intend_enrg"]
+    df_bad_lion.sort_values("loss_pct", ascending=False)
+
+