CN113848417A - Rail transit power supply equipment fault prediction method and device - Google Patents
Rail transit power supply equipment fault prediction method and device Download PDFInfo
- Publication number
- CN113848417A CN113848417A CN202110911227.1A CN202110911227A CN113848417A CN 113848417 A CN113848417 A CN 113848417A CN 202110911227 A CN202110911227 A CN 202110911227A CN 113848417 A CN113848417 A CN 113848417A
- Authority
- CN
- China
- Prior art keywords
- power supply
- data
- equipment
- supply equipment
- model
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 66
- 238000012545 processing Methods 0.000 claims abstract description 13
- 230000006870 function Effects 0.000 claims description 24
- 238000013145 classification model Methods 0.000 claims description 20
- 238000012549 training Methods 0.000 claims description 20
- 238000010606 normalization Methods 0.000 claims description 11
- 238000000354 decomposition reaction Methods 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 9
- 210000002569 neuron Anatomy 0.000 claims description 8
- 238000004458 analytical method Methods 0.000 claims description 7
- 238000001514 detection method Methods 0.000 claims description 7
- 238000010276 construction Methods 0.000 claims description 6
- 238000003062 neural network model Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 239000006185 dispersion Substances 0.000 claims description 2
- 238000012706 support-vector machine Methods 0.000 abstract description 40
- 230000007787 long-term memory Effects 0.000 abstract description 9
- 230000006403 short-term memory Effects 0.000 abstract description 8
- 230000007774 longterm Effects 0.000 abstract description 7
- 238000013528 artificial neural network Methods 0.000 description 15
- 238000010586 diagram Methods 0.000 description 11
- 230000007547 defect Effects 0.000 description 8
- 238000012423 maintenance Methods 0.000 description 8
- 238000003745 diagnosis Methods 0.000 description 7
- 238000007781 pre-processing Methods 0.000 description 5
- 230000015654 memory Effects 0.000 description 4
- 210000004027 cell Anatomy 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000009471 action Effects 0.000 description 2
- 230000004913 activation Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000007405 data analysis Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 230000002194 synthesizing effect Effects 0.000 description 2
- 230000009897 systematic effect Effects 0.000 description 2
- 238000007418 data mining Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 230000001939 inductive effect Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/08—Locating faults in cables, transmission lines, or networks
- G01R31/088—Aspects of digital computing
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Supply And Distribution Of Alternating Current (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method and a device for predicting faults of rail transit power supply equipment. Firstly, generating a power supply system fault prediction model by adopting an LSTM algorithm and system equipment historical data, then carrying out denoising and standardization processing on the current data of the system equipment, then inputting the processed data into the prediction model to obtain the prediction data of target equipment, finally classifying the data output by the prediction model by using an SVM model, and outputting an equipment state result. The method combines the advantages that the long-term and short-term memory model can carry out long-term prediction and the support vector machine has good nonlinear classification, firstly adopts the LSTM + SVM model to predict the fault of the power supply system, establishes a method for accurately predicting the fault for the rail transit power supply system, and can effectively improve the stability and the safety of the rail transit power supply system.
Description
Technical Field
The invention relates to the field of fault prediction of rail transit power systems, in particular to a fault prediction method for rail transit power supply equipment based on an LSTM neural network model.
Background
With the rapid development of high-speed rail and urban rail transit in China, the scale of trains and the frequency of train transportation are continuously improved, the damage of the fault of a power supply system to social economy and personal safety is more and more serious, and the rail transit power supply system faces a serious challenge. The method is an effective method for avoiding the power supply system from being out of order and further avoiding bringing a series of serious consequences. The method for predicting the fault of the rail transit power supply equipment with high reliability and high speed prediction is researched, and has very important significance for guaranteeing the safety and the economy of the whole rail transit power supply system.
The former people do a lot of research on the rail transit power supply system equipment fault diagnosis method, but do not predict equipment future data and then accurately classify equipment states. In the prior art, fault diagnosis is performed on power supply system equipment by combining the technologies of an artificial neural network, a Bayesian network, an expert system, data mining and the like in the field of artificial intelligence. The method in the prior art is combined with an artificial intelligence technology to intelligently diagnose the occurred faults by inducing expert experience, can carry out high-dimensional classification on fault data, and has higher accuracy. However, the method can only analyze the type and the cause of the fault, cannot predict the fault, and belongs to a mode of diagnosis after the fault.
Disclosure of Invention
Aiming at the defects in the prior art, the technical problems to be solved by the invention are as follows: the rail transit power supply system cannot accurately predict before the power supply equipment fails to work so as to avoid the occurrence of power supply accidents.
In order to solve the above problems, the present invention is realized by the following technical solutions:
the invention provides a rail transit power supply equipment fault prediction method, which comprises the following steps:
s1, obtaining power supply equipment data of the target power supply equipment at the time t, wherein the power supply equipment data comprise voltage, current, power and equipment operating temperature;
s2, denoising the power supply equipment data to obtain an equipment data set without interference elements, and standardizing the equipment data set to obtain input parameters at the time t;
s3, inputting the input parameters into a power supply equipment data LSTM prediction model with a prediction time step length of n, and outputting the target equipment operation parameter prediction value at the t + n moment;
s4, performing data classification on the predicted value as an input value of the SVM model, analyzing whether the equipment fails at the moment of t + n, and if not, returning to the step S1 for circular detection;
s5: if yes, the target power supply equipment is indicated to have the possibility of failure.
Further, the constructing step of the LSTM prediction model includes:
setting an LSTM input dimension and a prediction time step;
setting a model optimizer and a learning rate;
setting the number of hidden layer neuron nodes;
importing the target equipment operation historical data set into an LSTM neural network model for training;
and continuously optimizing and adjusting the hyper-parameters of the LSTM model in the training process to obtain an optimized LSTM prediction model.
Further, the construction step of the SVM classification model comprises the following steps:
selecting normal operation parameters and fault operation parameters in the historical data of the target equipment and combining the normal operation parameters and the fault operation parameters to form a data set;
carrying out standardization processing on the data set, wherein the processed data set is used as input data during training of an SVM model;
training the constructed SVM model, wherein the training process comprises the following steps: and adjusting a kernel function of the SVM, adjusting class weight, adjusting iteration time parameters, and optimizing the performance of the classification model to obtain the optimized SVM classification model.
Further, the value of the acquired target device data is influenced by the type of the target power supply device and/or the role played by the target device in the power supply system.
The method for denoising the data in step S2 is a wavelet denoising method.
The wavelet denoising method comprises the following steps:
constructing a function space, decomposing the signals into the function space for calculation, and acquiring useful data;
reconstructing the returned original signal, wherein the data decomposition and reconstruction formula is as follows:
decomposition formula: a. the0[f(t)]＝f(t)
Reconstructing a formula:
in the formula: t is a time sequence, f (t) is an original signal, j is a decomposed layer number, H, G is a wavelet decomposition filter in a time domain, h and g are wavelet reconstruction filters in a time domain, AjWavelet coefficients of the low-frequency part of the signal f (t) at the j-th level, DjIs the wavelet coefficient of the high frequency part of the signal f (t) at the j-th level.
Further, the normalization processing in step S2 is performed by a dispersion normalization method, which is established by the following formula:
wherein,
is the minimum value of the samples and is,
is the maximum value of the sample.
In another aspect of the present invention, a rail transit power supply equipment fault prediction apparatus is provided, and the rail transit power supply equipment fault prediction apparatus includes:
the data acquisition unit is used for acquiring power supply equipment data of target power supply equipment at time t, wherein the power supply equipment data comprises voltage, current, power and equipment operating temperature;
the de-noising standardization unit is used for de-noising the power supply equipment data to obtain an equipment data set without interference elements, and standardizing the equipment data set to obtain input parameters at the time t;
the prediction unit is used for inputting the input parameters into a power supply equipment data LSTM prediction model with a prediction time step length of n and outputting the target equipment operation parameter prediction value at the t + n moment;
the analysis unit is used for carrying out data classification on the predicted value serving as an input value of the SVM model and analyzing whether the equipment fails at the moment of t + n;
and the first execution unit is used for executing the circular detection returned to the data processing unit when the device does not have a fault at the moment of analyzing t + n.
And the second execution unit is used for showing that the target power supply equipment has the possibility of failure when the equipment fails at the moment of analyzing t + n.
The invention has the technical effects that:
the invention combines an LSTM prediction model with an SVM classification model; the LSTM prediction model input dimensions comprise: data of voltage, current, power and temperature of the target power supply equipment after preprocessing; the output dimension of the LSTM prediction model is predicted power supply equipment data; the input dimensionality of the SVM classification model is prediction data output by an LSTM prediction model; the output dimensionality of the SVM model is the state of the target power supply equipment, and the state comprises three states of normal, maintenance and fault. The power supply equipment fault prediction method adopted by the invention carries out state classification after predicting the future parameters of the target equipment, and evaluates the future state of the equipment. The data preprocessing algorithm preprocesses the parameter data of the equipment at the current moment to obtain standardized data; the prediction model outputs predicted data at the next moment by preprocessing the standardized data at the current moment and then taking the preprocessed data as input for analysis; further, the classification model takes the output of the prediction model as input, evaluates the state of the target equipment at the future moment by performing state classification on input data, and performs the next action on the equipment by outputting three states of normal, maintenance and fault of an evaluation result; if the equipment is predicted to be in a normal state, normal operation can be continued; if the equipment is in the maintenance state, the equipment is in the fault edge state, and maintenance and inspection are needed recently; when the equipment state is predicted to be a fault, the equipment needs to be shut down and maintained, so that electric power accidents are avoided, and the safety of people, vehicles and a power supply system is guaranteed. The invention aims to provide a heat supply pipeline leakage detection device based on machine vision, which can automatically identify the damage condition on a heat supply pipeline, timely make precautionary measures and inform workers of solving the heat supply pipeline with related defects.
Drawings
Other features, objects and advantages of the invention will become more apparent upon reading of the detailed description of non-limiting embodiments with reference to the following drawings:
FIG. 1 is a flow chart of a method for predicting a fault of a rail transit power supply apparatus according to the present invention;
FIG. 2 is a system flow diagram of a power supply system fault prediction incorporating an LSTM model and an SVM model in accordance with an embodiment of the present invention;
FIG. 3 is a schematic diagram of the structure of the input layer, the hidden layer and the output layer of the neural network according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of a recurrent neural network architecture in accordance with an embodiment of the present invention;
FIG. 5 is a diagram illustrating a structure of a long term memory model and a short term memory model according to an embodiment of the present invention;
FIG. 6 is a model diagram of the internal data processing method of the long-term and short-term memory model according to an embodiment of the present invention;
FIG. 7 is a diagram of actual prediction results for a fault using a model in an embodiment of the present invention.
Detailed Description
The present invention will be described in detail with reference to specific examples. The following examples will assist those skilled in the art in further understanding the invention, but are not intended to limit the invention in any way. It should be noted that it would be obvious to those skilled in the art that various changes and modifications can be made without departing from the spirit of the invention. All falling within the scope of the present invention.
As shown in fig. 1, a flowchart of a method for predicting a fault of a rail transit power supply apparatus provided by the present invention is provided, and the method for predicting a fault of a rail transit power supply apparatus provided by the present invention includes the following steps:
s1, obtaining power supply equipment data of the target power supply equipment at the time t, wherein the power supply equipment data comprise voltage, current, power and equipment operating temperature;
s2, denoising the power supply equipment data to obtain an equipment data set without interference elements, and standardizing the equipment data set to obtain input parameters at the time t;
s3, inputting the input parameters into a power supply equipment data LSTM prediction model with a prediction time step length of n, and outputting the target equipment operation parameter prediction value at the t + n moment;
and S4, performing data classification on the predicted value as an input value of the SVM model, analyzing whether the equipment has faults at the time of t + n, and if not, returning to the step S1 for circular detection.
S5: if yes, the target power supply equipment is indicated to have the possibility of failure.
FIG. 2 is a system flow diagram of the present invention incorporating an LSTM model and an SVM model for power supply system fault prediction; FIG. 3 is a block diagram of a neural artificial neural network, including an input layer, a hidden layer, and an output layer; FIG. 4 is a diagram of an RNN model that can also predict future state data from previous state data, but can only remember information from cell states, and therefore can only deal with short term dependency problems; fig. 5 is a schematic structural diagram of a long-short term memory model, fig. 6 is a model diagram of a data processing mode in the long-short term memory model, and the LSTM introduces a sigmoid function through an input gate, a forgetting gate and an output gate, combines a tanh function, adds summation operation, reduces the possibility of gradient disappearance and gradient explosion, and can process both short-term dependence problems and long-term dependence problems; fig. 7 shows the actual prediction result of the fault of the model, and after the system is trained by historical data, the state of the rail transit power supply equipment can be accurately predicted.
In an embodiment of the present invention, further, the denoising processing on the acquired device data is a wavelet denoising method, the method is divided into a data decomposition step and a data reconstruction step, and the formula is as follows:
decomposition formula: a. the0[f(t)]＝f(t)
Reconstructing a formula:
in the formula: t is a time sequence, f (t) is an original signal, j is a decomposed layer number, H, G is a wavelet decomposition filter in a time domain, h and g are wavelet reconstruction filters in a time domain, AjWavelet coefficients of the low-frequency part of the signal f (t) at the j-th level, DjIs the wavelet coefficient of the high frequency part of the signal f (t) at the j-th level.
Further, the step S3 of normalizing the denoised device data is established by using the following formula:
wherein,
is the minimum value of the samples and is,
is the maximum value of the sample.
The method for predicting the fault of the rail transit power supply equipment combines an LSTM prediction model with an SVM classification model; the LSTM prediction model input dimensions include: data of voltage, current, power and temperature of the target power supply equipment after preprocessing; the output dimension of the LSTM prediction model is predicted power supply equipment data; the input dimensionality of the SVM classification model is prediction data output by an LSTM prediction model; the output dimension of the SVM model is the state of the target power supply equipment, and the state comprises three states of normal state, maintenance state and fault state.
The power supply equipment fault prediction method adopted by the invention carries out state classification after predicting the future parameters of the target equipment, and evaluates the future state of the equipment. The data preprocessing algorithm preprocesses the parameter data of the equipment at the current moment to obtain standardized data; the prediction model outputs the predicted data of the next moment by taking the preprocessed standardized data of the current moment as input analysis; further, the classification model takes the output of the prediction model as input, evaluates the state of the target equipment at the future moment by performing state classification on input data, and performs the next action on the equipment by outputting three states of normal, maintenance and fault of an evaluation result; if the equipment is predicted to be in a normal state, normal operation can be continued; if the equipment is in the maintenance state, the equipment is in the fault edge state, and maintenance inspection is needed recently; when the equipment state is predicted to be a fault, the equipment needs to be shut down and maintained, so that the power accident is avoided, and the safety of people, vehicles and a power supply system is guaranteed.
In another aspect, the present invention further provides a rail transit power supply equipment prediction apparatus, where the rail transit power supply equipment fault prediction apparatus includes:
the data acquisition unit is used for acquiring power supply equipment data of target power supply equipment at time t, wherein the power supply equipment data comprises voltage, current, power and equipment operating temperature;
the de-noising standardization unit is used for de-noising the power supply equipment data to obtain an equipment data set without interference elements, and standardizing the equipment data set to obtain input parameters at the time t;
the prediction unit is used for inputting the input parameters into a power supply equipment data LSTM prediction model with a prediction time step length of n and outputting the target equipment operation parameter prediction value at the t + n moment;
the analysis unit is used for carrying out data classification on the predicted value serving as an input value of the SVM model and analyzing whether the equipment fails at the moment of t + n;
and the first execution unit is used for executing the circular detection returned to the data processing unit when the device does not have a fault at the moment of analyzing t + n.
And the second execution unit is used for showing that the target power supply equipment has the possibility of failure when the equipment fails at the moment of analyzing t + n.
On the other hand, the invention also provides a track traffic power supply equipment prediction system which mainly comprises a power supply equipment parameter unit, a power supply equipment data denoising unit, a power supply equipment data normalization unit, a power supply equipment data prediction unit and a power supply equipment data operation state classification unit;
the power supply equipment parameter unit is used for acquiring the structural parameter information of the power supply equipment;
the power supply equipment data denoising unit removes interference factors in equipment data through a wavelet denoising method, and ensures the high correlation between the obtained data and target equipment;
the power supply equipment data normalization unit is used for performing normalization processing on the denoised data to avoid influencing the result of data analysis due to different data dimensions;
the power supply equipment data prediction unit is used for synthesizing the environment of the target equipment in the claim 3, and predicting the future operation parameters of the equipment by taking the real-time operation data of the equipment as input to obtain the prediction data of the target equipment.
And the power supply equipment data operation state classification unit is used for performing state classification on the power supply equipment prediction data and analyzing to obtain the future operation state of the target equipment.
The data prediction and state classification system for the rail transit power supply equipment can effectively overcome the defects of the conventional power supply system fault diagnosis method. The existing method for diagnosing the faults of the rail transit power supply system mainly analyzes and classifies equipment parameters generated after the faults occur and then diagnoses the faults. After the fault occurs, the diagnosis method has fault influence, so that not only can large-scale delay be caused to the line to cause economic loss, but also the safety of people and vehicles can be damaged.
The invention also provides a method for acquiring the LSTM model of the predicted target equipment data and the SVM model for carrying out state classification on the equipment data, which comprises the following steps:
the construction of the LSTM model for acquiring the predicted target device data is realized by the following steps:
a1: setting an LSTM input dimension and a prediction time step;
a2: setting a model optimizer and a learning rate;
a3: setting the number of hidden layer neuron nodes;
a4: importing a target device operation historical data set into an LSTM neural network model for training;
a5: and continuously optimizing and adjusting the hyper-parameters of the LSTM model in the training process to obtain an optimal model.
The neural network LSTM model in step a4 adopts a tanh function as an activation function, where the formula of the tanh function is:
wherein, x represents the component of the hidden layer neuron input feature vector, namely the data feature of the target device parameter after normalization; ω represents the weight of the input component; θ represents the threshold of the neuron.
The construction of the SVM model for state classification of the equipment data is realized by the following steps:
b1: selecting normal operation parameters and fault operation parameters in the historical data of the target equipment and combining the normal operation parameters and the fault operation parameters to form a data set;
b2: carrying out standardization processing on the formed data set, wherein the processed data set is used as input data during training of the SVM model;
b3: training the constructed SVM model, continuously adjusting parameters of the SVM such as kernel function, class weight, iteration times and the like in the process, and optimizing the performance of the classification model to obtain an optimal classification model.
The time series classification SVM model in the step B2 adopts a sigmoid function as a kernel function:
the method for acquiring the long-term and short-term memory network (LSTM) model of the predicted target equipment data is characterized in that the LSTM is an improvement of a circulating neural network, a channel structure of a cell state is added, the structure is more complex than that of the circulating neural network, the LSTM overcomes the defect that the circulating neural network cannot memorize the long-term state, the future state can be effectively predicted through long-term historical state analysis, the prediction result is more accurate, and the method is more suitable for prediction of the rail transit power supply equipment data.
The long-short term memory network (LSTM) is a deep network model capable of predicting unknown duration delay time sequences in deep learning. The fault prediction system needs various sensors to input a large amount of parameter information, aiming at the characteristics of large data volume and high data repetition rate, the forgetting stage and the selecting and memorizing stage of long-term and short-term memory can screen a large amount of parameters input in the fault prediction network, useful information is extracted, the information with high repeatability is ignored, and the calculation amount of the parameters is greatly reduced. After useful information in the time sequence is extracted, the model can analyze the incidence relation of parameters in the time sequence and predict the trend of time sequence data, and meanwhile, the defect that other neural networks cannot predict for a long time is overcome. The method for obtaining the Support Vector Machine (SVM) model for carrying out state classification on the equipment data is characterized in that the SVM model is an efficient and accurate classification model, the data of different input dimensions can be effectively classified by setting different types of kernel functions, and the generalization capability of the classification model can be adjusted by adjusting punishment parameters of the kernel functions. The embodiment integrates various neural network algorithms, establishes accurate steps, and finally can predict whether the power supply equipment is in a fault state in the future. In addition, model training is carried out by combining multi-dimensional input data to obtain predicted parameter data of the power supply equipment, and the predicted parameter data at the future time are classified by using a support vector machine model. The method combines the advantages that a long-term memory model can predict for a long time and a support vector machine has good nonlinear classification, adopts a systematic design, establishes an accurate fault prediction method for a rail transit power supply system, effectively improves the stability and the safety of the rail transit power supply system, and provides guarantee for the safety of passengers, trains and power supply systems.
The invention also provides a system for predicting and classifying the states of the rail transit power supply equipment, which mainly comprises a power supply equipment parameter unit, a power supply equipment data denoising unit, a power supply equipment data normalization unit, a power supply equipment data prediction unit and a power supply equipment data operation state classification unit; the power supply equipment parameter unit is used for acquiring structural parameter information of the power supply equipment; the power supply equipment data denoising unit removes interference factors in equipment data through a wavelet denoising method, and ensures the high correlation between the obtained data and target equipment; the power supply equipment data normalization unit is used for normalizing the denoised data and avoiding influencing the data analysis result due to different data dimensions;
the power supply equipment data prediction unit is used for synthesizing the environment of the target equipment and predicting the future operation parameters of the equipment by taking the real-time operation data of the equipment as input to obtain the prediction data of the target equipment.
And the power supply equipment data operation state classification unit is used for performing state classification on the power supply equipment prediction data and analyzing to obtain the future operation state of the target equipment.
The data prediction and state classification system for the rail transit power supply equipment can effectively overcome the defects of the conventional power supply system fault diagnosis method. The existing method for diagnosing the faults of the rail transit power supply system mainly analyzes and classifies equipment parameters generated after the faults occur and then diagnoses the faults. After the fault occurs, the diagnosis method has fault influence, so that not only can large-scale delay be caused to the line to cause economic loss, but also the safety of people and vehicles can be damaged.
The invention also provides a method for acquiring the LSTM model of the predicted target equipment data and the SVM model for carrying out state classification on the equipment data, which comprises the following steps:
the construction of the LSTM model for acquiring the predicted target device data is realized by the following steps:
a1: setting an LSTM input dimension and a prediction time step;
a2: setting a model optimizer and a learning rate;
a3: setting the number of hidden layer neuron nodes;
a4: importing a target device operation historical data set into an LSTM neural network model for training;
a5: and continuously optimizing and adjusting the hyper-parameters of the LSTM model in the training process to obtain an optimal model.
The neural network LSTM model in step a4 adopts a tanh function as an activation function, where the formula of the tanh function is:
wherein, x represents the component of the hidden layer neuron input feature vector, namely the data feature of the target device parameter after normalization; ω represents the weight of the input component; θ represents the threshold of the neuron.
The construction of the SVM model for state classification of the equipment data is realized by the following steps:
b1: selecting normal operation parameters and fault operation parameters in the historical data of the target equipment and combining the normal operation parameters and the fault operation parameters to form a data set;
b2: carrying out standardization processing on the formed data set, wherein the processed data set is used as input data during training of the SVM model;
b3: training the constructed SVM model, continuously adjusting parameters of the SVM such as kernel function, class weight, iteration times and the like in the process, and optimizing the performance of the classification model to obtain an optimal classification model.
The time series classification SVM model in the step B2 adopts a sigmoid function as a kernel function:
the method for acquiring the long-term and short-term memory network (LSTM) model of the predicted target equipment data is characterized in that the LSTM is an improvement of a circulating neural network, a channel structure of a cell state is added, the structure is more complex than that of the circulating neural network, the LSTM overcomes the defect that the circulating neural network cannot memorize the long-term state, the future state can be effectively predicted through long-term historical state analysis, the prediction result is more accurate, and the method is more suitable for prediction of the rail transit power supply equipment data.
The long-short term memory network (LSTM) is a deep network model capable of predicting unknown duration delay time sequences in deep learning. The fault prediction system needs various sensors to input a large amount of parameter information, aiming at the characteristics of large data volume and high data repetition rate, the forgetting stage and the selecting and memorizing stage of long-term and short-term memory can screen a large amount of parameters input in the fault prediction network, useful information is extracted, the information with high repeatability is ignored, and the calculation amount of the parameters is greatly reduced. After useful information in the time sequence is extracted, the model can analyze the incidence relation of parameters in the time sequence and predict the trend of time sequence data, and meanwhile, the defect that other neural networks cannot predict for a long time is overcome.
The method for obtaining the Support Vector Machine (SVM) model for carrying out state classification on the equipment data is characterized in that the SVM model is an efficient and accurate classification model, the data of different input dimensions can be effectively classified by setting different types of kernel functions, and the generalization capability of the classification model can be adjusted by adjusting punishment parameters of the kernel functions.
The method carries out model training by fusing multi-dimensional input data to obtain the predicted parameter data of the power supply equipment, and carries out equipment state classification on the predicted parameter data at the future time by using a support vector machine model. The method is combined with the advantages that a long-term prediction and support vector machine can be carried out by a long-term and short-term memory model, and the method has good nonlinear classification, adopts a systematic design, establishes an accurate fault prediction method for a rail transit power supply system, effectively improves the stability and safety of the rail transit power supply system, and provides guarantee for the safety of passengers, trains and power supply systems.
The above-mentioned embodiments are only specific embodiments of the present invention, not intended to limit the present invention, but to describe the objects, technical solutions and advantages of the present invention in further detail,
any modification, equivalent replacement, or improvement made without departing from the spirit and principle of the present invention shall fall within the protection scope of the present invention.
The foregoing description of specific embodiments of the present invention has been presented. It is to be understood that the present invention is not limited to the specific embodiments described above, and that various changes and modifications may be made by one skilled in the art within the scope of the appended claims without departing from the spirit of the invention.
Claims (9)
1. A rail transit power supply equipment fault prediction method is characterized by comprising the following steps:
s1, obtaining power supply equipment data of the target power supply equipment at the time t, wherein the power supply equipment data comprise voltage, current, power and equipment operating temperature;
s2, denoising the power supply equipment data to obtain an equipment data set without interference elements, and standardizing the equipment data set to obtain input parameters at the time t;
s3, inputting the input parameters into a power supply equipment data LSTM prediction model with a prediction time step length of n, and outputting the target equipment operation parameter prediction value at the t + n moment;
s4, performing data classification on the predicted value as an input value of the SVM model, analyzing whether the target power supply equipment fails at the moment of t + n, and if not, returning to the step S1 for circular detection;
and S5, if yes, indicating that the target power supply equipment has the possibility of failure.
2. The rail transit power supply equipment fault prediction method of claim 1, wherein the LSTM prediction model is constructed by the steps of:
setting an LSTM input dimension and a prediction time step;
setting a model optimizer and a learning rate;
setting the number of hidden layer neuron nodes;
importing the target equipment operation historical data set into an LSTM neural network model for training;
and continuously optimizing and adjusting the hyper-parameters of the LSTM model in the training process to obtain an optimized LSTM prediction model.
3. The rail transit power supply equipment fault prediction method according to claim 1, wherein the construction step of the SVM classification model comprises the following steps:
selecting normal operation parameters and fault operation parameters in the historical data of the target equipment and combining the normal operation parameters and the fault operation parameters to form a data set;
carrying out standardization processing on the data set, wherein the processed data set is used as input data during training of an SVM model;
training the constructed SVM model, wherein the training process comprises the following steps: and adjusting a kernel function of the SVM, adjusting class weight, adjusting iteration time parameters, and optimizing the performance of the classification model to obtain the optimized SVM classification model.
4. The rail transit power supply equipment fault prediction method according to claim 1, characterized by:
the value of the target device data obtained is influenced by the type of the target power supply device and/or the role the target device plays in the power supply system.
5. The rail transit power supply equipment fault prediction method according to claim 1, characterized by:
the value of the acquired target device data is affected by the location of the target device and/or the season in which the data was acquired.
6. The rail transit power supply equipment fault prediction method according to any one of claims 1 to 5, characterized by:
the method for denoising the data in step S2 is a wavelet denoising method.
7. The rail transit power supply equipment fault prediction method of claim 6, characterized by: the wavelet denoising method comprises the following steps:
constructing a function space, decomposing the signals into the function space for calculation, and acquiring useful data;
reconstructing the returned original signal, wherein the data decomposition and reconstruction formula is as follows:
decomposition formula: a. the0[f(t)]＝f(t)
Reconstructing a formula:
in the formula: t is a time sequence, f (t) is an original signal, j is a decomposed layer number, H, G is a wavelet decomposition filter in a time domain, h and g are wavelet reconstruction filters in a time domain, AjWavelet coefficients of the low-frequency part of the signal f (t) at the j-th level, DjIs the wavelet coefficient of the high frequency part of the signal f (t) at the j-th level.
8. The rail transit power supply equipment fault prediction method according to claim 1, characterized by: the normalization processing in step S2 is performed by dispersion normalization.
9. A rail transit power supply equipment fault prediction device is characterized by comprising:
the data acquisition unit is used for acquiring power supply equipment data of target power supply equipment at time t, wherein the power supply equipment data comprises voltage, current, power and equipment operating temperature;
the de-noising standardization unit is used for de-noising the power supply equipment data to obtain an equipment data set without interference elements, and standardizing the equipment data set to obtain input parameters at the time t;
the prediction unit is used for inputting the input parameters into a power supply equipment data LSTM prediction model with a prediction time step length of n and outputting the target equipment operation parameter prediction value at the t + n moment;
the analysis unit is used for carrying out data classification on the predicted value serving as an input value of the SVM model and analyzing whether the equipment fails at the moment of t + n;
the first execution unit is used for executing the circular detection returned to the data processing unit when the fact that the equipment does not have a fault at the moment of t + n is analyzed;
and the second execution unit is used for showing that the target power supply equipment has the possibility of failure when the equipment fails at the moment of analyzing t + n.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110911227.1A CN113848417A (en) | 2021-08-09 | 2021-08-09 | Rail transit power supply equipment fault prediction method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110911227.1A CN113848417A (en) | 2021-08-09 | 2021-08-09 | Rail transit power supply equipment fault prediction method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113848417A true CN113848417A (en) | 2021-12-28 |
Family
ID=78975696
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110911227.1A Pending CN113848417A (en) | 2021-08-09 | 2021-08-09 | Rail transit power supply equipment fault prediction method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113848417A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115037630A (en) * | 2022-04-29 | 2022-09-09 | 电子科技大学长三角研究院(湖州) | Weighted network link prediction method based on structural disturbance model |
| CN115063048A (en) * | 2022-08-16 | 2022-09-16 | 天津津轨汇海科技发展有限公司 | Intelligent management system for rail transit power supply operation and maintenance |
| CN115149653A (en) * | 2022-09-05 | 2022-10-04 | 天津津轨汇海科技发展有限公司 | Safety management system for rail transit power supply equipment |
| CN116340875A (en) * | 2023-05-30 | 2023-06-27 | 单县鑫和纺织有限公司 | Roving frame operation fault prediction system based on data analysis |
| CN116433226A (en) * | 2023-06-14 | 2023-07-14 | 浙江亿视电子技术有限公司 | Alternating current power supply equipment management method and system based on electrical parameter AI analysis |
| CN116662862A (en) * | 2023-06-05 | 2023-08-29 | 西安工程大学 | Transformer fault early warning method based on prediction and diagnosis of content of dissolved gas in oil |
| CN117516927A (en) * | 2024-01-05 | 2024-02-06 | 四川省机械研究设计院(集团)有限公司 | Gearbox fault detection method, system, equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108090606A (en) * | 2017-12-12 | 2018-05-29 | 上海应用技术大学 | Equipment fault finds method and system |
| CN109858345A (en) * | 2018-12-25 | 2019-06-07 | 华中科技大学 | A kind of intelligent failure diagnosis method suitable for pipe expanding equipment |
| CN110161343A (en) * | 2019-06-12 | 2019-08-23 | 中南大学 | A kind of non-intrusion type real-time dynamic monitoring method of intelligence train exterior power receiving device |
| CN111239672A (en) * | 2020-03-19 | 2020-06-05 | 东南大学 | Machine learning algorithm-based gradient fault prediction method for optical fiber current transformer |
| CN111368890A (en) * | 2020-02-26 | 2020-07-03 | 珠海格力电器股份有限公司 | Fault detection method and device and information physical fusion system |
-
2021
- 2021-08-09 CN CN202110911227.1A patent/CN113848417A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108090606A (en) * | 2017-12-12 | 2018-05-29 | 上海应用技术大学 | Equipment fault finds method and system |
| CN109858345A (en) * | 2018-12-25 | 2019-06-07 | 华中科技大学 | A kind of intelligent failure diagnosis method suitable for pipe expanding equipment |
| CN110161343A (en) * | 2019-06-12 | 2019-08-23 | 中南大学 | A kind of non-intrusion type real-time dynamic monitoring method of intelligence train exterior power receiving device |
| CN111368890A (en) * | 2020-02-26 | 2020-07-03 | 珠海格力电器股份有限公司 | Fault detection method and device and information physical fusion system |
| CN111239672A (en) * | 2020-03-19 | 2020-06-05 | 东南大学 | Machine learning algorithm-based gradient fault prediction method for optical fiber current transformer |
Non-Patent Citations (1)
| Title |
|---|
| 杨秀璋，颜娜: "Python 网络数据爬取及分析从入门到精通 分析篇", 北京航空航天大学出版社, pages: 147 - 148 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115037630A (en) * | 2022-04-29 | 2022-09-09 | 电子科技大学长三角研究院(湖州) | Weighted network link prediction method based on structural disturbance model |
| CN115037630B (en) * | 2022-04-29 | 2023-10-20 | 电子科技大学长三角研究院(湖州) | Weighted network link prediction method based on structure disturbance model |
| CN115063048A (en) * | 2022-08-16 | 2022-09-16 | 天津津轨汇海科技发展有限公司 | Intelligent management system for rail transit power supply operation and maintenance |
| CN115149653A (en) * | 2022-09-05 | 2022-10-04 | 天津津轨汇海科技发展有限公司 | Safety management system for rail transit power supply equipment |
| CN116340875A (en) * | 2023-05-30 | 2023-06-27 | 单县鑫和纺织有限公司 | Roving frame operation fault prediction system based on data analysis |
| CN116340875B (en) * | 2023-05-30 | 2023-10-17 | 单县鑫和纺织有限公司 | Roving frame operation fault prediction system based on data analysis |
| CN116662862A (en) * | 2023-06-05 | 2023-08-29 | 西安工程大学 | Transformer fault early warning method based on prediction and diagnosis of content of dissolved gas in oil |
| CN116433226A (en) * | 2023-06-14 | 2023-07-14 | 浙江亿视电子技术有限公司 | Alternating current power supply equipment management method and system based on electrical parameter AI analysis |
| CN117516927A (en) * | 2024-01-05 | 2024-02-06 | 四川省机械研究设计院(集团)有限公司 | Gearbox fault detection method, system, equipment and storage medium |
| CN117516927B (en) * | 2024-01-05 | 2024-04-05 | 四川省机械研究设计院(集团)有限公司 | Gearbox fault detection method, system, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113848417A (en) | Rail transit power supply equipment fault prediction method and device | |
| Li et al. | Data alignments in machinery remaining useful life prediction using deep adversarial neural networks | |
| Xue et al. | A novel intelligent fault diagnosis method of rolling bearing based on two-stream feature fusion convolutional neural network | |
| Yan et al. | Deep order-wavelet convolutional variational autoencoder for fault identification of rolling bearing under fluctuating speed conditions | |
| Lundgren et al. | Data-driven fault diagnosis analysis and open-set classification of time-series data | |
| Tang et al. | A wind turbine bearing fault diagnosis method based on fused depth features in time–frequency domain | |
| Jin et al. | A multi-scale convolutional neural network for bearing compound fault diagnosis under various noise conditions | |
| CN113642754B (en) | Complex industrial process fault prediction method based on RF noise reduction self-coding information reconstruction and time convolution network | |
| CN113610188B (en) | Bow net contact force non-section abnormality identification method and device | |
| Xiao et al. | Remaining useful life prediction based on intentional noise injection and feature reconstruction | |
| Lu et al. | Joint training of a predictor network and a generative adversarial network for time series forecasting: A case study of bearing prognostics | |
| Son et al. | Deep learning-based anomaly detection to classify inaccurate data and damaged condition of a cable-stayed bridge | |
| CN116610998A (en) | Switch cabinet fault diagnosis method and system based on multi-mode data fusion | |
| CN118154174B (en) | Intelligent operation and maintenance cloud platform for industrial equipment | |
| CN114926825A (en) | Vehicle driving behavior detection method based on space-time feature fusion | |
| CN116541790A (en) | New energy vehicle health assessment method and device based on multi-feature fusion | |
| Li et al. | Intelligent fault diagnosis of aeroengine sensors using improved pattern gradient spectrum entropy | |
| CN115098962A (en) | Method for predicting residual life of mechanical equipment in degradation state based on hidden half Markov model | |
| CN114781458A (en) | Power distribution network initial fault identification method based on time-frequency memory neural network of adaptive wavelet and attention mechanism | |
| Liu et al. | Air Pollutant Concentration Forecasting Using Long Short‐Term Memory Based on Wavelet Transform and Information Gain: A Case Study of Beijing | |
| CN118050591A (en) | Distribution network 10KV overhead line fault early warning and positioning system based on AI | |
| Hurtado et al. | A data-driven methodology for bridge indirect health monitoring using unsupervised computer vision | |
| Qin et al. | A novel physically interpretable end-to-end network for stress monitoring in laser shock peening | |
| CN116714437B (en) | Hydrogen fuel cell automobile safety monitoring system and monitoring method based on big data | |
| Qu et al. | Remaining Useful Life Prediction of High-Speed Railroad Contact Network Based on Stacking Integrated Attention-LSTM-CNN Deep Learning |
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 |