cnn-bigru-attention对电池寿命进行预测

import os from random import seed import matplotlib import matplotlib.pyplot as plt import numpy as np import pandas as pd from sklearn.metrics import mean_absolute_error from tensorflow.python.framework.random_seed import set_random_seed from tensorflow.python.keras import Input from tensorflow.python.keras.layers import Bidirectional, GRU from tensorflow.python.keras.layers import Dense, Conv1D from tensorflow.python.keras.layers import Flatten from tensorflow.python.keras.layers import Permute, Multiply from tensorflow.python.keras.models import Model matplotlib.rcParams['font.sans-serif'] = ['SimHei'] # 用黑体显示中文 set_random_seed(11) seed(7) def attention_3d_block(inputs): input_dim = int(inputs.shape[2]) a = inputs a = Dense(input_dim, activation='softmax')(a) a_probs = Permute((1, 2), name='attention_vec')(a) output_attention_mul = Multiply()([inputs, a_probs]) return output_attention_mul def bp_lstm(lstm_units, dense1, look_back): inputs = Input(shape=(look_back, INPUT_DIMS)) x = Conv1D(filters=128, kernel_size=1, activation='relu')(inputs) # , padding = 'same' lstm_out = Bidirectional(GRU(lstm_units, return_sequences=True, activation='relu'), name='bilstm')(x) # For GPU you can use CuDNNLSTM cpu LSTM # lstm_out = Bidirectional(CuDNNLSTM(lstm_units, return_sequences=True))(x) attention_mul = attention_3d_block(lstm_out) attention_mul = Flatten()(attention_mul) output = Dense(dense1)(attention_mul) output = Dense(pre_year, activation='linear')(output) model = Model(inputs=[inputs], outputs=output) return model def fit_size(x, y): from sklearn import preprocessing x_MinMax = preprocessing.MinMaxScaler() y_MinMax = preprocessing.MinMaxScaler() x = x_MinMax.fit_transform(x) y = y_MinMax.fit_transform(y) return x, y, y_MinMax # 预测未来1次的电池容量 def create_dataset(dataset, train_y, look_back): dataX, dataY = [], [] for i in range(len(dataset) - look_back - 1): a = dataset[i:(i + look_back), :] dataX.append(a) y = train_y[(i + look_back):(i + look_back + 1), 0].tolist() dataY.append(y) TrainX = np.array(dataX) Train_Y = pd.DataFrame(dataY).values return TrainX, Train_Y src = r'C:\Projects\CNN-LSTM-attention+cnn+lstm三个模型\锂电池状态评估\Batterydataset (2)\Batterydataset\\' os.chdir(src) train_path = 'B0005.csv' df = pd.read_csv(train_path, encoding='gbk') # train_size = int(df.shape[0] * 0.8) # 0.036 train_size = int(df.shape[0] * 0.7) train = df.iloc[:train_size, :] test = df.iloc[train_size:, :] train_y = df.values[:train_size, 0].reshape(-1, 1) test_y = df.values[train_size:, 0].reshape(-1, 1) # 归一化 train_x, train_y, train_y_MinMax = fit_size(train, train_y) test_x, test_y, test_y_MinMax = fit_size(test, test_y) # TRAIN INPUT_DIMS = train.shape[1] # 预测未来1次的容量 pre_year = 1 # 调参方法 def objective(trial): lr = trial.suggest_uniform('lr', 0.002, 0.1) batch_size = trial.suggest_int('batch_size', 4, 6) lstm_units = trial.suggest_int('lstm_units', 33, 55) dense1 = trial.suggest_int('dense1', 54, 99) epochs = trial.suggest_int('epochs', 10, 100) look_back = trial.suggest_int('look_back', 1, 9) train_X, train_Y = create_dataset(train_x, train_y, look_back) train_Y_data = pd.DataFrame(train_Y) train_Y_data.dropna(inplace=True) train_X = train_X[:train_Y_data.shape[0]] train_Y = train_Y_data.values train_X = train_X.astype('float64') m = bp_lstm(lstm_units, dense1, look_back) m.compile(loss='mae', optimizer='adam') hist = m.fit(train_X, train_Y, epochs=epochs, batch_size=batch_size, verbose=0) mae = hist.history['loss'][-1] return mae parameters = {'lr': 0.01782551732488114, 'batch_size': 6, 'lstm_units': 43, 'dense1': 83, 'epochs': 50, 'look_back': 2} lstm_units = parameters['lstm_units'] lr = parameters['lr'] look_back = parameters['look_back'] batch_size = parameters['batch_size'] dense1 = parameters['dense1'] epochs = parameters['epochs'] train_X, train_Y = create_dataset(train_x, train_y, look_back) m = bp_lstm(lstm_units, dense1, look_back) m.compile(loss='mae', optimizer='Adam') hist = m.fit(train_X, train_Y, epochs=epochs, batch_size=batch_size) test_X, test_Y = create_dataset(test_x, test_y, look_back) test_X = test_X.astype('float64') pred_y = m.predict(test_X) mae = mean_absolute_error(test_Y, pred_y) print(mae) inv_pred_y = test_y_MinMax.inverse_transform(pred_y.reshape(-1, 1)) # 画预测和真实值的对比图 plt.plot([x for x in range(train_size, 164)], inv_pred_y, color="red", label="pred") plt.plot([x for x in range(167)], df.values[:, 0], color="blue", label="pred") plt.ylabel('SOH') plt.xlabel('循环次数') plt.title('SOH pred vs test') plt.legend(['SOH预测值', 'SOH真实值'], loc='best') plt.savefig("SOH_test_real_pred.png", dpi=500, bbox_inches='tight') plt.show()