变压器涌流和内部故障仿真-Simulink.zip

# -*- coding: utf-8 -*- """ Created on Wed Jan 23 09:11:55 2019 @author: baishuhua """ import warnings warnings.filterwarnings("ignore") import numpy as np import matplotlib.pyplot as plt plt.rcParams['font.sans-serif'] = 'SimHei' plt.rcParams['axes.unicode_minus']=False import os import glob from scipy import interpolate import DwtFeatureExtraction import pandas as pd from sklearn import preprocessing from sklearn.model_selection import train_test_split,cross_val_score,GridSearchCV import time import sys;sys.path.append('E:\\大数据') import sampler from collections import Counter ''' # ********** 第一阶段：原始数据展示及预处理 ********** ''' # 读文本数据,返回Time，Signals，其中Signals按列依次代表一次侧三相电流和二次侧三相电流 def ReadData_Text(file): Time = [] Signals = [] with open(file) as f: for line in f: Attr = line.strip().split(' ') Content = list(map(float, Attr)) Time.append(Content[0]) Signals.append(Content[1:]) return np.array(Time), np.array(Signals) def WaveForm_Cut(t, y, before=0.04, after=0.2): y_abs = np.abs(y) x_peak = np.argmax(y_abs, axis=0) left,right = min(x_peak[:3]),max(x_peak[:3]) interval = np.logical_and(t>=(t[left]-before), t<=(t[right]+after)) return t[interval], y[interval,:] # 展示原始波形数据 def ViewWaveform_TimeDomain(t, y, issave=True, name=''): fig,axes = plt.subplots(figsize=(10,6), nrows=2, ncols=2, sharex=True) axes[0,0].plot(t, y[:,:3]);axes[0,0].legend(['Ia','Ib','Ic']);axes[0,0].set_ylabel('一次侧') axes[0,1].plot(t, y[:,:3]-y[:,[1,2,0]]);axes[0,1].legend(['Iab','Ibc','Ica']) axes[1,0].plot(t, y[:,3:]);axes[1,0].legend(['Ia','Ib','Ic']);axes[1,0].set_ylabel('二次侧') axes[1,1].plot(t, y[:,3:]-y[:,[-1,-2,-3]]);axes[1,1].legend(['Iab','Ibc','Ica']) # plt.xlim([0.0,0.6]) plt.suptitle(os.path.basename(name)) if issave: plt.savefig(name) plt.close() # 数据预处理，插值成固定采样率的时间序列 def Interp(x, y, delta=1/1000, kind='slinear', axis=0): # delta为相邻自变量的间隔 # 沿axis轴拟合x，y，因变量y沿axis维度应等于x维度 function=interpolate.interp1d(x, y, kind, axis) x_new = np.arange(min(x), max(x)-delta, step=delta) y_new = function(x_new) return x_new, y_new # 现场录波读取，注意数据书写格式，前三行电压电流通道信息，后变采样率信息 def ReadFaultRecord(file): # 读取录波数据 ReadFlag=open(file) ReadFlag.readline();ReadFlag.readline();ReadFlag.readline() segnum=ReadFlag.readline().strip().split(':')[1] Time=[] for seg in range(int(segnum)): SampleAttr=ReadFlag.readline().strip().split(' ') [fs,start,terminal]=list(map(int,map(float,SampleAttr))) if len(Time)<1: Time.extend(1/fs*(np.linspace(1,terminal-start+1,num=terminal-start+1))) else: Time.extend(Time[-1]+1.0/fs*(np.linspace(1,terminal-start+1,num=terminal-start+1))) ReadFlag.readline();ReadFlag.readline() # 跳过故障相信息 SignalNames=ReadFlag.readline().strip().split(' ') SignalNames=[name for name in SignalNames if name!=''] Data=[];ValidIndex=[];row=0 for record in ReadFlag: detail=record.strip().split(' ') if len(detail)==len(SignalNames): try: Data.append(list(map(float,detail))) ValidIndex.append(row) row=row+1 except ValueError: row=row+1 else: row=row+1 continue ReadFlag.close() ValidIndex=np.array(ValidIndex) rows=ValidIndex[ValidIndex<len(Time)] Time=np.array(Time);Time=Time[rows];Data=np.array(Data);Data=Data[rows]; return Time,Data # 动模试验仿真，注意数据书写格式，先变采样率信息，后所有模拟通道 def ReadSimuRecord(file): # 读取仿真数据 ReadFlag=open(file) segnum=ReadFlag.readline().strip().split(':')[1] Time=[] for seg in range(int(segnum)): SampleAttr=ReadFlag.readline().strip().split(' ') [fs,start,terminal]=list(map(int,map(float,SampleAttr))) if len(Time)<1: Time.extend(1/fs*(np.linspace(1,terminal-start+1,num=terminal-start+1))) else: Time.extend(Time[-1]+1.0/fs*(np.linspace(1,terminal-start+1,num=terminal-start+1))) SignalNames=ReadFlag.readline().strip().split(' ') SignalNames=[name for name in SignalNames if name!=''] SignalNames=SignalNames[:41] Data=[];ValidIndex=[];row=0 for record in ReadFlag: detail=record.strip().split(' ') detail=[value for value in detail if value!=''] if len(detail)==len(SignalNames): try: Data.append(list(map(float,detail))) ValidIndex.append(row) row=row+1 except ValueError: row=row+1 else: row=row+1 continue ReadFlag.close() ValidIndex=np.array(ValidIndex) rows=ValidIndex[ValidIndex<len(Time)] Time=np.array(Time);Time=Time[rows];Data=np.array(Data);Data=Data[rows]; return Time,Data ''' # ********** 第二阶段：特征工程，生成特征集和类别标签 ********** ''' # 特征提取 def ExtractFeature(file): time, signals = ReadData_Text(file) time_new, signals_new = Interp(time, signals) signals_input = signals_new - signals_new[:,[1,2,0,4,5,3]] # 差流信号 names_input = ['Iab1','Ibc1','Ica1','Iab2','Ibc2','Ica2'] # 1,2代表一，二次侧 featurename = [];featurevalue = [] for col in range(np.shape(signals_new)[-1]): sample = signals_input[:,col] # ********** DWT小波变换域特征提取 ********** # wavelet = 'bior3.5';level = 4 coeffs = DwtFeatureExtraction.DwtRec(sample, wavelet=wavelet, level=level) name,value = DwtFeatureExtraction.FeatureExtractDwt(coeffs) # 每个特征名添加对待处理物理量的描述 name_new = list(map(lambda x:names_input[col]+'-'+x,name)) featurename.extend(name_new) featurevalue.extend(value) featurename=np.array(featurename);featurevalue=np.array(featurevalue) return featurename, featurevalue # 特征样本集生成，并将其保存至文件中 def ExtractFeatures_AllFiles(FilePath, SavePath, filename='Features.csv'): print('Start extracting features ...\n') BigFeatures = [];BigSamplenames = [];BigLabels = [] SaveFile = os.path.join(SavePath, filename) import csv file_object = open(SaveFile, 'w', newline='') writer=csv.writer(file_object) FileLists = glob.glob(FilePath, recursive=True) FileNums=len(FileLists) for fileNo,file in enumerate(FileLists, start=1): try: featurename, featurevalue=ExtractFeature(file) filename=os.path.basename(file) attr=filename.split('_')[0] BigFeatures.append(featurevalue) # 特征矩阵 BigSamplenames.append(file) # 样本名列表 BigLabels.append(attr) # 标签列表 except: continue if fileNo == 1: head = np.concatenate((np.array(['Filename','Label']), featurename)) writer.writerow(head) line=np.concatenate((np.array([filename,attr]), featurevalue)) writer.writerow(line) if fileNo%1000 == 0 & fileNo != FileNums: print('Completed {0} %'.format(round(fileNo/FileNums*100,4))) elif fileNo == FileNums: print('Finish extracting features!!!') BigFeatures = np.array(BigFeatures) BigSamplenames = np.array(BigSamplenames) BigLabels = np.array(BigLabels) file_object.close() return BigSamplenames,