new utilities for time dilation

solardatatools/utilities.py

@@ -7,6 +7,7 @@
 import sys
 import numpy as np
 import cvxpy as cvx
+from scipy.interpolate import interp1d
 
 
 def basic_outlier_filter(x, outlier_constant=1.5):
@@ -78,3 +79,44 @@ def find_runs(x):
         run_lengths = np.diff(np.append(run_starts, n))
 
         return run_values, run_starts, run_lengths
+
+
+def time_dilate(data, mask, power=8, scale=None):
+    """Process an entire PV power matrix at once
+    :return:
+    """
+    if scale is None:
+        scale = 1
+    N = 2**power
+    output = np.zeros((N, data.shape[1]))
+    xs_new = np.linspace(0, 1, N)
+    for col_ix in range(data.shape[1]):
+        y = data[:, col_ix] * scale
+        msk = mask[:, col_ix]
+        xs = np.linspace(0, 1, int(np.sum(msk)))
+        interp_f = interp1d(xs, y[msk])
+        resampled_signal = interp_f(xs_new)
+        if np.sum(resampled_signal) > 0:
+            output[:, col_ix] = resampled_signal
+        else:
+            output[:, col_ix] = 0
+    return output
+
+
+def undo_time_dilate(data, mask, scale=None):
+    if scale is None:
+        scale = 1
+    output = np.zeros_like(mask, dtype=float)
+    xs_old = np.linspace(0, 1, data.shape[0])
+    for col_ix in range(data.shape[1]):
+        # the number of non-zero data points on that day
+        n_pts = np.sum(mask[:, col_ix])
+        # a linear space of n_pts values between 0 and 1
+        xs_new = np.linspace(0, 1, n_pts)
+        # the mapping from the old index to the data
+        interp_f = interp1d(xs_old, data[:, col_ix] / scale)
+        # transform the signal to the new length
+        resampled_signal = interp_f(xs_new)
+        # insert the data into the daytime index values
+        output[mask[:, col_ix], col_ix] = resampled_signal
+    return output