能耗

1import pandas as pd
2import matplotlib.pyplot as plt
3
4df = pd.read_csv('AEP.csv')
5
6# 转换数据类型
7df["Date"] = pd.to_datetime(df['Date'], format='%Y-%m-%d')
8 
9# 查看数据框的前几行和后几行
10print(df.head())
11print(df.tail())
12
13# 将日期设为索引
14df = df.set_index('Date')
15
16# 绘图
17fig, ax = plt.subplots(figsize=(7, 4.5), dpi=300)
18ax.plot(df)
19
20# 添加标签
21plt.xlabel("日期")
22plt.ylabel("兆瓦（MW）")
23plt.title("能源消耗")
24
25fig.savefig("output/output.png")
26plt.close(fig)

1import pandas as pd
2import matplotlib.pyplot as plt
3from statsmodels.tsa.seasonal import seasonal_decompose
4
5df = pd.read_csv('AEP.csv')
6
7# 转换数据类型
8df["Date"] = pd.to_datetime(df['Date'], format='%Y-%m-%d')
9 
10# 将日期设为索引
11df = df.set_index('Date')
12
13# 去除异常值
14df = df[:-1]
15
16# 季节性分解
17result = seasonal_decompose(
18    df['AEP_MW'],
19    model='additive',
20    period=12,
21    extrapolate_trend='freq'
22    )
23
24# 绘制季节性分解图
25fig = result.plot(observed=False)
26fig.savefig("output/output.png")
27plt.close(fig)

1import pandas as pd
2import matplotlib.pyplot as plt
3from statsmodels.graphics.tsaplots import plot_acf, plot_pacf
4import warnings
5
6warnings.filterwarnings("ignore")
7
8df = pd.read_csv('AEP.csv')
9
10# 转换数据类型
11df["Date"] = pd.to_datetime(df['Date'], format='%Y-%m-%d')
12 
13# 将日期设为索引
14df = df.set_index('Date')
15
16# 去除异常值
17df = df[:-1]
18
19# 对数据进行一次差分
20df_dif = df['AEP_MW'] - df['AEP_MW'].shift(1)
21df_dif = df_dif - df_dif.shift(12)
22df_dif = df_dif.dropna()
23
24# 绘制自相关图（ACF）和偏自相关图（PACF）
25fig = plt.figure(figsize=(6,5))
26
27ax1 = fig.add_subplot(211)
28ax1.set_xlabel('滞后阶数')
29ax1.set_ylabel('相关性')
30plot_acf(df_dif, ax=ax1, alpha=0.05)
31
32ax2 = fig.add_subplot(212)
33ax2.set_xlabel('滞后阶数')
34ax2.set_ylabel('偏相关性')
35plot_pacf(df_dif, ax=ax2, alpha=0.05)
36
37plt.subplots_adjust(left=0.2,
38                    bottom=0.1,
39                    right=0.9,
40                    top=0.9,
41                    wspace=0.4,
42                    hspace=0.4)
43
44# 保存图像
45fig.savefig("output/output.png")
46plt.close(fig)

1import pandas as pd
2import matplotlib.pyplot as plt
3from sklearn.metrics import *
4from statsmodels.tsa.statespace.sarimax import SARIMAX
5from statsmodels.tsa.holtwinters import ExponentialSmoothing
6import warnings
7
8warnings.filterwarnings("ignore")
9
10df = pd.read_csv('AEP.csv')
11
12# 转换数据类型
13df["Date"] = pd.to_datetime(df['Date'], format='%Y-%m-%d')
14 
15# 将日期设为索引
16df = df.set_index('Date')
17
18# 去除异常值
19df = df[:-1]
20
21# 计算朴素预测的时间滞后
22freq = 12
23df['t-'+str(freq)] = df['AEP_MW'].shift(freq)
24
25# 划分训练集和测试集
26df_train = df[:-freq]
27df_test = df[-freq:]
28
29# 使用朴素预测作为基准模型
30naive_forecast = df_test['t-'+str(freq)]
31baseline_error = round(mean_absolute_error(df_test['AEP_MW'], naive_forecast))
32
33# 使用SARIMA模型进行预测
34sarima_results = SARIMAX(
35    df_train['AEP_MW'], order=(12,1,12)).fit(disp=0)
36
37sarima_predictions = sarima_results.predict(
38    start=len(df_train),
39    end=len(df_train)+freq-1
40    )
41
42sarima_error =  round(mean_absolute_error(df_test['AEP_MW'], sarima_predictions))
43
44# 使用Holt-Winters模型进行预测
45hw_results = ExponentialSmoothing(
46    df_train['AEP_MW'],
47    trend='multiplicative',
48    seasonal='multiplicative').fit()
49
50hw_predictions = hw_results.predict(
51    start=df_test.index.min(), end=df_test.index.max()
52    )
53
54hw_error = round(mean_absolute_error(df_test['AEP_MW'], hw_predictions))
55
56# 绘图对比
57fig, ax = plt.subplots(figsize=(7, 4.5), dpi=300)
58
59ax.plot(df['AEP_MW'][-freq:], label = "真实值")
60ax.plot(naive_forecast, label = "基准模型 (MAE={0})".format(baseline_error))
61ax.plot(sarima_predictions, label = "SARIMA (MAE={0})".format(sarima_error))
62ax.plot(hw_predictions, label = "Holt-Winters (MAE={0})".format(hw_error))
63plt.title("能源消耗")
64
65plt.legend()
66fig.savefig("output/output.png")
67plt.close(fig)

1import pandas as pd
2import matplotlib.pyplot as plt
3from sklearn.metrics import *
4from statsmodels.tsa.statespace.sarimax import SARIMAX
5from statsmodels.tsa.holtwinters import ExponentialSmoothing
6from datetime import timedelta
7import warnings
8
9warnings.filterwarnings("ignore")
10
11df = pd.read_csv('AEP.csv')
12
13# 转换数据类型
14df["Date"] = pd.to_datetime(df['Date'], format='%Y-%m-%d')
15 
16# 将日期设为索引
17df = df.set_index('Date')
18
19# 去除异常值
20df = df[:-1]
21
22# 使用Holt-Winters模型进行未来预测
23freq=12
24hw_results = ExponentialSmoothing(
25    df['AEP_MW'],
26    trend='multiplicative',
27    seasonal='multiplicative', freq='M').fit()
28
29hw_predictions = hw_results.predict(
30    start=df.index.max()+timedelta(days=30),
31    end=df.index.max()+timedelta(days=12*30)
32    )
33
34# 绘图展示历史数据与预测结果
35fig, ax = plt.subplots(figsize=(7, 4.5), dpi=300)
36
37ax.plot(df['AEP_MW'], label = "真实值")
38ax.plot(hw_predictions, label = "Holt-Winters 预测")
39plt.title("能源消耗")
40
41plt.legend()
42fig.savefig("output/output.png")
43plt.close(fig)