clustering/light-fcm-clustering/env_data.py

#!/usr/bin/env python
# -*- coding: utf-8 -*-
# time: 2023/6/12 13:24
# file: env_data.py
# author: David
# company: shenyang JY
import datetime
import math
import pandas as pd
import numpy as np


def process_env_data():
    path = './xiangzhou/weather/weather-1-process.csv'
    envn = pd.read_csv(path, usecols=['C_TIME', 'C_GLOBALR', 'C_DIFFUSER', 'C_RH'])     # C_GLOBALR 总辐射 C_DIFFUSER 散辐射 C_RH 湿度
    envn['C_TIME'] = pd.to_datetime(envn['C_TIME'])
    path1 = './xiangzhou/power5.csv'
    power = pd.read_csv(path1, usecols=['C_TIME', 'C_REAL_VALUE'])
    power['C_TIME'] = pd.to_datetime(power['C_TIME'])
    envn = pd.merge(envn, power, on='C_TIME')
    envn_filter = envn[envn['C_GLOBALR'] > 0].reset_index(drop=True)  # 过滤夜间环境 （总辐射大于0）
    envn = normalize(envn)
    pre = envn_filter.iloc[0, 0].hour
    envn.set_index('C_TIME', inplace=True)
    envs, env = [], []
    for index, row in envn_filter.iterrows():
        if pre != row[0].hour:
            con = pd.concat(env, axis=1).T
            # 数据不齐，要么联立后缺点，要不是日出或日落时分
            if len(con) != 12:
                con = envn.loc[str(con.iloc[0, 0])[:-6]].reset_index()
                print("数据不齐，该时间点为：", row[0], "新的长度为：", len(con))
            envs.append(con.reset_index(drop=True))
            pre = row[0].hour
            env = [row]
        else:
            env.append(row)
    return envs


def envn_features(envs, path):
    for i, env in enumerate(envs):
        zero_indexs = env[env['C_GLOBALR'] == 0].index
        print("----", env)
        if len(zero_indexs) > 0:
            env.iloc[zero_indexs, env.columns.get_loc('C_GLOBALR')] = 0.1
        print("++++", env)
        x = list(map(lambda x,y: x/y, env['C_DIFFUSER'], env['C_GLOBALR']))
        f1 = round(np.mean(x), 2)
        env['diff1'] = env['C_REAL_VALUE'].diff()
        env['diff_1'] = env['C_REAL_VALUE'].diff(-1)
        ei = (env['diff1']*env['diff_1']).tolist()[1:-1]
        ei = [1 if e > 0 else 0 for e in ei]
        f2 = round(np.mean(ei), 2)
        f3 = round(np.mean(env['C_RH'].tolist()), 2)
        time = env.iloc[-1]['C_TIME'].replace(minute=0)
        time += datetime.timedelta(hours=1)     # 反应的是下一个小时的环境特征
        envs[i] = [time, f1, f2, f3]
    envn_features = pd.DataFrame(envs, columns=['C_TIME', 'f1', 'f2', 'f3'])
    # envn_features = normalize(envn_features)
    envn_features.to_csv(path, index=False)


def normalize(df):
    """
    暂时不将C_TIME归一化
    :param dfs:
    :return: 归一化后的DataFrame
    """
    df1 = df.iloc[:, 1:]
    mean = np.mean(df1, axis=0)  # 数据的均值
    std = np.std(df1, axis=0)  # 标准差
    print("归一化参数，均值为：{}，方差为：{}".format(mean.to_dict(), std.to_dict()))
    df_Zscore = df1.apply(lambda x: np.around((x - x.mean())/math.sqrt(sum((x-x.mean())**2/len(x))), decimals=2))
    df_Zscore.insert(0, 'C_TIME', df["C_TIME"])
    return df_Zscore


if __name__ == '__main__':
    feaP = './xiangzhou/features.csv'
    envs = process_env_data()
    envn_features(envs, feaP)