"""
Lab 4 - Distribution modelling (normal distribution) for PV inverter data
Reads the ORIGINAL sheet exported from Excel directly (no manual pre-cleaning),
using the same reading technique fixed for Labs 1, 2 and 3.
"""

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from scipy.stats import norm

FILE = "PV2013January.csv"


def load_pv_sheet(path):
    """
    Load one day's PV sheet exactly as Excel exports it:
      line 1: empty
      line 2: sensor names (TimeStamp / SENS0500... / WR11LV02...)
      line 3: the real column names (TimeStamp, ExlSolIrr, IntSolIrr, ...)
      line 4: units (hh:mm, W/m^2, W/m^2, ...)
      line 5+: data, with decimal numbers written as "11227,85" (comma
               instead of a dot, wrapped in quotes because a comma is
               otherwise the column separator).
    Returns a DataFrame with a proper TimeStamp column and every other
    column converted to a real number.
    """
    df = pd.read_csv(path, skiprows=2)             # ignore the 2 rows above the header
    df = df.drop(index=0).reset_index(drop=True)   # drop the units row
    df = df.rename(columns={df.columns[0]: "TimeStamp"})

    numeric_cols = df.columns.drop("TimeStamp")
    df[numeric_cols] = df[numeric_cols].apply(
        lambda col: pd.to_numeric(col.astype(str).str.replace(",", ".", regex=False), errors="coerce")
    )
    return df


df = load_pv_sheet(FILE)
print("Loaded", len(df), "rows,", len(df.columns), "columns")
print(df[["TimeStamp", "TmpMdul C", "Fac", "Pac"]].head())


def fit_and_report(name, data):
    """Compute mean and standard deviation, the parameters of the modelled
    normal distribution N(mean, std)."""
    mu = data.mean()
    sigma = data.std()
    print(f"{name}: n={len(data)}  mean={mu:.3f}  std={sigma:.3f}")
    return mu, sigma


def plot_hist_with_normal(data, mu, sigma, xlabel, title, filename, color):
    """Plot the real data as a density histogram, with the modelled normal
    PDF -- built from the data's own mean and std -- drawn on top."""
    fig, ax = plt.subplots(figsize=(6, 4.2))
    ax.hist(data, bins=30, density=True, color=color, alpha=0.55, edgecolor="white", label="real data")
    if sigma > 0:
        xs = np.linspace(data.min(), data.max(), 300)
        ax.plot(xs, norm.pdf(xs, mu, sigma), color="black", linewidth=2,
                 label=f"modelled N({mu:.2f}, {sigma:.2f})")
    ax.set_xlabel(xlabel)
    ax.set_ylabel("density")
    ax.set_title(title)
    ax.legend()
    fig.tight_layout()
    fig.savefig(filename, dpi=150)
    plt.close(fig)


# =====================================================================
# Column F: TmpMdul C (module temperature)
# =====================================================================
tmod = df["TmpMdul C"]
print("\n--- F: TmpMdul C ---")
mu_f, sigma_f = fit_and_report("TmpMdul C (raw)", tmod)
plot_hist_with_normal(tmod, mu_f, sigma_f, "Module temperature, C",
                       "F -- TmpMdul C: real data vs. modelled normal (raw)",
                       "F_normal_raw.png", "#2F3C7E")

# TmpMdul C has no physical zeros to filter (module temperature is never
# exactly 0 C in this dataset), but we check anyway, per the general
# "filter if needed" instruction established in Lab 3.
tmod_nz = tmod[tmod != 0]
if len(tmod_nz) != len(tmod):
    mu_f2, sigma_f2 = fit_and_report("TmpMdul C (zeros removed)", tmod_nz)
    plot_hist_with_normal(tmod_nz, mu_f2, sigma_f2, "Module temperature, C",
                           "F -- TmpMdul C: real data vs. modelled normal (filtered)",
                           "F_normal_filtered.png", "#2F3C7E")
else:
    print("No zeros found in TmpMdul C -- filtered model is identical to the raw one.")

# =====================================================================
# Column I: Fac (AC / grid frequency) -- dominated by zeros (inverter idle)
# =====================================================================
fac = df["Fac"]
print("\n--- I: Fac ---")
mu_i, sigma_i = fit_and_report("Fac (raw, incl. zeros)", fac)
plot_hist_with_normal(fac, mu_i, sigma_i, "Fac, Hz",
                       "I -- Fac: real data vs. modelled normal (raw, incl. zeros)",
                       "I_normal_raw.png", "#F96167")

fac_nz = fac[fac > 0]
mu_i2, sigma_i2 = fit_and_report("Fac (zeros removed)", fac_nz)
plot_hist_with_normal(fac_nz, mu_i2, sigma_i2, "Fac, Hz",
                       "I -- Fac: real data vs. modelled normal (zeros removed)",
                       "I_normal_filtered.png", "#F96167")

# =====================================================================
# Column P: Pac (AC power) -- also dominated by zeros (no generation at night)
# =====================================================================
pac = df["Pac"]
print("\n--- P: Pac ---")
mu_p, sigma_p = fit_and_report("Pac (raw, incl. zeros)", pac)
plot_hist_with_normal(pac, mu_p, sigma_p, "Pac, W",
                       "P -- Pac: real data vs. modelled normal (raw, incl. zeros)",
                       "P_normal_raw.png", "#2F3C7E")

pac_nz = pac[pac > 0]
mu_p2, sigma_p2 = fit_and_report("Pac (zeros removed)", pac_nz)
plot_hist_with_normal(pac_nz, mu_p2, sigma_p2, "Pac, W",
                       "P -- Pac: real data vs. modelled normal (zeros removed)",
                       "P_normal_filtered.png", "#2F3C7E")

print("\nAll figures saved.")