CNN_BiLSTM_Attention模型，LSTM模型，DNN(全连接神经网络)三种模型对短期日负荷曲线的预测

# -*- coding: utf-8 -*- """ Created on Wed Apr 13 11:02:02 2022 @author: pony """ import matplotlib.pyplot as plt import pandas as pd import numpy as np import pickle import keras import warnings warnings.filterwarnings("ignore") data = pd.read_csv("p.csv") parts = 14 this_one = data.iloc[parts:]#取14行以后数据 bak_index = this_one.index#取索引 for k in range(1, parts + 1): last_one = data.iloc[(parts - k):(this_one.shape[0] - k + parts)] this_one.set_index(last_one.index, drop=True, inplace=True) this_one = this_one.join(last_one, lsuffix="", rsuffix="_p" + str(k)) this_one.set_index(bak_index, drop=True, inplace=True) this_one = this_one.fillna(0)#用0填充缺失值 t0 = this_one.iloc[:, 0:96]#取0-96列 t0_min = t0.apply(lambda x: np.min(x), axis=0).values#每一列的最小值 t0_ptp = t0.apply(lambda x: np.ptp(x), axis=0).values#最大值和最小值的差值 this_one = this_one.apply(lambda x: (x - np.min(x)) / np.ptp(x), axis=0)#归一化 ''' 划分训练集与测试集 ''' test_data = this_one.iloc[-30:]#最后30天当做测试集 train_data = this_one.iloc[:-30] train_y_df = train_data.iloc[:, 0:96]#取前96列当做标签 train_y = np.array(train_y_df) train_x_df = train_data.iloc[:, 96:]#取96列以后，也就是前14天的数据作为特征输入 train_x = np.array(train_x_df) test_y_df = test_data.iloc[:, 0:96] test_y = np.array(test_y_df) test_x_df = test_data.iloc[:, 96:] test_x = np.array(test_x_df) test_y_real = t0.iloc[-30:]#测试集归一化前的真实标签 ''' 建立模型 ''' import keras import pickle init = keras.initializers.glorot_uniform(seed=1) simple_adam = keras.optimizers.Adam() model = keras.models.Sequential() model.add(keras.layers.Dense(units=512, input_dim=1434, kernel_initializer=init, activation='relu')) model.add(keras.layers.Dense(units=256, kernel_initializer=init, activation='relu')) model.add(keras.layers.Dense(units=128, kernel_initializer=init, activation='relu')) model.add(keras.layers.Dropout(0.1)) model.add(keras.layers.Dense(units=96, kernel_initializer=init, activation='tanh')) model.compile(loss='mse', optimizer=simple_adam, metrics=['accuracy']) model.fit(train_x, train_y, epochs=100, batch_size=32, shuffle=True, verbose=True) pred_y = model.predict(test_x) pred_y = (pred_y*t0_ptp)+t0_min import pickle import matplotlib.pyplot as plt import matplotlib # 以下 font.family 设置仅适用于 Mac系统，其它系统请使用对应字体名称 matplotlib.rcParams['font.family'] = 'Arial Unicode MS' import numpy as np realdf = test_y*t0_ptp[0:96]+t0_min[0:96] base = 0 error = 0 plt.figure(figsize=(20, 10)) for index in range(0, 30): real_array = realdf[index][0:96] pred_array = pred_y[index][0:96] pred_array[np.where(pred_array < 0)] = 0 plt.subplot(5, 7, index + 1) plt.plot(range(96), real_array, '-', label="real",c='red') plt.plot(range(96), pred_array, '--', label="pred",c='green') plt.ylim(0, 250) base = base + np.sum(real_array) error = error + np.sum(np.abs(real_array-pred_array)) plt.show() v = 100*(1-error/base) print("Evaluation on test data: accuracy = %0.2f%% \n" % v) # Evaluation on test data: accuracy = 74.95%