2021-06-21bp负荷预测的代码

import numpy as np
import matplotlib.pyplot as plt
from sklearn.preprocessing import MinMaxScaler,StandardScaler
from sklearn.metrics import r2_score
from sklearn.neural_network import MLPRegressor

import matplotlib as mpl
mpl.rcParams['font.sans-serif'] = ['KaiTi']
mpl.rcParams['font.serif'] = ['KaiTi']         #解决matplotlib无法显示中文的问题
plt.rcParams['axes.unicode_minus']=False       #解决负数坐标显示问题
# In[]
def split_data(data,n):
    # 前n时刻的值为输入
    # 预测n+1个时刻的值
    in_=[]
    out_=[]
    N=data.shape[0]-n
    for i in range(N):
        in_.append(data[i:i+n,])
        out_.append(data[i+n,])
    in_=np.array(in_).reshape(-1,n)
    out_=np.array(out_).reshape(-1,1)
    return in_,out_

   
fname='7day'# 7day 30day 60day
y=np.load('result/'+fname+'.npz')['data_imfs']

test_pred_each_mode,test_real_each_mode=[],[]
imf=y

train_again=0#是否重新训练测试
if train_again:
    for mode in range(imf.shape[0]):
        print('对第',mode+1,'个分量进行bp回归建模')
        
        in_,out_=split_data(imf[mode,:],n=10)
    
        m=-144#最后一天做测试集
        pt=in_[:m,:]
        tt=out_[:m,:]
        pe=in_[m:,:]
        te=out_[m:,:]
        # 归一化
        #ss_X=MinMaxScaler().fit(pt)
        #ss_y=MinMaxScaler().fit(tt)
        ss_X=StandardScaler().fit(pt)
        ss_y=StandardScaler().fit(tt)
        
        pt = ss_X.transform(pt)
        train_y = ss_y.transform(tt)
        pe = ss_X.transform(pe)
        test_y = ss_y.transform(te)
        
        model = MLPRegressor(
            hidden_layer_sizes=(6),  activation='relu', solver='adam', alpha=0.0001, batch_size='auto',
            learning_rate='constant', learning_rate_init=0.001, power_t=0.5, max_iter=50,random_state=1)
        model.fit (pt,train_y)
        yhat=model.predict(pe).reshape(-1,1)
        
        
        test_pred = ss_y.inverse_transform(yhat)
        test_label = ss_y.inverse_transform(test_y)

        
        test_pred_each_mode.append(test_pred)
        test_real_each_mode.append(test_label)
        print('-----------')