question: D3 Part 2 Q2

data/setpoints/setpoints_1718013923.json

@@ -0,0 +1,4 @@
+{"unit": "dumpload_sp", "value": 0.0, "time": 1718013928.6948164}
+{"unit": "dumpload_sp", "value": 20.0, "time": 1718013959.6977577}
+{"unit": "dumpload_sp", "value": 10.0, "time": 1718013989.6989655}
+{"unit": "dumpload_sp", "value": 0.0, "time": 1718014018.716735}

demo_datalogger.py

@@ -1,3 +1,4 @@
+#!./venv/bin/python3
 import syslab
 from json import dump
 from time import sleep, time

demo_plotter.py

@@ -1,11 +1,13 @@
+#!./venv/bin/python3
 import pandas as pd
+import numpy as np
 import json
 import matplotlib.pyplot as plt
 import os
 from datetime import timedelta
 
 ## Read the measurements data file ##
-DATA_MEAS_DIR = 'data\measurements'
+DATA_MEAS_DIR = 'data/measurements'
 # Always plot latest datafile - replace [-1] with another index if you want to plot a specific file.
 MEAS_LOG_FILE = sorted(os.listdir(DATA_MEAS_DIR))[-1]
 
@@ -19,7 +21,7 @@ use_setpoint_log = False
 
 ## Read the setpoints data file ##
 if use_setpoint_log:
-    DATA_SP_DIR = 'data\setpoints'
+    DATA_SP_DIR = 'data/setpoints'
     # Always plot latest datafile
     SP_LOG_FILE = sorted(os.listdir(DATA_SP_DIR))[-1]
 
@@ -33,107 +35,39 @@ if use_setpoint_log:
 else:
     data = meas_data
 
+################################################################################
+################## Part 2 ######################################################
+################################################################################
+
+def overshoot(df, T1, T2):
+    yT1, yT2 = df[T1], df[T2]
+    over = 1 / (yT1 - yT2) * np.max(yT2 - df)
+    return over
+
+SETPOINT_UNIX = 1718013959.6977577
+SETPOINT_TS = pd.to_datetime(SETPOINT_UNIX, unit='s')
+WINDOW = pd.to_datetime(SETPOINT_UNIX+25, unit='s')
+
+## The controller is reasonably fast at reacting to changes; the sum of in and
+## out is at zero roughly 5-10 seconds after a change.
 
 # Construct a dataframe and pivot it to obtain a dataframe with a column per unit, and a row per timestamp.
 df = pd.DataFrame.from_records(data)
+df['time'] = pd.to_datetime(df['time'], unit='s')
 df_pivot = df.pivot_table(values='value', columns='unit', index='time')
-
+df_resampled = df_pivot.resample('0.1s').mean()
+df_resampled.interpolate(method='linear', inplace=True)
+df_resampled = pd.DataFrame(df_resampled)
 
 # Plot the data. Note, that the data will mostly not be plotted with lines.
 plt.ion()  # Turn interactive mode on
 plt.figure()
-ax1 = plt.subplot(211)  # Make two separate figures
+ax1, ax2 = plt.subplot(211), plt.subplot(212)
-ax2 = plt.subplot(212)
+df_resampled[[c for c in df_resampled.columns if '_p' in c]].plot(marker='.', ax=ax1, linewidth=3)
-df_pivot[[c for c in df_pivot.columns if "_p" in c]].plot(marker='.', ax=ax1, linewidth=3)
+ax2.plot(df_resampled['pcc_p'][SETPOINT_TS:WINDOW], marker='.', linewidth=3, label='pcc_p')
-df_pivot[[c for c in df_pivot.columns if "_q" in c]].plot(marker='.', ax=ax2, linewidth=3)
+ax2.plot(df_resampled['dumpload_p'][SETPOINT_TS:WINDOW], marker='.', linewidth=3, label='dumpload')
+plt.legend()
+
+# print(overshoot(df_resampled['pcc_p'][SETPOINT_TS:WINDOW], SETPOINT_TS, WINDOW))
+
 plt.show(block=True)
-
-
-## TODO Q1: Your code here
-
-
-
-
-
-## TODO Q2:
-# Convert time column (index) of df_pivot to datetime
-# TODO Your code here
-# Hint1: You can use pandas to_numeric() to prepare the index for pandas to_datetime function
-# Hint2: Remember to define the unit within pandas to_datetime function
-
-# Resample the data
-# TODO Your code here
-
-
-# Interpolate the measurements
-# TODO Your code here
-# Hint: For part two of the exercise ("collecting fresh data") the nan rows after a setpoint
-# in the recorded step function should be filled with the value of the setpoint until the row of the next setpoint is reached
-# You can use the df.fillna(method="ffill") function for that purpose. However, the measurements should still be interpolated!
-
-
-
-# Plot the resampled data
-# TODO Your code here
-
-
-
-## TODO Q3: Your code here
-
-
-
-
-## TODO Q4: Your code here
-
-
-
-## Part two: "Collecting fresh data"
-
-# Hint 1: You can build up on the "read_and_plot_data.py" from day 2
-# Hint 2: Yoy may want to store your response metric functions from day 2 in the "util.py" and import all of them with
-# "from util import *"
-
-if use_setpoint_log:
-
-    # Add a column to df_pivot containing the reference/target signal
-    # TODO your code here
-
-    # Loop over all steps and extract T_1, T_2 and the step size
-    results = {}
-
-    for idx in range(0, len(sp_data)-1):
-        label = f"Step_{sp_data[idx]['value']}kW"
-
-        # Extract T_1 and T_2 from the setpoint JSON
-        # TODO your code here
-
-
-        # Change timestamp format
-        T_1 = pd.to_datetime(pd.to_numeric(T_1), unit="s").round("0.1S")
-        T_2 = pd.to_datetime(pd.to_numeric(T_2), unit="s").round("0.1S")
-
-        # To ensure we are not considering values of the next load step
-        T_2 = T_2 - timedelta(seconds=0.2)
-
-
-        # define measured output y and target setpoint r
-        # TODO your code here
-
-        # Derive step direction from the setpoint data
-        if ...:  # TODO your code here
-            Positive_step = True
-        else:
-            Positive_step = False
-
-        # Collect response metrics results
-        results[label] = {
-            # TODO your code here
-        }
-
-    pd.DataFrame.from_dict(results).plot(kind='bar')
-    plt.title("Metrics")
-    plt.tight_layout()
-    plt.savefig('data/test_metrics'+MEAS_LOG_FILE[-10:]+'.png')
-    plt.show(block=True)
-
-

plotter.py

@@ -0,0 +1,28 @@
+#!./venv/bin/python3
+import pandas as pd
+import json
+import matplotlib.pyplot as plt
+import os
+
+DATA_MEAS_DIR = 'data/measurements'
+SPECIFIC_FILE = ''
+MEAS_LOG_FILE = sorted(os.listdir(DATA_MEAS_DIR))[-1] if not SPECIFIC_FILE else SPECIFIC_FILE
+
+with open(os.path.join(DATA_MEAS_DIR, MEAS_LOG_FILE)) as f:
+    meas_data = [json.loads(line) for line in f]
+data = meas_data
+
+df = pd.DataFrame.from_records(data)
+df['time'] = pd.to_datetime(df['time'], unit='s')
+df_pivot = df.pivot_table(values='value', columns='unit', index='time')
+df_resampled = df_pivot.resample('s').mean()
+df_resampled.interpolate(method='linear', inplace=True)
+df_resampled = pd.DataFrame(df_resampled)
+
+# Plot the data. Note, that the data will mostly not be plotted with lines.
+plt.ion()  # Turn interactive mode on
+plt.figure()
+ax1, ax2 = plt.subplot(211), plt.subplot(212)
+df_resampled[[c for c in df_resampled.columns if '_p' in c]].plot(marker='.', ax=ax1, linewidth=3)
+df_resampled[[c for c in df_resampled.columns if '_q' in c]].plot(marker='.', ax=ax2, linewidth=3)
+plt.show(block=True)