使用python编写基于紫外-可见光谱的水质COD检测代码，读取三份excel文件，分别为10mg/L,15mg/L,20mg/L的标准溶液，excel文件中第一列数据为波长，第二列为透过率，对数据通过导数法进行预处理，得到一阶导数谱图，并绘制在同一个折线图上，再计算一阶导数谱的排列熵，通过排列熵算法进行特征波长的提取，最后，利用提取的特征波长处的特征值与对应的COD浓度进行PLS建模，得到COD的预测模型

首先，需要安装并导入所需的库，包括`pandas`、`numpy`、`matplotlib`、`scipy`和`sklearn`。可以使用以下代码安装和导入这些库：

!pip install pandas numpy matplotlib scipy scikit-learn
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from scipy.signal import savgol_filter
from sklearn.cross_decomposition import PLSRegression

接下来，读取并处理excel数据，得到一阶导数谱图。可以使用以下代码读取数据并绘制谱图：

# 读取excel数据
data1 = pd.read_excel('10mg_L.xlsx')
data2 = pd.read_excel('15mg_L.xlsx')
data3 = pd.read_excel('20mg_L.xlsx')

# 获取波长和透过率数据
wavelength = data1.iloc[:, 0].values
tr1 = data1.iloc[:, 1].values
tr2 = data2.iloc[:, 1].values
tr3 = data3.iloc[:, 1].values

# 对透过率数据进行预处理，得到一阶导数谱图
def get_derivative_spectrum(tr):
    # 对透过率数据进行Savitzky-Golay滤波
    tr_smoothed = savgol_filter(tr, 31, 3)
    # 计算一阶导数
    dy = np.diff(tr_smoothed)
    dx = np.diff(wavelength)
    dy_dx = dy / dx
    # 对一阶导数进行Savitzky-Golay滤波
    dy_dx_smoothed = savgol_filter(dy_dx, 31, 3)
    return dy_dx_smoothed

ds1 = get_derivative_spectrum(tr1)
ds2 = get_derivative_spectrum(tr2)
ds3 = get_derivative_spectrum(tr3)

# 绘制一阶导数谱图
plt.plot(wavelength[1:], ds1, label='10mg/L')
plt.plot(wavelength[1:], ds2, label='15mg/L')
plt.plot(wavelength[1:], ds3, label='20mg/L')
plt.xlabel('Wavelength (nm)')
plt.ylabel('First Derivative')
plt.legend()
plt.show()

接下来，计算一阶导数谱的排列熵，并通过排列熵算法提取特征波长。可以使用以下代码实现：

# 计算一阶导数谱的排列熵
def get_permutation_entropy(ds):
    # 对一阶导数谱进行离散化
    ds_discrete = np.digitize(ds, np.histogram(ds, bins=20)[1])
    # 计算排列熵
    n = len(ds_discrete)
    pe = 0
    for m in range(2, 7):
        count = {}
        for i in range(n - m + 1):
            seg = tuple(ds_discrete[i:i+m])
            if seg in count:
                count[seg] += 1
            else:
                count[seg] = 1
        pe_m = 0
        for seg in count:
            p = count[seg] / (n - m + 1)
            pe_m -= p * np.log(p)
        pe += pe_m
    return pe

# 提取特征波长
pe1 = get_permutation_entropy(ds1)
pe2 = get_permutation_entropy(ds2)
pe3 = get_permutation_entropy(ds3)
f1 = wavelength[np.argmin(np.abs(ds1 - pe1))]
f2 = wavelength[np.argmin(np.abs(ds2 - pe2))]
f3 = wavelength[np.argmin(np.abs(ds3 - pe3))]

print('Feature wavelengths:', f1, f2, f3)

最后，利用特征波长处的特征值与对应的COD浓度进行PLS建模，得到COD的预测模型。可以使用以下代码实现：

# 读取COD浓度数据
cod1 = 10
cod2 = 15
cod3 = 20

# 构建PLS模型
X_train = np.array([ds1[np.argmin(np.abs(wavelength - f1))], ds2[np.argmin(np.abs(wavelength - f2))], ds3[np.argmin(np.abs(wavelength - f3))]]).reshape(3, 1)
y_train = np.array([cod1, cod2, cod3]).reshape(3, 1)
pls = PLSRegression(n_components=1)
pls.fit(X_train, y_train)

# 预测COD浓度
X_test = np.array([ds1[np.argmin(np.abs(wavelength - f1))], ds2[np.argmin(np.abs(wavelength - f2))], ds3[np.argmin(np.abs(wavelength - f3))]]).reshape(3, 1)
y_pred = pls.predict(X_test)

print('Predicted COD concentrations:', y_pred[:, 0])

这样就完成了基于紫外-可见光谱的水质COD检测代码的编写。完整代码如下：

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from scipy.signal import savgol_filter
from sklearn.cross_decomposition import PLSRegression

# 读取excel数据
data1 = pd.read_excel('10mg_L.xlsx')
data2 = pd.read_excel('15mg_L.xlsx')
data3 = pd.read_excel('20mg_L.xlsx')

# 获取波长和透过率数据
wavelength = data1.iloc[:, 0].values
tr1 = data1.iloc[:, 1].values
tr2 = data2.iloc[:, 1].values
tr3 = data3.iloc[:, 1].values

# 对透过率数据进行预处理，得到一阶导数谱图
def get_derivative_spectrum(tr):
    # 对透过率数据进行Savitzky-Golay滤波
    tr_smoothed = savgol_filter(tr, 31, 3)
    # 计算一阶导数
    dy = np.diff(tr_smoothed)
    dx = np.diff(wavelength)
    dy_dx = dy / dx
    # 对一阶导数进行Savitzky-Golay滤波
    dy_dx_smoothed = savgol_filter(dy_dx, 31, 3)
    return dy_dx_smoothed

ds1 = get_derivative_spectrum(tr1)
ds2 = get_derivative_spectrum(tr2)
ds3 = get_derivative_spectrum(tr3)

# 绘制一阶导数谱图
plt.plot(wavelength[1:], ds1, label='10mg/L')
plt.plot(wavelength[1:], ds2, label='15mg/L')
plt.plot(wavelength[1:], ds3, label='20mg/L')
plt.xlabel('Wavelength (nm)')
plt.ylabel('First Derivative')
plt.legend()
plt.show()

# 计算一阶导数谱的排列熵
def get_permutation_entropy(ds):
    # 对一阶导数谱进行离散化
    ds_discrete = np.digitize(ds, np.histogram(ds, bins=20)[1])
    # 计算排列熵
    n = len(ds_discrete)
    pe = 0
    for m in range(2, 7):
        count = {}
        for i in range(n - m + 1):
            seg = tuple(ds_discrete[i:i+m])
            if seg in count:
                count[seg] += 1
            else:
                count[seg] = 1
        pe_m = 0
        for seg in count:
            p = count[seg] / (n - m + 1)
            pe_m -= p * np.log(p)
        pe += pe_m
    return pe

# 提取特征波长
pe1 = get_permutation_entropy(ds1)
pe2 = get_permutation_entropy(ds2)
pe3 = get_permutation_entropy(ds3)
f1 = wavelength[np.argmin(np.abs(ds1 - pe1))]
f2 = wavelength[np.argmin(np.abs(ds2 - pe2))]
f3 = wavelength[np.argmin(np.abs(ds3 - pe3))]

print('Feature wavelengths:', f1, f2, f3)

# 读取COD浓度数据
cod1 = 10
cod2 = 15
cod3 = 20

# 构建PLS模型
X_train = np.array([ds1[np.argmin(np.abs(wavelength - f1))], ds2[np.argmin(np.abs(wavelength - f2))], ds3[np.argmin(np.abs(wavelength - f3))]]).reshape(3, 1)
y_train = np.array([cod1, cod2, cod3]).reshape(3, 1)
pls = PLSRegression(n_components=1)
pls.fit(X_train, y_train)

# 预测COD浓度
X_test = np.array([ds1[np.argmin(np.abs(wavelength - f1))], ds2[np.argmin(np.abs(wavelength - f2))], ds3[np.argmin(np.abs(wavelength - f3))]]).reshape(3, 1)
y_pred = pls.predict(X_test)

print('Predicted COD concentrations:', y_pred[:, 0])