CN114997749B - Method and system for intelligent dispatching of electric power personnel - Google Patents

Method and system for intelligent dispatching of electric power personnel Download PDF

Info

Publication number
CN114997749B
CN114997749B CN202210919919.5A CN202210919919A CN114997749B CN 114997749 B CN114997749 B CN 114997749B CN 202210919919 A CN202210919919 A CN 202210919919A CN 114997749 B CN114997749 B CN 114997749B
Authority
CN
China
Prior art keywords
audio signal
fault
power personnel
electric power
personnel
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202210919919.5A
Other languages
Chinese (zh)
Other versions
CN114997749A (en
Inventor
杨健
陈春玲
周焱
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing University of Posts and Telecommunications
Original Assignee
Nanjing University of Posts and Telecommunications
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nanjing University of Posts and Telecommunications filed Critical Nanjing University of Posts and Telecommunications
Priority to CN202210919919.5A priority Critical patent/CN114997749B/en
Publication of CN114997749A publication Critical patent/CN114997749A/en
Application granted granted Critical
Publication of CN114997749B publication Critical patent/CN114997749B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
    • G06Q10/06311—Scheduling, planning or task assignment for a person or group
    • G06Q10/063112—Skill-based matching of a person or a group to a task
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00—Speech recognition
    • G10L15/06—Creation of reference templates; Training of speech recognition systems, e.g. adaptation to the characteristics of the speaker's voice
    • G10L15/063—Training
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
    • G10L25/24—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being the cepstrum
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/27—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique
    • G10L25/30—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique using neural networks
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Economics (AREA)
  • Strategic Management (AREA)
  • Computational Linguistics (AREA)
  • Multimedia (AREA)
  • Human Computer Interaction (AREA)
  • Acoustics & Sound (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Educational Administration (AREA)
  • Signal Processing (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Theoretical Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Marketing (AREA)
  • General Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • Tourism & Hospitality (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Game Theory and Decision Science (AREA)
  • Development Economics (AREA)
  • Evolutionary Computation (AREA)
  • Public Health (AREA)
  • Water Supply & Treatment (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Electric Propulsion And Braking For Vehicles (AREA)

Abstract

本发明公开了电力人员智能调度方法和系统，方法包括：获取机器工作时的音频信号样本，根据音频信号样本确定机器是否存在故障；若判断机器存在故障，则根据音频信号样本确定机器故障类型；根据可调度的电力人员人数、电力人员故障修复次数、电力人员修复成功率以及机器故障类型构建智能调度问题；求解所述智能调度问题获得智能调度方案。本发明基于机器故障类型，能够第一时间对电力人员进行精准调度，从而完成机器故障的维修。

Figure 202210919919

The invention discloses a method and a system for intelligent dispatching of electric power personnel. The method includes: acquiring audio signal samples when a machine is working, and determining whether the machine is faulty according to the audio signal samples; if it is judged that the machine is faulty, determining the machine failure type according to the audio signal samples; An intelligent dispatching problem is constructed according to the number of dispatchable electric power personnel, the number of electric personnel fault repairs, the success rate of electric personnel repairing, and the type of machine failure; an intelligent dispatching scheme is obtained by solving the intelligent dispatching problem. Based on the type of machine failure, the present invention can precisely schedule electric power personnel at the first time, so as to complete the maintenance of the machine failure.

Figure 202210919919

Description

电力人员智能调度方法和系统Method and system for intelligent dispatching of electric power personnel

技术领域technical field

本发明涉及电力人员智能调度方法和系统，属于电力物联网技术领域以及电力系统智能调度技术领域。The invention relates to an intelligent dispatching method and system for electric power personnel, and belongs to the technical field of electric power Internet of Things and electric power system intelligent dispatching technology.

背景技术Background technique

电力系统作为关系到国家安全和国民经济命脉的重要基础设施，不仅要满足正常环境下系统稳定可靠地运行，更需要在系统出现某些问题的时候，能够及时准确地找出系统问题所在，并对电力人员进行精准调度，以最小化经济损失，从而更好地保证居民的用电体验。因此，如何在系统出现问题后快速找出问题所在，并对电力人员进行精准调度，减少停电造成的经济损失成为一项亟需解决的关键问题。As an important infrastructure related to national security and the lifeline of the national economy, the power system not only needs to meet the requirements of stable and reliable operation of the system under normal circumstances, but also needs to be able to find out the problem in time and accurately when there are some problems in the system, and Precise dispatch of electric power personnel to minimize economic losses, so as to better ensure residents' experience in using electricity. Therefore, how to quickly find out the problem after a system problem occurs, and accurately dispatch power personnel to reduce economic losses caused by power outages has become a key problem that needs to be solved urgently.

然而现存的一些电力人员智能调度方法大部分都存在着一些缺点，比如当某一台机器出现故障的时候，只能判别出电力系统出了问题，但是不能精确地反馈某一机器故障类型。基于这种情况，对电力人员的调度也会存在一些问题，比如调度了某一位电力人员，但他对发生的机器故障没有维修经验，那么只能调度其他电力人员来进行维修，这样的话就会浪费大量的时间，在此期间，可能会伴随着极大的经济损失。However, most of the existing intelligent dispatching methods for power personnel have some shortcomings. For example, when a certain machine fails, it can only identify the power system as a problem, but cannot accurately feed back the type of failure of a certain machine. Based on this situation, there will also be some problems in the dispatch of electric power personnel. For example, if a certain electric power personnel is dispatched, but he has no maintenance experience for the machine failure, then he can only dispatch other electric personnel to carry out maintenance. A lot of time will be wasted, and during this period, it may be accompanied by great economic losses.

发明内容Contents of the invention

本发明针对现有技术中存在的问题与不足，提供一种电力人员智能调度方法，结合机器故障预测故障分类结果，以及调度人员维修经验实现对电力人员进行精准调度。Aiming at the problems and deficiencies in the prior art, the present invention provides an intelligent dispatching method for electric power personnel, which realizes precise dispatching of electric power personnel in combination with machine failure prediction fault classification results and dispatcher maintenance experience.

为实现上述技术目的，本发明采用以下方案。In order to achieve the above-mentioned technical purpose, the present invention adopts the following solutions.

一方面，本发明提供电力人员智能调度方法，包括以下步骤：On the one hand, the present invention provides an intelligent dispatching method for electric power personnel, comprising the following steps:

获取机器工作时的音频信号样本，根据音频信号样本确定机器是否存在故障；Obtain audio signal samples when the machine is working, and determine whether the machine is faulty according to the audio signal samples;

若判断机器存在故障，则根据音频信号样本确定机器故障类型；If it is judged that there is a fault in the machine, then determine the type of machine fault according to the audio signal sample;

根据可调度的电力人员人数、电力人员故障修复成功次数、修复成功率以及机器故障类型构建智能调度问题；According to the number of power personnel who can be dispatched, the number of power personnel failure repairs, the success rate of repairs, and the type of machine failure, an intelligent scheduling problem is constructed;

求解所述智能调度问题获得智能调度方案。An intelligent scheduling solution is obtained by solving the intelligent scheduling problem.

进一步地，构建智能调度问题，包括：确定可调度的电力人员人数，将可调度的电力人员人数和存在故障的音频信号样本的数量进行比较；Further, constructing an intelligent scheduling problem includes: determining the number of dispatchable electric personnel, and comparing the dispatchable number of electric personnel with the number of faulty audio signal samples;

若可调度的电力人员人数小于等于存在故障的音频信号样本的数量，则求解第一电力人员智能调度问题；If the number of power personnel that can be dispatched is less than or equal to the number of faulty audio signal samples, then solve the intelligent dispatching problem of the first electric power personnel;

假设可调度的电力人员人数记为L，电力人员的集合表示为

Figure 96967DEST_PATH_IMAGE001
；根据电力人员的历史维修记录，建立每个电力人员的维修档案，电力人员l的维修档案=｛机器故障类型k，电力人员l成功修复故障类型k的次数
Figure 996921DEST_PATH_IMAGE002
，电力人员l修复故障类型k的成功率
Figure 939469DEST_PATH_IMAGE003
，
Figure 908562DEST_PATH_IMAGE004
｝，K代表已知故障类型数目，第K+1种故障类型代表其它未知类型故障；Assuming that the number of dispatchable electric personnel is denoted as L , the set of electric personnel is expressed as
Figure 96967DEST_PATH_IMAGE001
; According to the historical maintenance records of the electric personnel, establish the maintenance file of each electric personnel, the maintenance file of the electric personnel l ={machine fault type k , the number of times that the electric personnel l successfully repairs the fault type k
Figure 996921DEST_PATH_IMAGE002
, the success rate of power personnel l repairing fault type k
Figure 939469DEST_PATH_IMAGE003
,
Figure 908562DEST_PATH_IMAGE004
}, K represents the number of known fault types, and the K+ 1th fault type represents other unknown types of faults;

假设当前时刻共有M个存在故障的音频信号样本，每个样本

Figure 125917DEST_PATH_IMAGE005
,m=1,2,…,M的故障分类结果表示为
Figure 611869DEST_PATH_IMAGE006
，其中
Figure 256477DEST_PATH_IMAGE007
表示第m个音频信号样本中的机器故障类型为K+1个类型中故障类型k的概率；Assume that there are M faulty audio signal samples at the current moment, and each sample
Figure 125917DEST_PATH_IMAGE005
, m =1,2,…, the fault classification result of M is expressed as
Figure 611869DEST_PATH_IMAGE006
,in
Figure 256477DEST_PATH_IMAGE007
Indicates that the machine failure type in the m audio signal sample is the probability of failure type k in K+ 1 types;

计算每个存在故障的音频信号样本的重要性

Figure 712866DEST_PATH_IMAGE008
，其中
Figure 202753DEST_PATH_IMAGE009
代表故障类型k的优先级；Calculate the importance of each faulty audio signal sample
Figure 712866DEST_PATH_IMAGE008
,in
Figure 202753DEST_PATH_IMAGE009
Represents the priority of fault type k ;

根据重要性

Figure 811720DEST_PATH_IMAGE010
大小，从大到小选出前L个存在故障的音频信号样本，并将这些选出的样本索引放在故障音频信号样本集
Figure 361650DEST_PATH_IMAGE011
中；according to importance
Figure 811720DEST_PATH_IMAGE010
Size, select the first L faulty audio signal samples from large to small, and put these selected sample indexes in the faulty audio signal sample set
Figure 361650DEST_PATH_IMAGE011
middle;

所述第一电力人员智能调度问题表示如下：The intelligent scheduling problem of the first electric power personnel is expressed as follows:

Figure 39756DEST_PATH_IMAGE012
；
Figure 39756DEST_PATH_IMAGE012
;

其中，

Figure 598913DEST_PATH_IMAGE013
表示电力人员
Figure 311655DEST_PATH_IMAGE014
的调度策略，电力人员l的调度策略表示为：如果电力人员l被调度维修第m个音频信号样本的机器故障，那么
Figure 783218DEST_PATH_IMAGE015
，否则
Figure 214200DEST_PATH_IMAGE016
；
Figure 311469DEST_PATH_IMAGE017
为电力人员
Figure 878716DEST_PATH_IMAGE018
成功修复故障类型k的次数，
Figure 770449DEST_PATH_IMAGE019
为电力人员
Figure 173879DEST_PATH_IMAGE018
修复故障类型k的成功率，
Figure 74839DEST_PATH_IMAGE020
表示电力人员
Figure 496593DEST_PATH_IMAGE018
的调度策略，电力人员
Figure 824807DEST_PATH_IMAGE018
的调度策略表示为：如果电力人员
Figure 712604DEST_PATH_IMAGE018
被调度维修第m个音频信号样本中的机器故障，那么
Figure 151675DEST_PATH_IMAGE021
，否则
Figure 427936DEST_PATH_IMAGE022
。in,
Figure 598913DEST_PATH_IMAGE013
means electric staff
Figure 311655DEST_PATH_IMAGE014
The dispatching strategy of electric power personnel l is expressed as: if electric power personnel l is dispatched to repair the machine failure of the m -th audio signal sample, then
Figure 783218DEST_PATH_IMAGE015
,otherwise
Figure 214200DEST_PATH_IMAGE016
;
Figure 311469DEST_PATH_IMAGE017
for electricians
Figure 878716DEST_PATH_IMAGE018
the number of successful repairs of fault type k ,
Figure 770449DEST_PATH_IMAGE019
for electricians
Figure 173879DEST_PATH_IMAGE018
the success rate of repairing fault type k ,
Figure 74839DEST_PATH_IMAGE020
means electric staff
Figure 496593DEST_PATH_IMAGE018
Scheduling strategy, power staff
Figure 824807DEST_PATH_IMAGE018
The scheduling strategy of is expressed as: If the power personnel
Figure 712604DEST_PATH_IMAGE018
is scheduled to repair the machine failure in the mth audio signal sample, then
Figure 151675DEST_PATH_IMAGE021
,otherwise
Figure 427936DEST_PATH_IMAGE022
.

进一步地，构建智能调度问题，包括：确定可调度的电力人员人数，将可调度的电力人员人数和确定的存在故障的音频信号样本的数量进行比较；若可调度的电力人员人数大于存在故障的音频信号样本的数量，则求解第二电力人员智能调度问题，最终获得电力人员智能调度方案；Further, constructing an intelligent scheduling problem includes: determining the number of dispatchable electric personnel, comparing the dispatchable number of electric personnel with the determined number of faulty audio signal samples; if the dispatchable number of electric personnel is greater than the faulty The number of audio signal samples is used to solve the second electric power personnel intelligent dispatching problem, and finally obtain the electric power personnel intelligent dispatching scheme;

所述第二电力人员智能调度问题表示如下：The intelligent scheduling problem of the second electric power personnel is expressed as follows:

Figure 927050DEST_PATH_IMAGE023
；
Figure 927050DEST_PATH_IMAGE023
;

其中，

Figure 288761DEST_PATH_IMAGE024
表示电力人员
Figure 282256DEST_PATH_IMAGE014
的调度策略，电力人员l的调度策略表示为：如果电力人员l被调度维修第m个音频信号样本中的机器故障，那么
Figure 678603DEST_PATH_IMAGE025
，否则
Figure 348618DEST_PATH_IMAGE026
；L为可调度的电力人员人数，
Figure 932046DEST_PATH_IMAGE027
为电力人员l成功修复故障类型k的次数，
Figure 978500DEST_PATH_IMAGE028
为电力人员l修复故障类型k的成功率，
Figure 229353DEST_PATH_IMAGE029
，K代表已知故障类型数目，第K+1种故障类型代表其它未知类型故障，
Figure 821002DEST_PATH_IMAGE030
为电力人员
Figure 891726DEST_PATH_IMAGE031
的成功修复故障类型k的次数，
Figure 476291DEST_PATH_IMAGE032
为电力人员
Figure 581651DEST_PATH_IMAGE033
修复故障类型k的成功率，
Figure 593469DEST_PATH_IMAGE034
表示电力人员
Figure 433380DEST_PATH_IMAGE031
的调度策略，电力人员
Figure 290478DEST_PATH_IMAGE035
的调度策略表示为：如果电力人员
Figure 515923DEST_PATH_IMAGE036
被调度维修第m个音频信号样本中的机器故障，那么
Figure 698642DEST_PATH_IMAGE037
，否则
Figure 743959DEST_PATH_IMAGE038
，
Figure 418129DEST_PATH_IMAGE039
表示第m个音频信号样本中的机器故障类型为K+1个类型中的某一种故障类型的概率，M为存在故障的音频信号样本总数。in,
Figure 288761DEST_PATH_IMAGE024
means electric staff
Figure 282256DEST_PATH_IMAGE014
The dispatching strategy of electric power personnel l is expressed as: if electric power personnel l is dispatched to repair the machine failure in the mth audio signal sample, then
Figure 678603DEST_PATH_IMAGE025
,otherwise
Figure 348618DEST_PATH_IMAGE026
; L is the number of electric power personnel that can be dispatched,
Figure 932046DEST_PATH_IMAGE027
is the number of times that the electrician l successfully repairs the fault type k ,
Figure 978500DEST_PATH_IMAGE028
is the success rate of repairing fault type k for electrician l ,
Figure 229353DEST_PATH_IMAGE029
, K represents the number of known fault types, the K+ 1th fault type represents other unknown types of faults,
Figure 821002DEST_PATH_IMAGE030
for electricians
Figure 891726DEST_PATH_IMAGE031
The number of times of successfully repairing the fault type k ,
Figure 476291DEST_PATH_IMAGE032
for electricians
Figure 581651DEST_PATH_IMAGE033
the success rate of repairing fault type k ,
Figure 593469DEST_PATH_IMAGE034
means electric staff
Figure 433380DEST_PATH_IMAGE031
Scheduling strategy, power staff
Figure 290478DEST_PATH_IMAGE035
The scheduling strategy of is expressed as: If the power personnel
Figure 515923DEST_PATH_IMAGE036
is scheduled to repair the machine failure in the mth audio signal sample, then
Figure 698642DEST_PATH_IMAGE037
,otherwise
Figure 743959DEST_PATH_IMAGE038
,
Figure 418129DEST_PATH_IMAGE039
Indicates the probability that the machine fault type in the mth audio signal sample is one of the K +1 types, and M is the total number of audio signal samples with faults.

进一步地，根据音频信号样本确定机器是否存在故障，包括：Further, determining whether the machine is faulty according to the audio signal samples includes:

提取机器工作时的音频信号样本的梅尔频率倒谱系数特征向量，通过编码器和解码器对梅尔频率倒谱系数特征向量进行重建；Extracting the Mel-frequency cepstral coefficient eigenvector of the audio signal sample when the machine is working, and reconstructing the Mel-frequency cepstral coefficient eigenvector through an encoder and a decoder;

将重建结果与异常分数阈值进行比较确定机器是否存在故障。The rebuild results are compared to an anomaly score threshold to determine if the machine is faulty.

再进一步地，提取机器工作时的音频信号样本的梅尔频率倒谱系数特征向量，包括：Further, the Mel frequency cepstral coefficient eigenvector of the audio signal sample when the machine is extracted includes:

将音频信号样本频率转化为梅尔频率；Convert audio signal sample frequency to Mel frequency;

将音频信号样本通过高通滤波器预加重，使音频信号样本保持在低频到高频的整个频带中，高通滤波器如：The audio signal sample is pre-emphasized through a high-pass filter to keep the audio signal sample in the entire frequency band from low frequency to high frequency. The high-pass filter is as follows:

Figure 763660DEST_PATH_IMAGE040
；
Figure 763660DEST_PATH_IMAGE040
;

其中，a为预加重系数；

Figure 851702DEST_PATH_IMAGE041
表示高通滤波器的系统函数；Among them, a is the pre-emphasis coefficient;
Figure 851702DEST_PATH_IMAGE041
Represents the system function of the high-pass filter;

对每段单帧信号进行加窗操作，将预加重后的音频信号样本分成多段单帧信号，每段单帧信号对应于一个频谱；Perform a windowing operation on each single-frame signal, divide the pre-emphasized audio signal samples into multiple single-frame signals, and each single-frame signal corresponds to a frequency spectrum;

加窗操作如下：The windowing operation is as follows:

Figure 649893DEST_PATH_IMAGE042
；
Figure 649893DEST_PATH_IMAGE042
;

其中，

Figure 114373DEST_PATH_IMAGE043
为加窗函数，N为帧长，n为帧数。 in,
Figure 114373DEST_PATH_IMAGE043
is the windowing function, N is the frame length, and n is the number of frames.

利用短时快速傅里叶变换，计算频率与振幅的关系；短时快速傅里叶变换公式为：Use the short-time fast Fourier transform to calculate the relationship between frequency and amplitude; the short-time fast Fourier transform formula is:

Figure 65142DEST_PATH_IMAGE044
；
Figure 65142DEST_PATH_IMAGE044
;

其中，x(n)为第n帧的信号幅值，E(n)是变换后的能量信号；Among them, x ( n ) is the signal amplitude of the nth frame, E ( n ) is the transformed energy signal;

对音频信号样本进行离散余弦变换得到梅尔频率倒谱系数，公式如下：The discrete cosine transform is performed on the audio signal samples to obtain the Mel frequency cepstral coefficients, the formula is as follows:

Figure 324085DEST_PATH_IMAGE045
；
Figure 324085DEST_PATH_IMAGE045
;

Figure 343994DEST_PATH_IMAGE046
；
Figure 343994DEST_PATH_IMAGE046
;

其中，f(n) 表示音频信号样本第n帧在时域上的信号，F(n) 是余弦变换后的系数，C(n) 为音频信号样本第n帧的补偿系数；Among them, f ( n ) represents the signal of the nth frame of the audio signal sample in the time domain, F ( n ) is the coefficient after cosine transformation, and C ( n ) is the compensation coefficient of the nth frame of the audio signal sample;

最终得到描述音频数据的低阶梅尔频率倒谱系数特征向量MFCC，其数据形式为：Finally, the low-order Mel frequency cepstral coefficient feature vector MFCC describing the audio data is obtained, and its data form is:

Figure 408902DEST_PATH_IMAGE047
；
Figure 408902DEST_PATH_IMAGE047
;

其中F11表示第1个音频信号样本的第1帧的梅尔频率倒谱系数；F1N 表示第1个音频信号样本的第N帧的梅尔频率倒谱系数；F M1表示第M个音频信号样本的第1帧的梅尔频率倒谱系数，F MN 表示第M个音频信号样本的第N帧的梅尔频率倒谱系数。Among them, F 11 represents the Mel frequency cepstral coefficient of the first frame of the first audio signal sample; F 1 N represents the Mel frequency cepstral coefficient of the Nth frame of the first audio signal sample; F M 1 represents the Mth Mel frequency cepstral coefficient of the first frame of the audio signal sample, F MN represents the Mel frequency cepstral coefficient of the Nth frame of the M audio signal sample.

进一步地，通过编码器和解码器对梅尔频率倒谱系数特征向量进行重建，包括：Further, the eigenvectors of the Mel-frequency cepstral coefficients are reconstructed through the encoder and the decoder, including:

编码器E输入特征向量

Figure 197866DEST_PATH_IMAGE048
，将获取到的音频信号样本的梅尔频率倒谱系数特征向 量转换为潜在特征向量
Figure 644022DEST_PATH_IMAGE049
，表示为： Encoder E input feature vector
Figure 197866DEST_PATH_IMAGE048
, convert the eigenvectors of Mel-frequency cepstral coefficients of acquired audio signal samples into latent eigenvectors
Figure 644022DEST_PATH_IMAGE049
,Expressed as:

Figure 416806DEST_PATH_IMAGE050
；
Figure 416806DEST_PATH_IMAGE050
;

其中

Figure 223088DEST_PATH_IMAGE051
是编码器的参数； in
Figure 223088DEST_PATH_IMAGE051
is the parameter of the encoder;

解码器D将潜在特征向量

Figure 132138DEST_PATH_IMAGE052
重建成输入特征向量
Figure 998463DEST_PATH_IMAGE053
，表示为： Decoder D converts the latent feature vector
Figure 132138DEST_PATH_IMAGE052
Reconstruct into input feature vector
Figure 998463DEST_PATH_IMAGE053
,Expressed as:

Figure 743697DEST_PATH_IMAGE054
；
Figure 743697DEST_PATH_IMAGE054
;

其中

Figure 353669DEST_PATH_IMAGE055
是解码器的参数； in
Figure 353669DEST_PATH_IMAGE055
is the parameter of the decoder;

上述两个式子中：

Figure 117226DEST_PATH_IMAGE056
，下标i表示当前是第i个特征向量；s为特征向量总 数。 In the above two formulas:
Figure 117226DEST_PATH_IMAGE056
, the subscript i indicates that it is the i -th eigenvector; s is the total number of eigenvectors.

再进一步地，将重建结果与异常分数阈值进行比较确定机器是否存在故障，具体步骤包括：调整编码器和解码器的参数，使重建数据和输入数据之间的误差最小化，重建误差函数为：Furthermore, the reconstruction result is compared with the abnormal score threshold to determine whether the machine is faulty. The specific steps include: adjusting the parameters of the encoder and decoder to minimize the error between the reconstruction data and the input data. The reconstruction error function is:

Figure 420031DEST_PATH_IMAGE057
；
Figure 420031DEST_PATH_IMAGE057
;

Figure 649631DEST_PATH_IMAGE058
是误差函数；
Figure 797716DEST_PATH_IMAGE059
表示输入特征向量为
Figure 681358DEST_PATH_IMAGE060
的时候，所得出的重建误差值；
Figure 649631DEST_PATH_IMAGE058
is the error function;
Figure 797716DEST_PATH_IMAGE059
Indicates that the input feature vector is
Figure 681358DEST_PATH_IMAGE060
When , the resulting reconstruction error value;

将重建误差作为音频信号的异常分数S，确定机器是否存在故障的方式如下：Using the reconstruction error as the anomaly fraction S of the audio signal, the way to determine whether a machine is faulty is as follows:

Figure 889485DEST_PATH_IMAGE061
；
Figure 889485DEST_PATH_IMAGE061
;

Figure 858579DEST_PATH_IMAGE062
是输入特征向量为X的时候，所得出的重建误差值，
Figure 810354DEST_PATH_IMAGE063
为设定的异常分数阈值。
Figure 858579DEST_PATH_IMAGE062
is the reconstruction error value obtained when the input feature vector is X,
Figure 810354DEST_PATH_IMAGE063
Anomaly score threshold set for .

进一步地，若判断机器存在故障，则利用CNN模型进行分类获得机器故障类型，包括：Further, if it is judged that there is a fault in the machine, the CNN model is used to classify and obtain the type of machine fault, including:

将被判定为机器存在故障的音频信号样本的梅尔频率倒谱系数特征向量作为CNN模型的输入，通过softmax函数输出此音频信号为已知故障类型的概率结果；对于一个音频信号样本X，假设经过softmax层输出的机器故障类型结果为：The eigenvector of the Mel-frequency cepstral coefficient of the audio signal sample that is judged to be faulty in the machine is used as the input of the CNN model, and the probability result that the audio signal is a known fault type is output through the softmax function; for an audio signal sample X, suppose The result of the machine failure type output by the softmax layer is:

Figure 299235DEST_PATH_IMAGE064
；
Figure 299235DEST_PATH_IMAGE064
;

其中，

Figure 678264DEST_PATH_IMAGE065
为softmax单个节点输出的机器故障类型为k的概率，K代表已知故障类型数目，第K+1种故障类型代表其它未知类型故障，且满足
Figure 134653DEST_PATH_IMAGE066
；in,
Figure 678264DEST_PATH_IMAGE065
It is the probability that the machine fault type output by a single softmax node is k , K represents the number of known fault types, and the K+ 1th fault type represents other unknown types of faults, and satisfies
Figure 134653DEST_PATH_IMAGE066
;

输出的概率最大的机器故障类型就为最终确定的结果。The machine failure type with the highest output probability is the final result.

第二方面，本发明还提供了电力人员智能调度系统，包括：故障判断模块、故障类型判断模块以及调度方案求解模块；In the second aspect, the present invention also provides an intelligent dispatching system for electric power personnel, including: a fault judgment module, a fault type judgment module, and a dispatch scheme solution module;

故障判断模块，用于获取机器工作时的音频信号样本，根据音频信号样本确定机器是否存在故障；The fault judgment module is used to obtain audio signal samples when the machine is working, and determine whether the machine is faulty according to the audio signal samples;

故障类型判断模块，用于若判断机器存在故障，则根据音频信号样本确定机器故障类型；The fault type judging module is used to determine the machine fault type according to the audio signal sample if it is judged that the machine has a fault;

调度方案求解模块，用于根据可调度的电力人员人数、电力人员故障修复成功次数、修复成功率以及机器故障类型构建智能调度问题；求解所述智能调度问题获得智能调度方案。The dispatching plan solving module is used to construct an intelligent dispatching problem according to the number of dispatchable electric power personnel, the number of electric power personnel's fault repair successes, the repair success rate, and the type of machine fault; solving the intelligent dispatching problem to obtain an intelligent dispatching plan.

本发明提供的一种电力人员智能调度方法，具有如下优点：An intelligent scheduling method for electric power personnel provided by the present invention has the following advantages:

本发明基于机器的故障类型精准调度到对应维修经验丰富的电力人员，以最小的经济损失，快速高效地实现调度；针对人员调度，建立了不同情况下最大化机器故障平均修复率的电力人员调度问题，可以实现电力人员的智能调度，能够有效减少电力人员调度所消耗的时间；Based on the type of machine faults, the present invention accurately dispatches electric personnel with rich maintenance experience to realize dispatching quickly and efficiently with minimal economic loss; for personnel dispatching, a power personnel dispatching that maximizes the average repair rate of machine faults under different circumstances is established Problems, can realize the intelligent scheduling of electric personnel, and can effectively reduce the time consumed by electric personnel scheduling;

本方法采用自编码器对输入特征向量进行重建，其具有很强的特征表示能力，并且网络的结构相对简单、容易训练，从而能够提高系统的可实现性；This method uses an autoencoder to reconstruct the input feature vector, which has a strong feature representation ability, and the structure of the network is relatively simple and easy to train, thereby improving the feasibility of the system;

本方法采用的是无监督学习和监督学习相结合，有效解决了机器异常工作的音频信号的采集问题，大大降低了采集训练样本的难度。This method uses the combination of unsupervised learning and supervised learning, which effectively solves the problem of collecting audio signals of abnormal machine work, and greatly reduces the difficulty of collecting training samples.

附图说明Description of drawings

图1为本发明实施例提供的电力人员调度方法的流程图；Fig. 1 is the flow chart of the electric power personnel scheduling method provided by the embodiment of the present invention;

图2为实施例中确定机器是否存在故障的原理框架图；Fig. 2 is the principle frame diagram of determining whether the machine has a fault in the embodiment;

图3为本发明实施例中运用的CNN模型的结构。Fig. 3 is the structure of the CNN model used in the embodiment of the present invention.

具体实施方式Detailed ways

下面结合附图以及具体实施例，进一步阐明本发明，应理解这些实施例仅用于说明本发明而不用于限制本发明的范围，在阅读了本发明之后，本领域技术人员对本发明的各种等价形式的修改均落于本申请所附权利要求所限定的范围。Below in conjunction with accompanying drawing and specific embodiment, further illustrate the present invention, should be understood that these embodiments are only for illustrating the present invention and are not intended to limit the scope of the present invention, after having read the present invention, those skilled in the art will understand various aspects of the present invention Modifications in equivalent forms all fall within the scope defined by the appended claims of this application.

实施例1Example 1

电力人员智能调度方法，包括以下步骤：The intelligent scheduling method for electric power personnel includes the following steps:

获取机器工作时的音频信号样本，根据音频信号样本确定机器是否存在故障；Obtain audio signal samples when the machine is working, and determine whether the machine is faulty according to the audio signal samples;

若判断机器存在故障，则根据音频信号样本确定机器故障类型；If it is judged that there is a fault in the machine, then determine the type of machine fault according to the audio signal sample;

根据可调度的电力人员人数、电力人员故障修复成功次数、电力人员修复成功率以及机器故障类型构建智能调度问题；Construct an intelligent scheduling problem based on the number of dispatchable electric personnel, the number of successful fault repairs by electric personnel, the success rate of electric personnel repairs, and the type of machine failure;

求解所述智能调度问题获得智能调度方案。An intelligent scheduling solution is obtained by solving the intelligent scheduling problem.

本实施例中，构建智能调度问题，包括：确定可调度的电力人员人数，将可调度的电力人员人数和存在故障的音频信号样本的数量进行比较；In this embodiment, constructing an intelligent scheduling problem includes: determining the number of dispatchable electric personnel, and comparing the dispatchable number of electric personnel with the number of faulty audio signal samples;

若可调度的电力人员人数小于等于存在故障的音频信号样本的数量，则求解第一电力人员智能调度问题；If the number of power personnel that can be dispatched is less than or equal to the number of faulty audio signal samples, then solve the intelligent dispatching problem of the first electric power personnel;

在建立第一电力人员智能调度问题之前，假设可调度的电力人员人数记为L，其中电力人员的集合可以表示为

Figure 624540DEST_PATH_IMAGE067
。根据电力人员的历史维修记录，建立每个电力维护人员的维修档案，电力人员l的维修档案=｛机器故障类型k，电力人员l成功修复故障类型k的次数
Figure 482775DEST_PATH_IMAGE068
，电力人员l修复故障类型k的成功率
Figure 517858DEST_PATH_IMAGE069
，
Figure 461543DEST_PATH_IMAGE070
｝， K代表已知故障类型数目，第K+1种故障类型代表其它未知类型故障；Before setting up the first electric power personnel intelligent scheduling problem, it is assumed that the number of dispatchable electric personnel is denoted as L , where the set of electric personnel can be expressed as
Figure 624540DEST_PATH_IMAGE067
. According to the historical maintenance records of electric personnel, the maintenance files of each electric maintenance personnel are established, the maintenance files of electric personnel l = {machine fault type k , the number of times electric personnel l successfully repairs fault type k
Figure 482775DEST_PATH_IMAGE068
, the success rate of power personnel l repairing fault type k
Figure 517858DEST_PATH_IMAGE069
,
Figure 461543DEST_PATH_IMAGE070
}, K represents the number of known fault types, and the K+ 1th fault type represents other unknown types of faults;

假设当前时刻共有M个存在故障的音频信号样本，每个样本

Figure 755121DEST_PATH_IMAGE071
,m=1,2,…,M的故障分类结果表示为
Figure 467863DEST_PATH_IMAGE072
，其中
Figure 188694DEST_PATH_IMAGE073
表示第m个音频信号样本中的机器故障类型为K+1个类型中故障类型k的概率；Assume that there are M faulty audio signal samples at the current moment, and each sample
Figure 755121DEST_PATH_IMAGE071
, m =1,2,…, the fault classification result of M is expressed as
Figure 467863DEST_PATH_IMAGE072
,in
Figure 188694DEST_PATH_IMAGE073
Indicates that the machine failure type in the m audio signal sample is the probability of failure type k in K+ 1 types;

计算每个存在故障的音频信号样本的重要性

Figure 104828DEST_PATH_IMAGE074
，其中
Figure 936518DEST_PATH_IMAGE075
代表故障类型k的优先级；Calculate the importance of each faulty audio signal sample
Figure 104828DEST_PATH_IMAGE074
,in
Figure 936518DEST_PATH_IMAGE075
Represents the priority of fault type k ;

根据重要性

Figure 769345DEST_PATH_IMAGE076
大小，从大到小选出前L个存在故障的音频信号样本，并将这些选出的样本索引放在故障音频信号样本集
Figure 661078DEST_PATH_IMAGE077
中；according to importance
Figure 769345DEST_PATH_IMAGE076
Size, select the first L faulty audio signal samples from large to small, and put these selected sample indexes in the faulty audio signal sample set
Figure 661078DEST_PATH_IMAGE077
middle;

所述第一电力人员智能调度问题表示如下：The intelligent scheduling problem of the first electric power personnel is expressed as follows:

Figure 313776DEST_PATH_IMAGE012
；
Figure 313776DEST_PATH_IMAGE012
;

其中，

Figure 949157DEST_PATH_IMAGE078
表示电力人员
Figure 384293DEST_PATH_IMAGE014
的调度策略，电力人员l的调度策略表示为：如果电力人员l被调度维修第m个音频信号样本中的机器故障，那么
Figure 712506DEST_PATH_IMAGE079
，否则
Figure 852500DEST_PATH_IMAGE080
；第一电力人员智能调度问题（P1）采用匈牙利算法求解；L为可调度的电力人员人数，k为机器故障类型，K代表已知故障类型数目，第K+1种故障类型代表其它未知类型故障，
Figure 25992DEST_PATH_IMAGE081
为电力人员
Figure 318565DEST_PATH_IMAGE082
成功修复故障类型k的次数，
Figure 552100DEST_PATH_IMAGE083
为电力人员
Figure 179390DEST_PATH_IMAGE082
修复故障类型k的成功率，
Figure 422153DEST_PATH_IMAGE084
表示电力人员
Figure 552920DEST_PATH_IMAGE082
的调度策略，电力人员
Figure 957356DEST_PATH_IMAGE082
的调度策略表示为：如果电力人员
Figure 822675DEST_PATH_IMAGE082
被调度维修第m个音频信号样本中的机器故障，那么
Figure 869129DEST_PATH_IMAGE085
，否则
Figure 854402DEST_PATH_IMAGE086
。in,
Figure 949157DEST_PATH_IMAGE078
means electric staff
Figure 384293DEST_PATH_IMAGE014
The dispatching strategy of electric power personnel l is expressed as: if electric power personnel l is dispatched to repair the machine failure in the mth audio signal sample, then
Figure 712506DEST_PATH_IMAGE079
,otherwise
Figure 852500DEST_PATH_IMAGE080
; The first electric power personnel intelligent scheduling problem (P1) is solved by the Hungarian algorithm; L is the number of electric power personnel that can be dispatched, k is the type of machine fault, K represents the number of known fault types, and the K+ 1th fault type represents other unknown types of faults ,
Figure 25992DEST_PATH_IMAGE081
for electricians
Figure 318565DEST_PATH_IMAGE082
the number of successful repairs of fault type k ,
Figure 552100DEST_PATH_IMAGE083
for electricians
Figure 179390DEST_PATH_IMAGE082
the success rate of repairing fault type k ,
Figure 422153DEST_PATH_IMAGE084
means electric staff
Figure 552920DEST_PATH_IMAGE082
Scheduling strategy, power staff
Figure 957356DEST_PATH_IMAGE082
The scheduling strategy of is expressed as: If the power personnel
Figure 822675DEST_PATH_IMAGE082
is scheduled to repair the machine failure in the mth audio signal sample, then
Figure 869129DEST_PATH_IMAGE085
,otherwise
Figure 854402DEST_PATH_IMAGE086
.

若可调度的电力人员人数大于存在故障的音频信号样本的数量，则求解第二电力人员智能调度问题，最终获得电力人员智能调度方案。If the number of power personnel that can be dispatched is greater than the number of faulty audio signal samples, the second intelligent dispatching problem of electric personnel is solved, and finally the intelligent dispatching scheme of electric personnel is obtained.

所述第二电力人员智能调度问题表示如下：The intelligent scheduling problem of the second electric power personnel is expressed as follows:

Figure 695319DEST_PATH_IMAGE087
；
Figure 695319DEST_PATH_IMAGE087
;

其中，

Figure 31623DEST_PATH_IMAGE088
表示电力人员l的调度策略，电力人员l的调度策略表示为：如果电力人员l被调度维修第m个音频信号样本中的机器故障，那么
Figure 366920DEST_PATH_IMAGE089
，否则
Figure 472279DEST_PATH_IMAGE090
；第二电力人员智能调度问题（P2）采用匈牙利算法求解；L为可调度的电力人员人数，k为机器故障类型，
Figure 484098DEST_PATH_IMAGE091
为电力人员l的修复次数，
Figure 307697DEST_PATH_IMAGE092
为电力人员l的修复成功率， K代表已知故障类型数目，第K+1种故障类型代表其它未知类型故障，
Figure 430374DEST_PATH_IMAGE093
为电力人员
Figure 427059DEST_PATH_IMAGE094
的成功修复故障类型k的次数，
Figure 609779DEST_PATH_IMAGE095
为电力人员
Figure 920675DEST_PATH_IMAGE094
的修复故障类型k的成功率，
Figure 581463DEST_PATH_IMAGE096
表示电力人员
Figure 661415DEST_PATH_IMAGE094
的调度策略，电力人员
Figure 765768DEST_PATH_IMAGE094
的调度策略表示为：如果电力人员
Figure 298380DEST_PATH_IMAGE094
被调度维修第m个音频信号样本中的机器故障，那么
Figure 28439DEST_PATH_IMAGE097
，否则
Figure 228476DEST_PATH_IMAGE098
，
Figure 487419DEST_PATH_IMAGE099
表示第m个音频信号样本中的机器故障类型为K+1中类型中的某一种故障类型的概率，M为存在故障的音频信号样本总数。in,
Figure 31623DEST_PATH_IMAGE088
Represents the scheduling strategy of electrician l , and the dispatching strategy of electrician l is expressed as: if electrician l is dispatched to repair the machine failure in the mth audio signal sample, then
Figure 366920DEST_PATH_IMAGE089
,otherwise
Figure 472279DEST_PATH_IMAGE090
; The second power personnel intelligent scheduling problem (P2) is solved by the Hungarian algorithm; L is the number of power personnel that can be dispatched, k is the type of machine failure,
Figure 484098DEST_PATH_IMAGE091
is the number of repairs by the electrician l ,
Figure 307697DEST_PATH_IMAGE092
is the repair success rate of electrical personnel l , K represents the number of known fault types, and the K+ 1 fault type represents other unknown types of faults,
Figure 430374DEST_PATH_IMAGE093
for electricians
Figure 427059DEST_PATH_IMAGE094
The number of times of successfully repairing the fault type k ,
Figure 609779DEST_PATH_IMAGE095
for electricians
Figure 920675DEST_PATH_IMAGE094
The success rate of repairing fault type k ,
Figure 581463DEST_PATH_IMAGE096
means electric staff
Figure 661415DEST_PATH_IMAGE094
Scheduling strategy, power staff
Figure 765768DEST_PATH_IMAGE094
The scheduling strategy of is expressed as: If the power personnel
Figure 298380DEST_PATH_IMAGE094
is scheduled to repair the machine failure in the mth audio signal sample, then
Figure 28439DEST_PATH_IMAGE097
,otherwise
Figure 228476DEST_PATH_IMAGE098
,
Figure 487419DEST_PATH_IMAGE099
Indicates the probability that the machine failure type in the mth audio signal sample is a certain type of failure type in K+ 1, and M is the total number of audio signal samples with failures.

本发明基于机器的故障类型精准调度到对应维修经验丰富的电力人员，以最小的经济损失，快速高效地实现调度；针对人员调度，建立了不同情况下最大化机器故障平均修复率的电力人员调度问题，可以实现电力人员的智能调度，能够有效减少电力人员调度所消耗的时间。Based on the type of machine faults, the present invention accurately dispatches electric personnel with rich maintenance experience to realize dispatching quickly and efficiently with minimal economic loss; for personnel dispatching, a power personnel dispatching that maximizes the average repair rate of machine faults under different circumstances is established To solve the problem, the intelligent scheduling of electric power personnel can be realized, which can effectively reduce the time consumed by electric power personnel dispatching.

实施例2Example 2

本实施例提供一种电力人员智能调度方法，利用不同机器正常工作时的音频信号样本和环境中可能会存在的干扰音频信号样本对自编码器进行训练，让它能够以最低的重建误差，使得音频信号样本的MFCC特征得到重建，然后用训练完成的自编码器对测试集中的音频信号样本的MFCC特征进行重建，从而根据重建误差来判断机器故障是否发生。除此之外，利用机器异常工作的音频信号样本对CNN模型进行训练，以实现故障分类的功能，根据可调度的电力人员人数、电力人员故障修复成功次数、电力人员修复成功率以及机器故障类型构建智能调度问题；求解所述智能调度问题获得智能调度方案，从而能够在某一个机器发生某个类型的故障的时候，对电力人员实现精准调度，提高机器维修的效率。This embodiment provides an intelligent scheduling method for electric power personnel, using audio signal samples of different machines in normal operation and interference audio signal samples that may exist in the environment to train the autoencoder, so that it can use the lowest reconstruction error, so that The MFCC features of the audio signal samples are reconstructed, and then the trained autoencoder is used to reconstruct the MFCC features of the audio signal samples in the test set, so as to judge whether the machine failure occurs according to the reconstruction error. In addition, the CNN model is trained by using the audio signal samples of abnormal machine work to realize the function of fault classification. Construct an intelligent scheduling problem; solve the intelligent scheduling problem to obtain an intelligent scheduling plan, so that when a certain type of failure occurs in a certain machine, the electric power personnel can be accurately dispatched and the efficiency of machine maintenance can be improved.

如图1所示，本实施例包括以下步骤：As shown in Figure 1, this embodiment includes the following steps:

步骤101：在电力场景中采集多组不同机器正常工作时的音频信号样本、可能会存在的干扰音频信号样本以及少量常出现的机器异常工作时的音频信号样本，然后通过多种滤波器滤除非主要影响因素产生的噪声干扰；Step 101: Collect multiple groups of audio signal samples of different machines in normal operation, possible interference audio signal samples, and a small number of audio signal samples of abnormal operation of machines in the power scene, and then filter non- Noise interference from main influencing factors;

步骤102：通过对采集的音频信号样本进行预加重、分帧、加窗、快速傅里叶变换、Mel滤波以及离散余弦变换，最终得到能够描述音频信号样本的低阶MFCC特征值即梅尔频率倒谱系数特征向量；Step 102: By performing pre-emphasis, framing, windowing, fast Fourier transform, Mel filtering, and discrete cosine transform on the collected audio signal samples, the low-order MFCC eigenvalues that can describe the audio signal samples, namely the Mel frequency, are finally obtained Cepstral coefficient eigenvector;

步骤103：将提取到的机器正常工作时的音频信号样本的梅尔频率倒谱系数特征向量作为自编码器的输入，让它充分学习音频信号样本的特征，从而能够以最低的重建误差，使得音频信号样本的梅尔频率倒谱系数特征向量得到重建；Step 103: The extracted Mel-frequency cepstral coefficient eigenvector of the audio signal sample when the machine is working normally is used as the input of the autoencoder, so that it can fully learn the characteristics of the audio signal sample, so that it can use the lowest reconstruction error, so that The eigenvectors of the Mel-frequency cepstral coefficients of the audio signal samples are reconstructed;

步骤104：利用训练完成的自编码器，对测试集中的音频信号的MFCC特征进行重建，其中测试集中包含了机器正常工作音频信号样本、环境干扰音频信号样本以及机器异常工作音频信号样本；Step 104: Utilize the trained self-encoder to reconstruct the MFCC features of the audio signal in the test set, wherein the test set contains samples of the audio signal of the machine working normally, samples of the audio signal of environmental interference and samples of the audio signal of the machine working abnormally;

步骤105：通过对比重建前的特征向量

Figure 523640DEST_PATH_IMAGE100
与重建后的特征向量
Figure 526231DEST_PATH_IMAGE101
，将两者之间的误差作为音频信号样本的异常分数S，如果异常分数S大于某一个设定的异常分数阈值
Figure 580774DEST_PATH_IMAGE102
，则输入的音频信号样本为机器异常工作（即机器故障）的音频信号，反之，输入的音频信号样本为机器正常工作的音频信号或者环境中的某种干扰音频信号；Step 105: By comparing the feature vector before reconstruction
Figure 523640DEST_PATH_IMAGE100
and the reconstructed eigenvectors
Figure 526231DEST_PATH_IMAGE101
, take the error between the two as the abnormal score S of the audio signal sample, if the abnormal score S is greater than a certain set abnormal score threshold
Figure 580774DEST_PATH_IMAGE102
, then the input audio signal sample is the audio signal of the abnormal operation of the machine (that is, the machine failure), otherwise, the input audio signal sample is the audio signal of the normal operation of the machine or some kind of interfering audio signal in the environment;

步骤106：在步骤105的基础上，如果

Figure 276198DEST_PATH_IMAGE103
，表明机器出现异常，此时将此音频信号样本的梅尔频率倒谱系数特征向量输入到CNN模型中进行机器故障类型判断。由于CNN模型的训练样本数量不足，因此可能会出现CNN模型输出概率普遍较低的情况，那么将此类机器故障定义为其它未知类型故障。Step 106: On the basis of step 105, if
Figure 276198DEST_PATH_IMAGE103
, indicating that the machine is abnormal. At this time, the eigenvector of the Mel frequency cepstral coefficient of this audio signal sample is input into the CNN model to judge the type of machine failure. Due to the insufficient number of training samples of the CNN model, the output probability of the CNN model may be generally low, so such machine failures are defined as other unknown types of failures.

步骤107：基于步骤106所判定的机器故障类型以及机器故障数量，如果电力人员数L小于机器故障数M即存在故障的音频信号样本总数，则将电力人员优先分配给重要性

Figure 48982DEST_PATH_IMAGE076
值最大的机器故障，以最小化机器故障带来的损失，因此选出前L个重要性
Figure 605996DEST_PATH_IMAGE076
值最大的机器故障，标记为
Figure 515046DEST_PATH_IMAGE104
，
Figure 115792DEST_PATH_IMAGE074
；Step 107: Based on the type of machine failure and the number of machine failures determined in step 106, if the number L of electric personnel is less than the number M of machine failures, that is, the total number of faulty audio signal samples, the electric personnel will be assigned priority to the importance
Figure 48982DEST_PATH_IMAGE076
The machine failure with the largest value to minimize the loss caused by machine failure, so select the top L importance
Figure 605996DEST_PATH_IMAGE076
The machine fault with the largest value, labeled as
Figure 515046DEST_PATH_IMAGE104
,
Figure 115792DEST_PATH_IMAGE074
;

其中

Figure 375872DEST_PATH_IMAGE075
代表故障类型k的优先级；in
Figure 375872DEST_PATH_IMAGE075
Represents the priority of fault type k ;

并建立如下调度问题，其目标函数为最大化机器故障平均修复率，可表示为：And establish the following scheduling problem, the objective function of which is to maximize the average repair rate of machine failures, which can be expressed as:

Figure 720266DEST_PATH_IMAGE012
；
Figure 720266DEST_PATH_IMAGE012
;

其中，

Figure 749402DEST_PATH_IMAGE078
表示电力人员的调度策略，如果电力人员l被调度维修第m个异常样本中的机器故障，那么
Figure 534430DEST_PATH_IMAGE079
，否则
Figure 16227DEST_PATH_IMAGE080
。问题（P1）采用匈牙利算法求解。in,
Figure 749402DEST_PATH_IMAGE078
Indicates the scheduling strategy of electric power personnel, if electric power personnel l is dispatched to repair the machine failure in the mth abnormal sample, then
Figure 534430DEST_PATH_IMAGE079
,otherwise
Figure 16227DEST_PATH_IMAGE080
. Problem (P1) is solved using the Hungarian algorithm.

如果电力人员数L大于存在故障的音频信号样本总数M，则建立另一个最大化机器故障平均修复率的电力人员调度问题：If the number of electrical personnel L is greater than the total number of faulty audio signal samples M , another electrical personnel scheduling problem that maximizes the average repair rate of machine failures is established:

Figure 164312DEST_PATH_IMAGE087
；
Figure 164312DEST_PATH_IMAGE087
;

问题（P2）仍然采用匈牙利算法求解。Problem (P2) is still solved using the Hungarian algorithm.

通过对上述两个调度问题进行求解，最终可以得出电力人员的最优调度策略。By solving the above two dispatching problems, the optimal dispatching strategy for electric power personnel can be finally obtained.

如图2所示，根据音频信号样本确定机器是否存在故障，包括：As shown in Figure 2, it is determined whether the machine is faulty based on the audio signal samples, including:

提取机器工作时的音频信号样本的梅尔频率倒谱系数特征向量，通过编码器和解码器对梅尔频率倒谱系数特征向量进行重建；Extracting the Mel-frequency cepstral coefficient eigenvector of the audio signal sample when the machine is working, and reconstructing the Mel-frequency cepstral coefficient eigenvector through an encoder and a decoder;

将重建结果与异常分数阈值进行比较确定机器是否存在故障。The rebuild results are compared to an anomaly score threshold to determine if the machine is faulty.

提取机器工作时的音频信号样本的梅尔频率倒谱系数特征向量，包括：Extract the eigenvectors of Mel frequency cepstral coefficients of audio signal samples when the machine is working, including:

将音频信号样本的频率转化为梅尔频率；Convert the frequency of an audio signal sample to a Mel frequency;

将音频信号样本通过高通滤波器预加重，使音频信号样本保持在低频到高频的整个频带中，高通滤波器如：The audio signal sample is pre-emphasized through a high-pass filter to keep the audio signal sample in the entire frequency band from low frequency to high frequency. The high-pass filter is as follows:

Figure 47954DEST_PATH_IMAGE105
；
Figure 47954DEST_PATH_IMAGE105
;

其中，a为预加重系数；

Figure 256082DEST_PATH_IMAGE106
表示高通滤波器的系统函数；Among them, a is the pre-emphasis coefficient;
Figure 256082DEST_PATH_IMAGE106
Represents the system function of the high-pass filter;

对每段单帧信号进行加窗操作，将预加重后的音频信号样本分成多段单帧信号，每段单帧信号对应于一个频谱；Perform a windowing operation on each single-frame signal, divide the pre-emphasized audio signal samples into multiple single-frame signals, and each single-frame signal corresponds to a frequency spectrum;

加窗操作如下：The windowing operation is as follows:

Figure 975907DEST_PATH_IMAGE107
；
Figure 975907DEST_PATH_IMAGE107
;

其中，

Figure 927683DEST_PATH_IMAGE108
为加窗函数，N为帧长，n为帧数；in,
Figure 927683DEST_PATH_IMAGE108
is the windowing function, N is the frame length, and n is the number of frames;

利用短时快速傅里叶变换，计算频率与振幅的关系；短时快速傅里叶变换公式为：Use the short-time fast Fourier transform to calculate the relationship between frequency and amplitude; the short-time fast Fourier transform formula is:

Figure 931411DEST_PATH_IMAGE109
；
Figure 931411DEST_PATH_IMAGE109
;

其中，x(n)为第n帧的信号幅值，E(n)是变换后的能量信号；Among them, x ( n ) is the signal amplitude of the nth frame, E ( n ) is the transformed energy signal;

由于频域信号有很多冗余信息，需要滤波器组对频域的能量幅值进行精简，每一个频段用一个值来表示，且将频率转化为人耳对声音感知的梅尔频率，转化公式为：Since the frequency domain signal has a lot of redundant information, the filter bank is required to simplify the energy amplitude in the frequency domain. Each frequency band is represented by a value, and the frequency is converted into the Mel frequency that the human ear perceives sound. The conversion formula is :

Figure 310439DEST_PATH_IMAGE110
；
Figure 310439DEST_PATH_IMAGE110
;

其中，f为音频原始频率，单位为Hz。Wherein, f is the original audio frequency, and the unit is Hz.

音频原始频率就是音频信号样本本身的频率，然后通过梅尔频率转化公式，转换成梅尔频率。预加重是由于发声构造体的摩擦会导致信号的高频部分丢失，因此需要进行预加重，使音频信号保持在低频到高频的整个频带中。分帧是为了便于研究信号。加窗是由于分帧后，帧与帧之间有重叠，帧间频谱不平稳，所以需要加窗操作，使频谱变得平坦。快速傅里叶变换操作为了使时域信号转换为时频域上的能量分布。The original frequency of the audio is the frequency of the audio signal sample itself, and then converted to the Mel frequency through the Mel frequency conversion formula. Pre-emphasis is because the friction of the sound-emitting structure will cause the high-frequency part of the signal to be lost, so pre-emphasis is required to keep the audio signal in the entire frequency band from low frequency to high frequency. Framing is for the convenience of studying the signal. Windowing is because after framing, there is overlap between frames, and the spectrum between frames is not stable, so a windowing operation is required to make the spectrum flat. The Fast Fourier Transform operates in order to transform the time-domain signal into an energy distribution in the time-frequency domain.

对音频信号样本进行离散余弦变换得到梅尔频率倒谱系数，公式如下：The discrete cosine transform is performed on the audio signal samples to obtain the Mel frequency cepstral coefficients, the formula is as follows:

Figure 766829DEST_PATH_IMAGE045
；
Figure 766829DEST_PATH_IMAGE045
;

Figure 7448DEST_PATH_IMAGE046
；
Figure 7448DEST_PATH_IMAGE046
;

其中，f(n) 表示音频信号样本第n帧在时域上的信号，F(n) 是余弦变换后的系数，C(n) 为音频信号样本第n帧的补偿系数；Among them, f ( n ) represents the signal of the nth frame of the audio signal sample in the time domain, F ( n ) is the coefficient after cosine transformation, and C ( n ) is the compensation coefficient of the nth frame of the audio signal sample;

最终得到描述音频数据的低阶梅尔频率倒谱系数特征向量MFCC，Finally, the low-order Mel frequency cepstral coefficient feature vector MFCC describing the audio data is obtained,

其数据形式为：Its data format is:

Figure 600104DEST_PATH_IMAGE047
；
Figure 600104DEST_PATH_IMAGE047
;

其中F11表示第1个音频信号样本的第1帧的梅尔频率倒谱系数；F1N 表示第1个音频信号样本的第N帧的梅尔频率倒谱系数；F M1表示第M个音频信号样本的第1帧的梅尔频率倒谱系数，F MN 表示第M个音频信号样本的第N帧的梅尔频率倒谱系数。Among them, F 11 represents the Mel frequency cepstral coefficient of the first frame of the first audio signal sample; F 1 N represents the Mel frequency cepstral coefficient of the Nth frame of the first audio signal sample; F M 1 represents the Mth Mel frequency cepstral coefficient of the first frame of the audio signal sample, F MN represents the Mel frequency cepstral coefficient of the Nth frame of the M audio signal sample.

根据梅尔频率倒谱系数特征通过编码器和解码器对梅尔频率倒谱系数特征向量进行重建，具体步骤包括：According to the feature of Mel-frequency cepstral coefficient, the feature vector of Mel-frequency cepstral coefficient is reconstructed through encoder and decoder, and the specific steps include:

编码器E将输入特征向量

Figure 150034DEST_PATH_IMAGE100
，即获取到的音频信号样本的梅尔频率倒谱系数特征向量转换为潜在特征向量
Figure 93719DEST_PATH_IMAGE049
，即：The encoder E will input the feature vector
Figure 150034DEST_PATH_IMAGE100
, that is, the eigenvectors of Mel-frequency cepstral coefficients of the obtained audio signal samples are converted into latent eigenvectors
Figure 93719DEST_PATH_IMAGE049
,which is:

Figure 387297DEST_PATH_IMAGE050
；
Figure 387297DEST_PATH_IMAGE050
;

其中

Figure 850770DEST_PATH_IMAGE111
是编码器的参数；in
Figure 850770DEST_PATH_IMAGE111
is the parameter of the encoder;

解码器D将潜在特征向量

Figure 571602DEST_PATH_IMAGE049
重建成输入特征向量
Figure 737004DEST_PATH_IMAGE101
，即：Decoder D converts the latent feature vector
Figure 571602DEST_PATH_IMAGE049
Reconstruct into input feature vector
Figure 737004DEST_PATH_IMAGE101
,which is:

Figure 834273DEST_PATH_IMAGE112
；
Figure 834273DEST_PATH_IMAGE112
;

其中

Figure 401520DEST_PATH_IMAGE113
是解码器的参数；in
Figure 401520DEST_PATH_IMAGE113
is the parameter of the decoder;

上述两个式子中：

Figure 41056DEST_PATH_IMAGE114
，下标i表示当前是第i个特征向量；s为特征向量总数。In the above two formulas:
Figure 41056DEST_PATH_IMAGE114
, the subscript i indicates that it is the i -th eigenvector; s is the total number of eigenvectors.

将重建结果与异常分数阈值进行比较确定机器是否存在故障，具体步骤包括：Compare the reconstruction result with the abnormal score threshold to determine whether the machine is faulty, and the specific steps include:

调整编码器和解码器的参数，使重建数据和输入数据之间的误差最小化，重建误差函数为：Adjust the parameters of the encoder and decoder to minimize the error between the reconstructed data and the input data, and the reconstruction error function is:

Figure 693754DEST_PATH_IMAGE115
；
Figure 693754DEST_PATH_IMAGE115
;

Figure 594714DEST_PATH_IMAGE116
是误差函数；
Figure 750889DEST_PATH_IMAGE117
表示输入特征向量为
Figure 79102DEST_PATH_IMAGE100
的时候，所得出的重建误差值。
Figure 594714DEST_PATH_IMAGE116
is the error function;
Figure 750889DEST_PATH_IMAGE117
Indicates that the input feature vector is
Figure 79102DEST_PATH_IMAGE100
When , the obtained reconstruction error value is obtained.

将重建误差作为音频信号的异常分数S，确定机器是否存在故障的方式如下：Using the reconstruction error as the anomaly fraction S of the audio signal, the way to determine whether a machine is faulty is as follows:

Figure 969829DEST_PATH_IMAGE118
；
Figure 969829DEST_PATH_IMAGE118
;

Figure 408900DEST_PATH_IMAGE119
是输入特征向量为X的时候，所得出的重建误差值，
Figure 950740DEST_PATH_IMAGE120
为设定的异常分数阈值。
Figure 408900DEST_PATH_IMAGE119
is the reconstruction error value obtained when the input feature vector is X,
Figure 950740DEST_PATH_IMAGE120
Anomaly score threshold set for .

本实施例中，若存在故障则利用CNN模型进行分类获得机器故障类型，CNN模型的结构如图3所示，包括：In this embodiment, if there is a fault, the CNN model is used to classify and obtain the machine fault type. The structure of the CNN model is as shown in Figure 3, including:

被判定为机器故障时的音频信号的MFCC特征(即梅尔频率倒谱系数特征向量)作为CNN模型的输入，通过softmax函数输出此音频信号为已知故障类型的概率结果，最终对概率数组进行分析，得出音频信号所属类型，其过程如下：The MFCC feature (i.e., Mel frequency cepstral coefficient feature vector) of the audio signal when it is judged to be a machine failure is used as the input of the CNN model, and the probability result that the audio signal is a known failure type is output through the softmax function, and finally the probability array is processed Analyze to get the type of audio signal, the process is as follows:

将被判定为机器存在故障时的音频信号样本的梅尔频率倒谱系数特征向量作为CNN模型的输入，由于卷积神经网络softmax层输出的结果为分类概率矩阵，需要将该矩阵转换为分类结果，本实施例通过softmax函数输出此音频信号为已知故障类型的概率结果；The eigenvector of the Mel-frequency cepstral coefficient of the audio signal sample when the machine is judged to be faulty is used as the input of the CNN model. Since the output result of the softmax layer of the convolutional neural network is a classification probability matrix, the matrix needs to be converted into a classification result. , the present embodiment outputs the probability result that the audio signal is a known fault type through the softmax function;

对于一个样本X，假设经过softmax层输出的故障分类结果为对于一个音频信号样本X，假设经过softmax层输出的机器故障类型结果为：For a sample X, it is assumed that the fault classification result output by the softmax layer is For an audio signal sample X, it is assumed that the machine fault type result output by the softmax layer is:

Figure 184275DEST_PATH_IMAGE121
；
Figure 184275DEST_PATH_IMAGE121
;

其中，

Figure 811566DEST_PATH_IMAGE122
为softmax单个节点输出的机器故障类型为k的概率，K代表已知故障类型数目，第K+1种故障类型代表其它未知类型故障，且满足
Figure 788749DEST_PATH_IMAGE123
；in,
Figure 811566DEST_PATH_IMAGE122
It is the probability that the machine fault type output by a single softmax node is k , K represents the number of known fault types, and the K + 1th fault type represents other unknown types of faults, and satisfies
Figure 788749DEST_PATH_IMAGE123
;

输出的概率最大的机器故障类型就为最终确定的结果。The machine failure type with the highest output probability is the final result.

本发明考虑基于自编码器的无监督学习以及CNN模型的监督学习，解决了需要大量异常音频信号样本的难题，并且自编码器结构较为简单，容易训练；通过构造一个重建误差函数来衡量机器发生异常的可能性，从而可以有效地判断机器是否发生故障；通过建立一个最大化机器故障平均修复率的电力人员调度问题，利用匈牙利算法，最终可以得出最优的电力人员调度策略。The invention considers unsupervised learning based on autoencoder and supervised learning of CNN model, solves the problem of requiring a large number of abnormal audio signal samples, and the structure of autoencoder is relatively simple and easy to train; by constructing a reconstruction error function to measure the machine occurrence The possibility of abnormality, so that it can effectively judge whether the machine is faulty; by establishing a power personnel scheduling problem that maximizes the average repair rate of machine failures, and using the Hungarian algorithm, the optimal power personnel scheduling strategy can finally be obtained.

实施例3Example 3

与以上实施例提供的电力人员智能调度方法相对应地，本实施例提供了电力人员智能调度系统，包括：故障判断模块、故障类型判断模块以及调度方案求解模块；Corresponding to the intelligent dispatching method for electric power personnel provided in the above embodiments, this embodiment provides an intelligent dispatching system for electric power personnel, including: a fault judgment module, a fault type judgment module, and a dispatch scheme solution module;

故障判断模块，用于获取机器工作时的音频信号样本，根据音频信号样本确定机器是否存在故障；The fault judgment module is used to obtain audio signal samples when the machine is working, and determine whether the machine is faulty according to the audio signal samples;

故障类型判断模块，用于若判断机器存在故障，则根据音频信号样本确定机器故障类型；The fault type judging module is used to determine the machine fault type according to the audio signal sample if it is judged that the machine has a fault;

调度方案求解模块，用于根据可调度的电力人员人数、电力人员故障修复成功次数、修复成功率以及机器故障类型构建智能调度问题；求解所述智能调度问题获得智能调度方案。The dispatching plan solving module is used to construct an intelligent dispatching problem according to the number of dispatchable electric power personnel, the number of electric power personnel's fault repair successes, the repair success rate, and the type of machine fault; solving the intelligent dispatching problem to obtain an intelligent dispatching plan.

所属领域的技术人员可以清楚地了解到，为描述的方便和简洁，上述描述的系统，装置和单元的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。Those skilled in the art can clearly understand that for the convenience and brevity of the description, the specific working process of the above-described system, device and unit can refer to the corresponding process in the foregoing method embodiment, which will not be repeated here.

本领域内的技术人员应明白，本申请的实施例可提供为方法、系统、或计算机程序产品。因此，本申请可采用完全硬件实施例、完全软件实施例、或结合软件和硬件方面的实施例的形式。而且，本申请可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质（包括但不限于磁盘存储器、CD-ROM、光学存储器等）上实施的计算机程序产品的形式。Those skilled in the art should understand that the embodiments of the present application may be provided as methods, systems, or computer program products. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

本申请是参照根据本申请实施例的方法、设备（系统）、和计算机程序产品的流程图和／或方框图来描述的。应理解可由计算机程序指令实现流程图和／或方框图中的每一流程和／或方框、以及流程图和／或方框图中的流程和／或方框的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其他可编程数据处理设备的处理器以产生一个机器，使得通过计算机或其他可编程数据处理设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和／或方框图一个方框或多个方框中指定的功能的装置。The present application is described with reference to flowcharts and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It should be understood that each procedure and/or block in the flowchart and/or block diagram, and a combination of procedures and/or blocks in the flowchart and/or block diagram can be realized by computer program instructions. These computer program instructions may be provided to a general purpose computer, special purpose computer, embedded processor, or processor of other programmable data processing equipment to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing equipment produce a Means for realizing the functions specified in one or more steps of the flowchart and/or one or more blocks of the block diagram.

这些计算机程序指令也可存储在能引导计算机或其他可编程数据处理设备以特定方式工作的计算机可读存储器中，使得存储在该计算机可读存储器中的指令产生包括指令装置的制造品，该指令装置实现在流程图一个流程或多个流程和／或方框图一个方框或多个方框中指定的功能。These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to operate in a specific manner, such that the instructions stored in the computer-readable memory produce an article of manufacture comprising instruction means, the instructions The device realizes the function specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

这些计算机程序指令也可装载到计算机或其他可编程数据处理设备上，使得在计算机或其他可编程设备上执行一系列操作步骤以产生计算机实现的处理，从而在计算机或其他可编程设备上执行的指令提供用于实现在流程图一个流程或多个流程和／或方框图一个方框或多个方框中指定的功能的步骤。These computer program instructions can also be loaded onto a computer or other programmable data processing device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby The instructions provide steps for implementing the functions specified in the flow chart flow or flows and/or block diagram block or blocks.

以上结合附图对本发明的实施例进行了描述，但是本发明并不局限于上述的具体实施方式，上述的具体实施方式仅仅是示意性的，而不是限制性的，本领域的普通技术人员在本发明的启示下，在不脱离本发明宗旨和权利要求所保护的范围情况下，还可做出很多形式，这些均属于本发明的保护之内。Embodiments of the present invention have been described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementations, and the above-mentioned specific implementations are only illustrative, rather than restrictive. Under the enlightenment of the present invention, many forms can also be made without departing from the gist of the present invention and the protection scope of the claims, and these all belong to the protection of the present invention.

Claims (7)

1. The intelligent scheduling method for the power personnel is characterized by comprising the following steps:
acquiring an audio signal sample when the machine works, and determining whether the machine has a fault according to the audio signal sample;
if the machine is judged to have a fault, determining the fault type of the machine according to the audio signal sample;
constructing an intelligent scheduling problem according to the number of dispatchable electric power personnel, the fault repair success times of the electric power personnel, the repair success rate and the machine fault type;
solving the intelligent scheduling problem to obtain an intelligent scheduling scheme;
constructing an intelligent scheduling problem, comprising: determining the number of dispatchable electric power personnel, and comparing the number of dispatchable electric power personnel with the number of audio signal samples with faults;
if the number of dispatchable electric power personnel is less than or equal to the number of audio signal samples with faults, solving the intelligent dispatching problem of the first electric power personnel;
electric staff number recording assuming schedulableLEstablishing maintenance files of each power personnel according to historical maintenance records of the power personnelCounter, electric power personnellMaintenance profile of = { machine fault typekElectric power personnellSuccessfully repairing a failure typekNumber of times of
Figure 747577DEST_PATH_IMAGE001
Electric power personnellRepairing failure typeskSuccess rate of
Figure 375611DEST_PATH_IMAGE002
，
Figure 547967DEST_PATH_IMAGE003
｝，KRepresenting the number of known fault types, the first K+1 fault type represents other unknown types of faults;
suppose that the current time is commonMA failed audio signal sample, each sample
Figure 437425DEST_PATH_IMAGE004
,m=1,2,…,MIs expressed as a fault classification result of (a),
Figure 378837DEST_PATH_IMAGE005
；
wherein
Figure 944816DEST_PATH_IMAGE006
Is shown asmThe type of machine fault in the audio signal sample isK+Failure type of 1 typekThe probability of (d);
calculating the importance of each audio signal sample with faults
Figure 604467DEST_PATH_IMAGE007
Wherein
Figure 297617DEST_PATH_IMAGE008
Representing the type of faultkThe priority of (2);
according to importance
Figure 93535DEST_PATH_IMAGE009
Size, before selection from large to smallLThe audio signal samples with faults are stored and the selected samples are indexed in the fault audio signal sample set
Figure 597459DEST_PATH_IMAGE010
Performing the following steps;
the first power personnel intelligent scheduling problem is expressed as follows:
Figure 213249DEST_PATH_IMAGE011
；
wherein,
Figure 710089DEST_PATH_IMAGE012
indicating electric personnellScheduling policy of, electric power personnellThe scheduling policy of (a) is expressed as: if the power personnellIs scheduled for maintenancemA machine failure of an audio signal sample, then
Figure 360513DEST_PATH_IMAGE014
Otherwise
Figure 550186DEST_PATH_IMAGE016
；
Figure 168118DEST_PATH_IMAGE018
For personnel of electric power
Figure 203070DEST_PATH_IMAGE019
Successfully repairing a failure typekThe number of times of the above-mentioned operations,
Figure 708001DEST_PATH_IMAGE020
for electric power personnel
Figure 68575DEST_PATH_IMAGE019
Repairing failure typeskThe success rate of the process is increased,
Figure 236837DEST_PATH_IMAGE021
indicating electric personnel
Figure 75480DEST_PATH_IMAGE019
Scheduling policy of, electric power personnel
Figure 700497DEST_PATH_IMAGE019
The scheduling policy of (a) is expressed as: if the power personnel
Figure 966393DEST_PATH_IMAGE019
Is scheduled for maintenancemA machine fault in a sample of the audio signal, then
Figure 309650DEST_PATH_IMAGE023
Otherwise
Figure 935672DEST_PATH_IMAGE025
；
If the number of dispatchable power personnel is larger than the number of audio signal samples with faults, solving a second power personnel intelligent dispatching problem, and finally obtaining a power personnel intelligent dispatching scheme;
the second power personnel intelligent scheduling problem is expressed as follows:
Figure 680774DEST_PATH_IMAGE026
；
wherein,
Figure 117571DEST_PATH_IMAGE027
indicating electric personnellScheduling policy of, electric power personnellThe scheduling policy of (a) is expressed as: if the power personnellIs scheduled for maintenancemA machine fault in a sample of the audio signal, then
Figure 682545DEST_PATH_IMAGE028
Otherwise
Figure 613723DEST_PATH_IMAGE029
；LThe number of the electric power personnel can be dispatched,
Figure 478911DEST_PATH_IMAGE030
for electric power personnellSuccessfully repairing failure typeskThe number of times of the operation of the motor,
Figure 86610DEST_PATH_IMAGE031
for electric power personnellRepairing failure typeskThe success rate of the process is increased,
Figure 138879DEST_PATH_IMAGE032
，Krepresenting the number of known fault types, the firstK+The 1 fault type represents the other unknown type of fault,
Figure 123016DEST_PATH_IMAGE033
for electric power personnel
Figure 826399DEST_PATH_IMAGE034
Successfully repairing the failure typekThe number of times of the operation of the motor,
Figure 870578DEST_PATH_IMAGE035
for electric power personnel
Figure 878985DEST_PATH_IMAGE034
Repairing failure typeskThe success rate of the process is increased,
Figure 666813DEST_PATH_IMAGE036
indicating electric personnel
Figure 975434DEST_PATH_IMAGE034
Scheduling policy of, electric power personnel
Figure 938318DEST_PATH_IMAGE034
The scheduling policy of (a) is expressed as: if the power personnel
Figure 965180DEST_PATH_IMAGE034
Is scheduled for maintenancemA machine fault in a sample of the audio signal, then
Figure 25539DEST_PATH_IMAGE037
Otherwise, otherwise
Figure 719826DEST_PATH_IMAGE038
，
Figure 840229DEST_PATH_IMAGE039
Is shown asmThe type of machine fault in the audio signal sample isKProbability of a certain fault type of +1 types,Mtotal number of audio signal samples for which a fault exists;
and solving the first power personnel intelligent scheduling problem and the second power personnel intelligent scheduling problem by adopting a Hungarian algorithm.
2. The intelligent dispatching method for electric power personnel as claimed in claim 1, wherein the step of determining whether the machine has a fault according to the audio signal samples comprises the following steps:
extracting the feature vector of the Mel frequency cepstrum coefficient of the audio signal sample when the machine works, and reconstructing the feature vector of the Mel frequency cepstrum coefficient through an encoder and a decoder;
and comparing the reconstruction result with an abnormal score threshold value to determine whether the machine has a fault.
3. The intelligent electric power personnel scheduling method of claim 2, wherein the extraction of mel-frequency cepstrum coefficient feature vectors of audio signal samples during the operation of the machine comprises the following specific steps:
converting the audio signal sample frequency into a mel frequency;
the audio signal samples are pre-emphasized by a high pass filter, such that the audio signal samples remain in the entire band from low frequencies to high frequencies, the high pass filter being:
Figure 603654DEST_PATH_IMAGE040
；
wherein,ais a pre-emphasis coefficient;
Figure 467705DEST_PATH_IMAGE041
a system function representing a high pass filter;
windowing each section of single-frame signal, dividing the audio signal sample after pre-emphasis into a plurality of sections of single-frame signals, wherein each section of single-frame signal corresponds to a frequency spectrum;
the windowing operation is as follows:
Figure 750919DEST_PATH_IMAGE042
；
wherein,
Figure 42223DEST_PATH_IMAGE043
in order to be a function of the windowing,Nin order to make the frame length longer,nis the frame number;
calculating the relation between the frequency and the amplitude by using short-time fast Fourier transform; the short-time fast fourier transform formula is:
Figure 528830DEST_PATH_IMAGE044
；
wherein,x(n) Is a firstnThe amplitude of the signal of the frame,E(n) Is the transformed energy signal;
discrete cosine transform is carried out on the audio signal sample to obtain a Mel frequency cepstrum coefficient, and the formula is as follows:
Figure 196572DEST_PATH_IMAGE045
；
Figure 599871DEST_PATH_IMAGE046
；
wherein,f(n) Representing samples of an audio signalnThe signal of the frame in the time domain,F(n) Is the coefficient after the cosine transform and,C(n) Is an audio signal samplenA compensation coefficient for the frame;
finally, the feature vector of the low-order Mel frequency cepstrum coefficient describing the audio data is obtainedMFCC，
The data form is as follows:
Figure 62077DEST_PATH_IMAGE047
；
wherein F 11 Mel-frequency cepstral coefficients representing a 1 st frame of the 1 st audio signal sample; f N1 Mel-frequency cepstral coefficients representing an nth frame of the 1 st audio signal sample; f M1 Merr-frequency cepstrum coefficient, F, representing the 1 st frame of the Mth audio signal sample MN The mel-frequency cepstral coefficients of the nth frame representing the mth audio signal sample.
4. The intelligent electric power personnel scheduling method of claim 2 wherein reconstructing the mel-frequency cepstrum coefficient feature vector by the encoder and decoder comprises:
encoder E input feature vector
Figure 285248DEST_PATH_IMAGE048
Converting the feature vector of the Mel frequency cepstrum coefficient of the obtained audio signal sample into potential feature vector
Figure 271527DEST_PATH_IMAGE049
Expressed as:
Figure 263754DEST_PATH_IMAGE050
；
wherein
Figure 162440DEST_PATH_IMAGE051
Is a parameter of the encoder;
decoder D will be latent feature vector
Figure 872907DEST_PATH_IMAGE052
Reconstructing into input feature vectors
Figure 882451DEST_PATH_IMAGE053
Expressed as:
Figure 482846DEST_PATH_IMAGE054
；
wherein
Figure 552434DEST_PATH_IMAGE055
Is a parameter of the decoder;
Figure 750197DEST_PATH_IMAGE056
subscriptiIndicates that it is currently the firstiA feature vector;sthe total number of feature vectors.
5. The intelligent dispatching method for the electric power personnel as claimed in claim 4, wherein the reconstruction result is compared with the abnormal score threshold value to determine whether the machine has a fault, and the specific steps comprise:
adjusting parameters of the encoder and decoder to minimize an error between the reconstructed data and the input data, the reconstruction error function being:
Figure 563432DEST_PATH_IMAGE057
；
Figure 530251DEST_PATH_IMAGE058
is an error function;
Figure 754428DEST_PATH_IMAGE059
representing the input feature vector as
Figure 173908DEST_PATH_IMAGE060
Obtaining a reconstruction error value;
taking reconstruction error as an abnormal score of an audio signalSThe manner of determining whether a machine has a fault is as follows:
Figure 790834DEST_PATH_IMAGE061
；
Figure 612159DEST_PATH_IMAGE062
when the input feature vector is X, the obtained reconstruction error value,
Figure 39861DEST_PATH_IMAGE063
is a set anomaly score threshold.
6. The intelligent scheduling method for power personnel according to claim 2, wherein if it is determined that the machine has a fault, classifying by using a CNN model to obtain a fault type of the machine comprises:
taking the Mel frequency cepstrum coefficient feature vector of the audio signal sample judged to have the fault on the machine as the input of the CNN model, and outputting the probability result that the audio signal is of the known fault type through the softmax function; for one audio signal sample X, assume that the machine fault type result output via the softmax layer is:
Figure 946637DEST_PATH_IMAGE064
；
wherein,
Figure 101675DEST_PATH_IMAGE065
the type of machine fault output for the single node of softmax iskThe probability of (a) of (b) being,Krepresenting the number of known fault types, the firstK+1 fault types represent other unknown types of faults and satisfy
Figure 777507DEST_PATH_IMAGE066
；
The type of the machine fault with the highest output probability is the final determined result.
7. Electric power personnel intelligent dispatching system, its characterized in that includes: the system comprises a fault judgment module, a fault type judgment module and a scheduling scheme solving module;
the fault judgment module is used for acquiring an audio signal sample when the machine works and determining whether the machine has a fault according to the audio signal sample;
the fault type judging module is used for determining the fault type of the machine according to the audio signal sample if the machine is judged to have a fault;
the scheduling scheme solving module is used for constructing an intelligent scheduling problem according to the number of schedulable electric personnel, the failure repair success times of the electric personnel, the repair success rate and the machine failure type; solving the intelligent scheduling problem to obtain an intelligent scheduling scheme;
determining the number of dispatchable electric power personnel, and comparing the number of dispatchable electric power personnel with the number of audio signal samples with faults;
if the number of dispatchable electric power personnel is less than or equal to the number of audio signal samples with faults, solving the intelligent dispatching problem of the first electric power personnel;
electric staff number recording assuming schedulableLAccording to the historical maintenance records of the electric power personnel, a maintenance file of each electric power personnel is established, and the electric power personnellOfRepair file = { machine fault typekElectric power personnellSuccessfully repairing a failure typekNumber of times of
Figure 609065DEST_PATH_IMAGE001
Electric power personnellRepairing failure typeskSuccess rate of
Figure 268717DEST_PATH_IMAGE002
，
Figure 961866DEST_PATH_IMAGE003
｝，KRepresenting the number of known fault types, the first K+1 fault type represents other unknown types of faults;
suppose that the current time is commonMA failed audio signal sample, each sample
Figure 23363DEST_PATH_IMAGE067
,m=1,2,…,MIs expressed as
Figure 510976DEST_PATH_IMAGE005
Wherein
Figure 140148DEST_PATH_IMAGE006
Is shown asmThe type of machine fault in the audio signal sample isK+Failure type of 1 typekThe probability of (d);
calculating the importance of each audio signal sample with faults
Figure 902567DEST_PATH_IMAGE007
Wherein
Figure 287412DEST_PATH_IMAGE008
Representing the type of faultkThe priority of (2);
according to importance
Figure 477085DEST_PATH_IMAGE009
Size, before selection from large to smallLThe audio signal samples with faults are stored and the selected samples are indexed in the fault audio signal sample set
Figure 580170DEST_PATH_IMAGE010
Performing the following steps;
the first power personnel intelligent scheduling problem is expressed as follows:
Figure 129969DEST_PATH_IMAGE068
；
wherein,
Figure 900479DEST_PATH_IMAGE069
indicating electric personnellScheduling policy of, electric power personnellThe scheduling policy of (a) is expressed as: if the power personnellIs scheduled for maintenancemA machine failure of an audio signal sample, then
Figure 995474DEST_PATH_IMAGE070
Otherwise
Figure 851435DEST_PATH_IMAGE071
；
Figure 690078DEST_PATH_IMAGE017
For electric power personnel
Figure 65826DEST_PATH_IMAGE019
Successfully repairing a failure typekThe number of times of the operation of the motor,
Figure 331723DEST_PATH_IMAGE072
for electric power personnel
Figure 674979DEST_PATH_IMAGE019
Repairing failure typeskThe success rate of the process is increased,
Figure 51734DEST_PATH_IMAGE021
indicating electric personnel
Figure 780524DEST_PATH_IMAGE019
Scheduling policy of, electric power personnel
Figure 482901DEST_PATH_IMAGE019
The scheduling policy of (a) is expressed as: if the power personnel
Figure 47875DEST_PATH_IMAGE019
Is scheduled for maintenancemA machine fault in a sample of the audio signal, then
Figure 228320DEST_PATH_IMAGE073
Otherwise, otherwise
Figure 38714DEST_PATH_IMAGE074
；
If the number of dispatchable power personnel is larger than the number of audio signal samples with faults, solving a second power personnel intelligent dispatching problem, and finally obtaining a power personnel intelligent dispatching scheme;
the second power personnel intelligent scheduling problem is expressed as follows:
Figure 380834DEST_PATH_IMAGE026
；
wherein,
Figure 698682DEST_PATH_IMAGE027
indicating electric personnellScheduling policy of, electric power personnellThe scheduling policy of (a) is expressed as: if the power personnellIs scheduled for maintenancemA machine fault in a sample of the audio signal, then
Figure 682819DEST_PATH_IMAGE028
Otherwise
Figure 136934DEST_PATH_IMAGE029
；LThe number of the dispatchable electric power personnel is,
Figure 164802DEST_PATH_IMAGE030
for electric power personnellSuccessfully repairing a failure typekThe number of times of the above-mentioned operations,
Figure 704367DEST_PATH_IMAGE031
for electric power personnellRepairing failure typeskThe success rate of the process is increased,
Figure 226616DEST_PATH_IMAGE032
，Krepresenting the number of known fault types, the firstK+The 1 fault type represents the other unknown type of fault,
Figure 800816DEST_PATH_IMAGE033
for electric power personnel
Figure 501050DEST_PATH_IMAGE034
Successfully repairing the failure typekThe number of times of the operation of the motor,
Figure 527912DEST_PATH_IMAGE035
for electric power personnel
Figure 853851DEST_PATH_IMAGE034
Repairing failure typeskThe success rate of the process is increased,
Figure 282559DEST_PATH_IMAGE036
indicating electric personnel
Figure 668541DEST_PATH_IMAGE034
Scheduling policy of, electric power personnel
Figure 166387DEST_PATH_IMAGE034
Is scheduled byThe strategy is represented as: if the power personnel
Figure 296017DEST_PATH_IMAGE034
Is scheduled for maintenancemA machine fault in a sample of the audio signal, then
Figure 579231DEST_PATH_IMAGE037
Otherwise, otherwise
Figure 870535DEST_PATH_IMAGE038
，
Figure 606410DEST_PATH_IMAGE039
Denotes the firstmThe type of machine fault in the audio signal sample isKProbability of one of +1 types of failure,Mthe total number of samples of the audio signal for which a fault exists.
CN202210919919.5A 2022-08-02 2022-08-02 Method and system for intelligent dispatching of electric power personnel Active CN114997749B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210919919.5A CN114997749B (en) 2022-08-02 2022-08-02 Method and system for intelligent dispatching of electric power personnel

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210919919.5A CN114997749B (en) 2022-08-02 2022-08-02 Method and system for intelligent dispatching of electric power personnel

Publications (2)

Publication Number Publication Date
CN114997749A CN114997749A (en) 2022-09-02
CN114997749B true CN114997749B (en) 2022-11-11

Family

ID=83021790

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210919919.5A Active CN114997749B (en) 2022-08-02 2022-08-02 Method and system for intelligent dispatching of electric power personnel

Country Status (1)

Country Link
CN (1) CN114997749B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115511397B (en) * 2022-11-23 2023-04-07 广东华居科技有限公司 Intelligent work order data dispatching method and system

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110689258A (en) * 2019-09-25 2020-01-14 重庆戴普思科技有限公司 Vehicle maintenance task scheduling method and system
CN111325095B (en) * 2020-01-19 2024-01-30 西安科技大学 Intelligent detection method and system for equipment health state based on acoustic wave signals
CN112258065B (en) * 2020-10-29 2024-12-20 上海电气集团股份有限公司 A method, device and equipment for determining maintenance work order

Also Published As

Publication number Publication date
CN114997749A (en) 2022-09-02

Similar Documents

Publication Publication Date Title
CN112885372B (en) Intelligent diagnosis method, system, terminal and medium for power equipment fault sound
CN111325095A (en) Method and system for intelligent detection of equipment health status based on acoustic signal
CN113314144B (en) Voice recognition and power equipment fault early warning method, system, terminal and medium
CN112383052A (en) Power grid fault repairing method and device based on power internet of things
CN109190672A (en) Operation of Electric Systems operating condition unsupervised clustering and device
JP2023143703A (en) Extraction of instantaneously renewable power generation from pure load measurement of power distribution system
CN113077812B (en) Voice signal generation model training method, echo cancellation method, device and equipment
CN114997749B (en) Method and system for intelligent dispatching of electric power personnel
CN118312734B (en) Power system equipment state monitoring and fault diagnosis method
CN107644231A (en) A kind of generator amature method for diagnosing faults and device
CN111325233B (en) Transformer fault detection method and device
CN118035923B (en) A method for identifying abnormal signals in power grid recording
CN112464708A (en) Double-fed asynchronous fan power quality abnormal fault diagnosis method
CN116882164A (en) Valve cooling equipment fault diagnosis method based on Fisher ratio and improved LSSVM algorithm
CN117217730A (en) Power equipment fault identification method, device, equipment, medium and product
CN116778956A (en) Transformer acoustic feature extraction and fault identification method
CN115982621A (en) Rotary machine residual service life prediction method based on time convolution network
CN108562853A (en) Method of Motor Fault Diagnosis based on error correcting output codes support vector machines and system
CN114121025A (en) Voiceprint fault intelligent detection method and device for substation equipment
CN115406630A (en) Method for detecting faults of wind driven generator blades through passive acoustic signals based on machine learning
CN114722964B (en) Digital audio tampering passive detection method and device based on power grid frequency space and time series feature fusion
CN116072146A (en) Pumped storage station detection method and system based on voiceprint recognition
CN106596111A (en) Adaptive water power set fault characteristic extraction method based on multiwavelets
CN114822590B (en) Digital audio tampering passive detection method and device based on power grid frequency phase timing sequence characterization
Cheng et al. Early Fault Diagnosis of Medium Voltage Covered Conductor Overhead Lines

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant