Update preparation.py

added the remove noise function in the same preparation.py module

preparation.py

@@ -99,3 +99,38 @@ def get_timeseries(path, datecolumn = 'date', datacolumn = 'CO2 (ppm)'):
 
 
     return co2_series
+
+def remove_noise(data, low_freq=None, high_freq=None):
+    """
+    Removes high/low frequency noise from a time-series by applying a bandpass filter in the frequency domain.
+
+    Parameters:
+    data (pd.Series): Time series data with a datetime index.
+    low_freq (float): The minimum frequency to retain; frequencies below this are considered noise and removed.
+    high_freq (float): The maximum frequency to retain; frequencies above this are considered noise and removed.
+    
+    Returns:
+    pd.Series: Filtered time series with noise removed, same length as the input.
+    """
+    # Ensure data is evenly spaced
+    n = len(data)
+    timestamp_sum = sum(data.index[i+1].timestamp() - data.index[i].timestamp() for i in range(n-1))
+    if not timestamp_sum / (n-1) == data.index[2].timestamp() - data.index[1].timestamp():
+        print("Data is not evenly spaced or data points are missing")
+        return None
+
+# FFT to get the frequency components
+    data_fft = np.fft.fft(data.values)
+    freqs = np.fft.fftfreq(n, d=(data.index[1] - data.index[0]).total_seconds())
+
+    # Apply the frequency filter
+    if low_freq is not None:
+        data_fft[np.abs(freqs) < low_freq] = 0
+    if high_freq is not None:
+        data_fft[np.abs(freqs) > high_freq] = 0
+
+    # Inverse FFT to get the filtered time-domain signal
+    filtered_data = np.fft.ifft(data_fft)
+
+    # Return as a pandas series with the original datetime index
+    return pd.Series(filtered_data.real, index=data.index)