Equipment fault prevention method and system
Through the combination of Kalman filtering, DPCA, GRU neural network and LOF algorithm, the problems of slow calculation speed and low accuracy in turbine generator fault prevention are solved, and efficient and accurate fault warning and prevention are achieved.
Patent Information
- Application Number
- CN202510869592.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-26
- Publication Date
- 2025-10-17
Smart Images
Figure CN120804571A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of electronic system equipment failure prevention, and in particular to an equipment failure prevention method and system. BACKGROUND
[0002] The steam turbine generator is a very critical equipment in the power plant, and its stable operation directly affects the safety and economy of the power plant. In order to prevent the equipment failure of the steam turbine generator, it is very important to take a series of preventive measures. The failure prevention method of the steam turbine generator can be carried out by various means, such as regular maintenance inspection, vibration, temperature and electrical parameter monitoring, etc. By comprehensively using the above methods, the equipment failure rate can be effectively reduced, the reliability and operation efficiency of the equipment can be improved, and the service life of the equipment can be prolonged. The core of preventive maintenance is "early detection and early treatment", and through real-time monitoring and intelligent analysis, potential problems can be found in time to avoid major failures.
[0003] Among the two prevention methods of regular maintenance inspection and data monitoring, regular maintenance inspection generally consumes time and effort, and the accuracy of the detection result is not necessarily high. Therefore, in order to avoid time and effort and improve the accuracy of detection, the existing technology can prevent the equipment failure in the steam turbine generator by monitoring the vibration, temperature and electrical parameters, and predicting by monitoring and analyzing the historical data. However, since the data is disturbed by noise during collection, there is a great error when using the historical data with noise to perform prediction analysis, thereby reducing the accuracy of equipment failure prevention.
[0004] Since there are a large amount of data in the process of analyzing the historical data, the calculation speed is slow when calculated and analyzed by a computer, which cannot quickly feedback the failure condition of the equipment in time. Therefore, a large amount of data also reduces the accuracy of equipment failure prevention. SUMMARY
[0005] The present application provides an equipment failure prevention method and system, which solves the problems of slow calculation speed and reduced accuracy of equipment failure prevention in the prior art.
[0006] The purpose of the present application can be achieved by the following technical solutions: The first aspect of the present application provides an equipment failure prevention method, comprising: Obtaining a plurality of original time sequence signals of the steam turbine generator in the working process; Filtering and dimensionality reduction preprocessing the plurality of original time sequence signals to obtain a plurality of characteristic time sequence signals; The original time sequence signal is filtered and dimensionally preprocessed to obtain a plurality of feature time sequence signals, and the plurality of feature time sequence signals are divided to obtain a plurality of data, and each time sequence signal sequence is obtained through the plurality of data; each time sequence signal sequence is divided into a front signal sequence and a rear signal sequence; an initial prediction sequence is obtained through prediction according to the front signal sequence, and an error value of each feature time sequence signal is obtained through an error between the initial prediction sequence and the rear signal sequence; the error values of the plurality of feature time sequence signals are obtained and combined into a sequence, and the sequence is recorded as an error value sequence; and a plurality of monitoring signals are screened from all feature time sequence signals according to the error value sequence; The monitoring feature space is constructed through the plurality of monitoring signals, each monitoring signal is divided to obtain a signal sequence of each monitoring signal; the signal sequences of the plurality of monitoring signals are mapped in the monitoring feature space to obtain a plurality of data points, which are recorded as reference points; the data of the plurality of monitoring signals corresponding to each time point in the future is predicted through the signal sequences of the plurality of monitoring signals to obtain the data of the plurality of monitoring signals corresponding to each time point in the future; the data of the plurality of monitoring signals corresponding to each time point in the future is mapped in the monitoring feature space to obtain a to-be-analyzed point corresponding to each time point in the future; and the device fault of the steam turbine generator is prevented according to the distribution of all reference points and to-be-analyzed points in the monitoring feature space.
[0007] Further, the original time sequence signal is filtered and dimensionally preprocessed to obtain a plurality of feature time sequence signals, and the plurality of feature time sequence signals are divided to obtain a plurality of data, and each time sequence signal sequence is obtained through the plurality of data; each time sequence signal sequence is divided into a front signal sequence and a rear signal sequence, and the plurality of feature time sequence signals are obtained through the plurality of data. Each original time sequence signal is filtered by using a Kalman filtering algorithm to obtain each filtered time sequence signal, which is recorded as each filtered time sequence signal. The plurality of main signals are extracted from all filtered time sequence signals through a DPCA algorithm, and the plurality of main signals are recorded as the plurality of feature time sequence signals.
[0008] Further, the original time sequence signal is filtered and dimensionally preprocessed to obtain a plurality of feature time sequence signals, and the plurality of feature time sequence signals are divided to obtain a plurality of data, and each time sequence signal sequence is obtained through the plurality of data; each time sequence signal sequence is divided into a front signal sequence and a rear signal sequence, and the plurality of feature time sequence signals are obtained through the plurality of data. Each feature time sequence signal is divided at a preset time interval to obtain a plurality of data corresponding to each feature time sequence signal after division, and the plurality of data is combined into a sequence in chronological order, which is recorded as each time sequence signal sequence. The data in front of each time sequence signal sequence is combined into a front signal sequence, and the data behind each time sequence signal sequence is combined into a rear signal sequence; wherein, is a preset division percentage factor.
[0009] Further, the prediction according to the previous signal sequence obtains a set of initial prediction sequences, and the error value of each characteristic time sequence signal is obtained through the error between the initial prediction sequence and the subsequent signal sequence, comprising: The previous signal sequence is put into the GRU neural network for prediction to obtain a set of initial prediction sequences; The error between the initial prediction sequence and the subsequent signal sequence is calculated through MAPE, and is recorded as the error value of each characteristic time sequence signal.
[0010] Further, the error values of the plurality of characteristic time sequence signals are obtained, and a set of sequences is formed, which is recorded as an error value sequence; and a plurality of monitoring signals are screened from all characteristic time sequence signals according to the error value sequence, comprising: The error values of all characteristic time sequence signals are calculated, and the error values of all characteristic time sequence signals are sorted in ascending order to obtain an error value sequence; The time sequence signal data corresponding to the error value at the front of the error value sequence is recorded as monitoring signal data; wherein, The preset distinguishing factor is represented.
[0011] Further, the monitoring feature space is constructed through the plurality of monitoring signals, each monitoring signal is divided to obtain the signal sequence of each monitoring signal, and the signal sequences of the plurality of monitoring signals are mapped in the monitoring feature space to obtain a plurality of data points, which are recorded as reference points, comprising: Each monitoring signal is taken as a dimension to construct the monitoring feature space; Each monitoring signal is divided into a plurality of data through a preset time interval , and the data in the signal sequence of each monitoring signal is composed in time sequence; all data in the signal sequences of all monitoring signals are mapped in the monitoring feature space to obtain a plurality of data points, and the plurality of data points are recorded as reference points.
[0012] Further, the prediction of the data corresponding to the plurality of monitoring signals at each subsequent moment is carried out through the signal sequences of the plurality of monitoring signals to obtain the monitoring signal data of the plurality of monitoring signals at each subsequent moment; and a to-be-analyzed point corresponding to each subsequent moment is obtained by mapping the monitoring signal data of the plurality of monitoring signals at each subsequent moment in the monitoring feature space, comprising: The prediction of each monitoring signal at a plurality of subsequent moments is carried out through the GRU neural network according to the signal sequence of each monitoring signal to obtain the monitoring signal data of each monitoring signal at a plurality of subsequent moments; Obtain all monitoring signal data of each subsequent time, map all monitoring signal data of each subsequent time in the monitoring feature space, and obtain a data point corresponding to each subsequent time; the data points mapped in the feature space of all subsequent times are recorded as analysis points.
[0013] Further, the prevention of the equipment fault of the turbogenerator according to the distribution of all reference points and analysis points in the monitoring feature space comprises: According to the LOF algorithm, the outlier factor of each analysis point is calculated according to all reference points and all analysis points in the monitoring feature space, and then the outlier factor of all analysis points is normalized to obtain the normalized outlier factor value of each analysis point. When the normalized outlier factor value of the analysis point of each subsequent time is greater than or equal to a preset outlier threshold , an alarm is given in the background, and the staff is notified to check, so as to prevent the occurrence of equipment failure.
[0014] The second aspect of the present application provides a system for preventing equipment failure, comprising a data acquisition module for obtaining several kinds of original time series signals of the turbogenerator during operation; a preprocessing module for filtering and dimensionality reduction preprocessing of the several kinds of original time series signals to obtain several kinds of feature time series signals; a feature dimension screening module for dividing each feature time series signal to obtain several data, obtaining each time series signal sequence through the several data, dividing each time series signal sequence into a front signal sequence and a rear signal sequence, predicting according to the front signal sequence to obtain an initial prediction sequence, obtaining the error value of each feature time series signal through the error between the initial prediction sequence and the rear signal sequence, obtaining an error value sequence by obtaining the error value of the several feature time series signals, and screening several monitoring signals from all feature time series signals according to the error value sequence; a fault prevention module for constructing a monitoring feature space through the several monitoring signals, dividing each monitoring signal to obtain a signal sequence of each monitoring signal, mapping the signal sequences of the several monitoring signals in the monitoring feature space to obtain several data points, recording the data points as reference points, predicting the corresponding data of the several monitoring signals at each subsequent time through the signal sequences of the several monitoring signals to obtain the monitoring signal data of each subsequent time, mapping the monitoring signal data of each subsequent time in the monitoring feature space to obtain an analysis point corresponding to each subsequent time, and preventing the equipment fault of the turbogenerator according to the distribution of all reference points and analysis points in the monitoring feature space.
[0015] The third aspect of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the device fault prevention method when executing the computer program.
[0016] Compared with the prior art, the present application has the beneficial effects that: a plurality of original time sequence signals are filtered and dimensionally preprocessed to obtain a plurality of characteristic time sequence signals, the calculation amount of data is reduced, and the degree of noise interference is reduced; each characteristic time sequence signal is divided into a front signal sequence and a rear signal sequence; an error between an initial prediction sequence obtained by prediction according to the front signal sequence and the rear signal sequence is obtained to obtain an error value of each characteristic time sequence signal; error values of a plurality of characteristic time sequence signals are obtained, which are combined to form a sequence, denoted as an error value sequence; according to the error value sequence, a plurality of monitoring signals are screened out from all characteristic time sequence signals, the dimension is reduced again through analysis of different signals, so as to reduce the calculation amount of data; each monitoring signal is divided to obtain a signal sequence of each monitoring signal; the signal sequences of a plurality of monitoring signals are mapped in a monitoring characteristic space to obtain reference points; prediction of data corresponding to a plurality of monitoring signals at each moment is performed through the signal sequences of a plurality of monitoring signals to obtain an analysis point; according to the distribution of all reference points and analysis points in the monitoring characteristic space, the device fault of the steam turbine generator is prevented, abnormal analysis is performed through the difference between the distribution of collected data and the distribution of predicted data, abnormal result analysis of the predicted data is performed through the abnormal analysis result, so as to prevent the device fault. BRIEF DESCRIPTION OF DRAWINGS
[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative effort.
[0018] Figure 1 A step flowchart diagram of a device fault prevention method is provided for the present application. Figure 2 A module flowchart diagram of a device fault prevention system is provided for the present application. DETAILED DESCRIPTION
[0019] In the following, the technical solutions in the embodiments of the present application will be described clearly and completely in conjunction with the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all the other embodiments obtained by a person of ordinary skill in the art without creative effort should belong to the scope of protection of the present application.
[0020] It should be noted that the terms "first", "second", and the like in the specification and claims of the present application and the above-described accompanying drawings are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or device including a series of steps or units does not have to be limited to only those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to the process, method, product, or device.
[0021] In view of the problems in the background art, a device fault prevention method and system are designed, which has important practical significance.
[0022] As shown in Figure 1 The first aspect of the present application provides a device fault prevention method, comprising the following steps: Step S001: Collecting several kinds of original time sequence signals of a steam turbine generator in a working process.
[0023] It should be noted that in order to prevent the steam turbine generator from appearing abnormal in the working process, the system will be paralyzed, and some damage will occur to various devices in the steam turbine generator. Therefore, in order to avoid the steam turbine generator from appearing abnormal in the working process, it is necessary to collect various data in the history for prediction analysis, and to prevent abnormal conditions of the steam turbine generator in the working process through the prediction data.
[0024] Specifically, several kinds of original time sequence signals of the steam turbine generator in the working process are obtained within a preset time length hours before the current time; wherein the several kinds of original time sequence signals include vibration data signals, current data signals, voltage data signals, power data signals, bearing temperature data signals, exciter temperature data signals, stator and rotor temperature data signals, etc. In this embodiment, the preset time length In this embodiment, the preset time length Without specific limitation, the implementer can be determined according to the specific situation.
[0025] Among them, the vibration data signal is collected by the speed sensor, the current data signal is collected by the current transformer, the voltage data signal is collected by the voltage transformer, the power data signal is collected by the power analyzer, the bearing temperature data signal, the exciter temperature data signal, the stator and rotor temperature data signal are collected by the temperature sensor.
[0026] At this point, several kinds of original time sequence signals of the steam turbine generator in the working process are obtained.
[0027] Step S002: filtering and dimensionality reduction preprocessing are performed on the several kinds of original time sequence signals to obtain several kinds of characteristic time sequence signals.
[0028] It should be noted that, since the state of the device is changing during operation, the Kalman filtering algorithm is used to filter the collected various time sequence data.
[0029] Specifically, the Kalman filtering algorithm is used to filter each original time sequence signal to obtain each filtered time sequence signal.
[0030] It should be further noted that, in order to reduce the computer's calculation amount, the main signals in all filtered time sequence signals are extracted by dimensionality reduction; the main signals can not only reflect the device fault information, but also reduce the calculation amount.
[0031] Specifically, several main signals are extracted from all filtered time sequence signals by DPCA (Dynamic Principal Component Analysis) algorithm, and the several main signals are denoted as several characteristic time sequence signals.
[0032] Among them, the Kalman filtering algorithm and the DPCA algorithm are known technologies, which will not be described in detail here.
[0033] At this point, several kinds of characteristic time sequence signals are obtained.
[0034] Step S003: each characteristic time sequence signal is divided into a front signal sequence and a rear signal sequence; the error between the initial prediction sequence obtained by predicting according to the front signal sequence and the rear signal sequence is obtained to obtain the error value of each characteristic time sequence signal; the error values of several characteristic time sequence signals are obtained to form a sequence, denoted as an error value sequence; and several monitoring signals are screened from all characteristic time sequence signals according to the error value sequence.
[0035] It should be noted that in order to analyze the data processing and the prediction effect after adjustment, each characteristic time series signal is segmented, the previous data segment is predicted, and the difference between the prediction result and the subsequent data segment is analyzed.
[0036] Specifically, each characteristic time series signal is divided at a preset time interval to obtain a plurality of data corresponding to each characteristic time series signal after division. According to the time sequence, the plurality of data is grouped into a sequence, which is denoted as each time series signal sequence. In this embodiment, the preset time interval is 1 second, and the preset time interval is not specifically limited and can be determined by the implementer according to the specific situation.
[0037] The data in the front of each time series signal sequence is grouped into a front signal sequence, and the data in the rear of each time series signal sequence is grouped into a rear signal sequence. In this embodiment, the preset division percentage factor is 0.5.
[0038] In this embodiment, the preset division percentage factor is 0.5, and the preset division percentage factor is not specifically limited and can be determined by the implementer according to the specific situation.
[0039] The front signal sequence is input into a GRU neural network for prediction to obtain an initial prediction sequence. The initial prediction sequence obtained by prediction has the same number of data as the rear signal sequence.
[0040] The error between the initial prediction sequence and the rear signal sequence is calculated by MAPE (Mean Absolute Percentage Error), which is denoted as the error value W of each characteristic time series signal. The GRU neural network and the MAPE are both known technologies, and will not be described in detail here.
[0041] The error values of all characteristic time series signals are calculated, and the error values of all characteristic time series signals are sorted in ascending order to obtain an error value sequence. The characteristic time series signal corresponding to the error value in the front of the error value sequence is denoted as a monitoring signal data. In this embodiment, the preset division factor is 0.5. In this embodiment, the preset division factor is 0.5, and the preset division factor is not specifically limited and can be determined by the implementer according to the specific situation.
[0042] At this point, a plurality of monitoring signals are obtained.
[0043] Step S004: dividing each monitoring signal to obtain a signal sequence of each monitoring signal; mapping the signal sequences of the several monitoring signals in a monitoring feature space to obtain reference points; predicting the data corresponding to the several monitoring signals at each subsequent moment through the signal sequences of the several monitoring signals to obtain an analysis point; and preventing equipment failure of the steam turbine generator according to the distribution of all the reference points and the analysis point in the monitoring feature space.
[0044] It should be noted that the difference between the distribution of the predicted monitoring signal data at the subsequent moment and the distribution of all the historical monitoring signal data is used to determine the failure condition of the equipment at the subsequent moment.
[0045] Specifically, each monitoring signal is taken as a dimension to construct the monitoring feature space. By presetting a time interval each monitoring signal is divided into several data, and the data of each monitoring signal is sequentially arranged to form a signal sequence of each monitoring signal; and all the data in the signal sequences of all the monitoring signals are mapped in the monitoring feature space to obtain several data points, which are denoted as reference points; wherein all the data at the same moment in the signal sequences of all the monitoring signals are mapped in the monitoring feature space as a data point.
[0046] According to the signal sequence of each monitoring signal, the GRU neural network is used to predict the monitoring signal data of each monitoring signal at the subsequent moments, to obtain the monitoring signal data of each monitoring signal at the subsequent moments; all the monitoring signal data at each subsequent moment are obtained, and the data at each subsequent moment are mapped in the monitoring feature space to obtain a data point corresponding to each subsequent moment; and the data points of all the subsequent moments mapped in the feature space are denoted as analysis points.
[0047] It should be noted that the reference points in the monitoring feature space represent normal data points, and the analysis points are data points at the subsequent moments, so the distribution difference between all the reference points and the analysis points in the monitoring feature space is analyzed to determine the possibility of equipment failure at the subsequent moment; the greater the distribution difference between all the reference points and the analysis points, the greater the distance between the analysis points at the subsequent moment and the normal reference points, and the greater the possibility of equipment failure at the subsequent moment; and the smaller the distribution difference between all the reference points and the analysis points, the smaller the distance between the analysis points at the subsequent moment and the normal reference points, and the smaller the possibility of equipment failure at the subsequent moment.
[0048] Specifically, the local outlier factor (LOF) algorithm is used to calculate the outlier factor of each to-be-analyzed point according to all reference points and all to-be-analyzed points in the feature space, and then the outlier factors of all to-be-analyzed points are normalized to obtain the normalized outlier factor value of each to-be-analyzed point. In this embodiment, the linear normalization method is used for normalization.
[0049] The LOF algorithm is a known technology, and will not be described in detail here.
[0050] When the normalized outlier factor value of each to-be-analyzed point at a subsequent moment is greater than or equal to the preset outlier threshold , the background is alarmed, and the staff is notified to check, so as to prevent the occurrence of equipment failure.
[0051] In this embodiment, the preset outlier threshold is not specifically limited, and can be determined by the implementer according to the specific situation.
[0052] As shown in Figure 2 , the second aspect of the present application provides a device failure prevention system, comprising the following modules: The data acquisition module 101 is used to acquire several kinds of original time sequence signals of the steam turbine generator during the working process. The preprocessing module 102 is used to filter and reduce the dimension of the several kinds of original time sequence signals to obtain several kinds of feature time sequence signals. The feature dimension screening module 103 is used to divide each feature time sequence signal to obtain several data, and obtain each time sequence signal sequence through the several data. Each time sequence signal sequence is divided into a front signal sequence and a rear signal sequence. According to the front signal sequence, an initial prediction sequence is obtained, and the error value of each feature time sequence signal is obtained through the error between the initial prediction sequence and the rear signal sequence. The error values of the several feature time sequence signals are obtained to form an error value sequence. According to the error value sequence, several monitoring signals are screened from all feature time sequence signals. The fault prevention module 104 is configured to construct a monitoring feature space by using the monitoring signals, divide each monitoring signal to obtain a signal sequence of each monitoring signal, map the signal sequences of the monitoring signals in the monitoring feature space to obtain a plurality of data points, which are denoted as reference points, predict the data corresponding to the monitoring signals at each subsequent moment by using the signal sequences of the monitoring signals, obtain the monitoring signal data at each subsequent moment, map the monitoring signal data at each subsequent moment in the monitoring feature space to obtain a to-be-analyzed point corresponding to each subsequent moment, and prevent the equipment fault of the turbogenerator according to the distribution of all the reference points and the to-be-analyzed points in the monitoring feature space.
[0053] The third aspect of the present application provides an electronic device, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the method for preventing the equipment fault when executing the computer program.
[0054] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can adopt a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, optical storage, etc.) containing computer usable program codes.
[0055] The present application is described with reference to flowcharts and / or block diagrams of the method, the system, and the computer program product according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one or more flows and / or blocks. Figure 1 The means for implementing the functions specified in one or more flows and / or blocks.
[0056] These computer program instructions can also be stored in a computer readable storage medium capable of guiding the computer or other programmable data processing devices to work in a specific manner, so that the instructions stored in the computer readable storage medium produce a product including instruction means, which implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one or more flows and / or blocks. Figure 1 The means for implementing the functions specified in one or more flows and / or blocks.
[0057] These computer program instructions can also be loaded into a computer or other programmable data processing devices, so that a series of operational steps are generated to realize the computer-implemented processes in the computer or other programmable devices, and the instructions executed in the computer or other programmable devices provide the processes for implementing the functions specified in the flowchart Figure 1 one flow or multiple flows and / or the functions specified in the block Figure 1 one flow or multiple flows and / or the functions specified in the block
[0058] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application, but not to limit it, although the above embodiments of the present application have been described in detail, those skilled in the art should understand: the specific embodiments of the present application can be modified or replaced by the same, without departing from the spirit and scope of the present application, any modification or equivalent replacement, which should be covered within the scope of protection of the present application.
Claims
1. A method for preventing equipment failure, characterized in that: include: Obtaining several original timing signals of the steam turbine generator during operation; Filter and reduce the dimension of several original time series signals to obtain several characteristic time series signals; Each characteristic time series signal is divided to obtain a number of data, and each time series signal sequence is obtained through the number of data; each time series signal sequence is divided into a front signal sequence and a back signal sequence; prediction is performed based on the front signal sequence to obtain a set of initial prediction sequences, and the error value of each characteristic time series signal is obtained based on the error between the initial prediction sequence and the back signal sequence; the error values of the plurality of characteristic time series signals are obtained and grouped into a set of sequences, recorded as an error value sequence; and a plurality of monitoring signals are screened from all characteristic time series signals based on the error value sequence; A monitoring feature space is constructed through several monitoring signals, and each monitoring signal is divided to obtain a signal sequence of each monitoring signal; the signal sequences of the several monitoring signals are mapped in the monitoring feature space to obtain several data points, which are recorded as reference points; the corresponding data of the several monitoring signals at each subsequent moment are predicted through the signal sequences of the several monitoring signals to obtain several monitoring signal data at each subsequent moment; the several monitoring signal data at each subsequent moment are mapped in the monitoring feature space to obtain a point to be analyzed corresponding to each subsequent moment; and equipment failure of the steam turbine generator is prevented based on the distribution of all reference points and points to be analyzed in the monitoring feature space.
2. A method for preventing equipment failure according to claim 1, characterized in that: The filtering and dimensionality reduction preprocessing of the multiple original time series signals are performed to obtain multiple characteristic time series signals, including: Use the Kalman filter algorithm to filter each original time series signal to obtain each filtered time series signal, which is recorded as each filtered time series signal; A DPCA algorithm is used to reduce the dimension of all filtered time series signals to extract several main signals, and the several main signals are recorded as several characteristic time series signals.
3. The method for preventing equipment failure according to claim 1, characterized in that: The method of dividing each characteristic time series signal to obtain a plurality of data, and obtaining each time series signal sequence through the plurality of data; and dividing each time series signal sequence into a front signal sequence and a rear signal sequence, comprises: At preset time intervals To divide each characteristic time series signal, obtain a number of data corresponding to each characteristic time series signal after division, and group the data into a set of sequences in chronological order, which are recorded as each time series signal sequence; The first The data is composed of the front signal sequence; the back signal sequence of each time series signal sequence The data is composed of, followed by a signal sequence; among them, Divide the preset into percentage factors.
4. The method for preventing equipment failure according to claim 1, characterized in that: The method of performing prediction based on the previous signal sequence to obtain a set of initial prediction sequences, and obtaining the error value of each characteristic time series signal through the error between the initial prediction sequence and the subsequent signal sequence, includes: Put the previous signal sequence into the GRU neural network for prediction to obtain a set of initial prediction sequences; Through MAPE, the error between the initial prediction sequence and the subsequent signal sequence is calculated and recorded as the error value of each characteristic time series signal.
5. A method for preventing equipment failure according to claim 4, characterized in that: The error values of the characteristic time series signals are obtained and grouped into a set of sequences, which are recorded as error value sequences; According to the error value sequence, several monitoring signals are screened out from all characteristic time series signals, including: Calculate the error values of all characteristic time series signals, sort the error values of all characteristic time series signals in ascending order, and obtain an error value sequence; The error value sequence The time series signal data corresponding to the error value is recorded as monitoring signal data; Represents the preset distinguishing factor.
6. The method for preventing equipment failure according to claim 1, characterized in that: The monitoring feature space is constructed by using several monitoring signals, each monitoring signal is divided to obtain a signal sequence of each monitoring signal; the signal sequences of the several monitoring signals are mapped in the monitoring feature space to obtain several data points, which are recorded as reference points, including: Each monitoring signal is used as a dimension to construct the monitoring feature space; By pre-set time interval Each monitoring signal is divided into several data, and a signal sequence of each monitoring signal is formed in chronological order; all data in the signal sequence of all monitoring signals are mapped into the monitoring feature space to obtain several data points, and the several data points are recorded as reference points.
7. A method for preventing equipment failure according to claim 6, characterized in that: The method includes predicting corresponding data of several monitoring signals at each subsequent moment through the signal sequences of several monitoring signals to obtain data of several monitoring signals at each subsequent moment; mapping the data of several monitoring signals at each subsequent moment in the monitoring feature space to obtain a point to be analyzed corresponding to each subsequent moment, including: According to the signal sequence of each monitoring signal, a GRU neural network is used to predict each monitoring signal at several subsequent moments to obtain monitoring signal data of each monitoring signal at several subsequent moments; Obtain all monitoring signal data at each subsequent moment, map all monitoring signal data at each subsequent moment in the monitoring feature space, and obtain a data point corresponding to each subsequent moment; record the data points mapped in the feature space at all subsequent moments as points to be analyzed.
8. The method for preventing equipment failure according to claim 1, characterized in that: The prevention of equipment failure of the steam turbine generator according to the distribution of all reference points and points to be analyzed in the monitoring feature space includes: The outlier factor of each point to be analyzed is calculated using the LOF algorithm based on all reference points and all points to be analyzed in the monitoring feature space. Then, the outlier factors of all points to be analyzed are normalized to obtain the normalized outlier factor value of each point to be analyzed. When the normalized outlier factor value of the point to be analyzed at each subsequent moment is greater than or equal to the preset outlier threshold When an error occurs, an alarm will be sounded in the background and the staff will be notified to conduct an inspection to prevent equipment failure.
9. A system for preventing equipment failure, characterized in that: include: Data acquisition module: used to obtain several original timing signals of the steam turbine generator during operation; Preprocessing module: used to filter and reduce the dimension of several original time series signals to obtain several characteristic time series signals; Feature dimension screening module: used to divide each characteristic time series signal to obtain a number of data, and obtain each time series signal sequence through the said data; divide each time series signal sequence into a front signal sequence and a back signal sequence; predict based on the front signal sequence to obtain a set of initial prediction sequences, and obtain the error value of each characteristic time series signal through the error between the initial prediction sequence and the back signal sequence; obtain the error values of several characteristic time series signals, form them into a set of sequences, and record them as error value sequences; based on the error value sequences, screen out several monitoring signals from all characteristic time series signals; Fault prevention module: used to construct a monitoring feature space through several monitoring signals, divide each monitoring signal, and obtain a signal sequence for each monitoring signal; map the signal sequences of several monitoring signals in the monitoring feature space to obtain several data points, which are recorded as reference points; predict the corresponding data of several monitoring signals at each subsequent moment through the signal sequences of several monitoring signals, and obtain several monitoring signal data at each subsequent moment; map the several monitoring signal data at each subsequent moment in the monitoring feature space to obtain a point to be analyzed corresponding to each subsequent moment; prevent equipment failure of the steam turbine generator based on the distribution of all reference points and points to be analyzed in the monitoring feature space.
10. An electronic device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for preventing equipment failure according to any one of claims 1 to 8 is implemented.