CN111008485B - Neural network-based multi-parameter life prediction method for three-phase alternating current asynchronous motor - Google Patents
Neural network-based multi-parameter life prediction method for three-phase alternating current asynchronous motor Download PDFInfo
- Publication number
- CN111008485B CN111008485B CN201911354074.4A CN201911354074A CN111008485B CN 111008485 B CN111008485 B CN 111008485B CN 201911354074 A CN201911354074 A CN 201911354074A CN 111008485 B CN111008485 B CN 111008485B
- Authority
- CN
- China
- Prior art keywords
- motor
- neural network
- service life
- parameter
- training
- 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.)
- Expired - Fee Related
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 50
- 238000000034 method Methods 0.000 title claims abstract description 30
- 238000012549 training Methods 0.000 claims abstract description 40
- 230000001012 protector Effects 0.000 claims abstract description 28
- 238000012360 testing method Methods 0.000 claims abstract description 18
- 238000012544 monitoring process Methods 0.000 claims description 6
- 238000012423 maintenance Methods 0.000 claims description 5
- 238000005265 energy consumption Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 230000017105 transposition Effects 0.000 claims description 3
- 238000003745 diagnosis Methods 0.000 claims description 2
- 238000013135 deep learning Methods 0.000 abstract description 6
- 230000002159 abnormal effect Effects 0.000 abstract description 4
- 238000007689 inspection Methods 0.000 abstract description 2
- 230000007774 longterm Effects 0.000 description 5
- 238000004519 manufacturing process Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 238000004891 communication Methods 0.000 description 3
- 238000009826 distribution Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 238000009776 industrial production Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000013075 data extraction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000010438 heat treatment Methods 0.000 description 1
- 238000011031 large-scale manufacturing process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 239000003208 petroleum Substances 0.000 description 1
- 238000005086 pumping 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
- G01R31/34—Testing dynamo-electric machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Control Of Electric Motors In General (AREA)
- Tests Of Circuit Breakers, Generators, And Electric Motors (AREA)
Abstract
The invention discloses a neural network-based multi-parameter life prediction method for a three-phase alternating current asynchronous motor, which belongs to the field of industrial motor security inspection systems, and realizes the multi-parameter life prediction of the motor by using the state parameter and the protection event parameter recording information of a microcomputer relay protector and combining a deep learning network, wherein the prediction step comprises the following steps: establishing a neural network training set, a half life training subset and a loss training subset, sending the training set into a neural network for training, correcting network parameters by using an error back propagation algorithm, obtaining parameterized description of the network on the service life of the motor, establishing a motor loss test set, inputting the established neural network and predicting; the complex modeling process is avoided, the accuracy of prediction is improved, the potential safety hazard of abnormal shutdown of the factory process is avoided, and the economic loss and the casualty risk caused by safety accidents are reduced.
Description
Technical Field
The invention belongs to the field of industrial motor security inspection systems, and particularly relates to a three-phase alternating current asynchronous motor service life prediction method.
Background
With the rapid development of economy, the demand of society for electric power energy is vigorous, so that the capacity of an electric power system is continuously enlarged, and the development scale is rapid. Among them, the three-phase ac asynchronous motors are widely used in industrial production, for example, in the field of petroleum and petrochemical industry, a chemical plant often needs to operate several hundreds or even thousands of three-phase ac asynchronous motors at the same time. These asynchronous motors are responsible for pumping or pressurizing materials and are the most important energy-consuming devices in the chemical industry field.
Especially, the main and standby motors at key production process nodes need large-scale production halt and shutdown for once replacement, which causes certain economic loss to manufacturers. These motors cannot be used for "sleeping for the end of life" at any time, and cannot be replaced when they are broken. The normal and reliable operation of each motor is closely related to the production life and property safety, so that the method has great practical significance for effective health management and life prediction of the motors.
In the prior art, the control and state monitoring of three-phase alternating current asynchronous motors in industrial production mainly have two forms: firstly, a frequency converter is used for direct drive, and state parameters of the frequency converter are networked by a communication interface of the frequency converter and uploaded to background management software; and secondly, a microcomputer relay protector (namely a motor protector) is used for management, and state parameters of the microcomputer relay protector are uploaded to background management software through a communication interface of the motor protector.
The motor protector is a microcomputer relay protector, and is one of the most widely applied devices in the monitoring and protection of the cluster state of the three-phase alternating current asynchronous motor. Generally, a motor protector is responsible for monitoring a three-phase alternating current asynchronous motor and can measure parameters such as three-phase voltage, three-phase current, leakage current and temperature of the motor in real time; and calculating by using the parameters to obtain various derived parameters of the operation of the motor, such as positive and negative sequence current, active power, reactive power, apparent power, power factor, heating capacity and the like.
Besides monitoring various parameters of the motor operation, the motor protector can delay fixed time to send a tripping command after detecting abnormal parameters, so as to cut off the power supply of the motor and realize a plurality of protection functions. Generally, a microcomputer relay protector has the following over ten motor protection functions: start-up overtime protection, locked rotor protection, thermal overload protection, three-phase current imbalance protection, single-phase ground protection, leakage protection, phase failure protection, overvoltage protection, undervoltage protection, undercurrent protection, anti-dazzling protection, delayed restart protection, TE time protection of an intrinsically safe motor, and the like. After the protection functions are performed, all detection parameters and the occurrence time are automatically recorded by the motor protector, and can be displayed on site or uploaded to a background through a communication interface. However, for the application field of industrial motors, the simple trip protection cannot meet the requirements of safety production, and even causes the safety accidents of pipeline blockage, tank body explosion and the like.
Meanwhile, after the records of the state parameters and the protection event parameters are uploaded to background software, the records are often only used for generating a production management report. However, in practice, these state parameters and protection records reflect detailed information of the whole life cycle of a motor, wherein the fluctuation of the state parameters, the occurrence frequency and the occurrence time of the abnormality directly influence the service life of the motor, and are very important data in the motor health management. The prior art is difficult to apply various parameters quantitatively, and management and maintenance personnel can only use the parameters as reference information for maintenance through personal experience.
Disclosure of Invention
The invention provides a method for realizing multi-parameter life prediction of a motor by using state parameters and protection event parameter recording information of a microcomputer relay protector and combining a deep learning network, aiming at solving the problems that the service life of an industrial motor cannot be accurately detected in real time, the utilization rate of each motor is poor and the safety risk is high in the prior art.
The technical scheme provided by the invention is as follows:
a neural network-based multi-parameter life prediction method for a three-phase alternating current asynchronous motor comprises the following prediction steps:
1) establishing a neural network training set:
a) constructing a full loss training subset: in the existing motor protector background management software, aiming at each motor which has faults and needs to be replaced, the running state parameters during fault maintenance are extracted, after the data are standardized, the neural network input vector I is constructedk(ii) a Since these motors are when they have failed/repaired, their used life is marked as 100%;
b) constructing half of life training subsets: in the existing motor protector background management software, aiming at each motor which has faults and needs to be replaced, operating state parameters of half of the service life of the motor are extracted, and after the data are standardized, the data are merged into a neural network input vector Ik(ii) a Extracting parameters of half of the service life of the failed motor, and marking the service life of the failed motor as 50%;
c) constructing a loss training subset: in the existing motor protector background management software, aiming at each motor which has faults and needs to be replaced, the running state parameters of which the service life is 10 percent are extracted, and the data are standardized and then merged into a neural network input vector Ik(ii) a Extracting parameters when the failed motor has 10% of service life, and marking the service life as 10%;
in each of the above training sets, k is equal to at least two of 2 and (8, …,12), and the operation status parameter includes a total operation time i2And includes a maximum average operating current i8Maximum operating current i9Average power factor i10Minimum power factor i11And average operating temperature i12At least two of;
2) sending the training set into a neural network for training, wherein the input layer of the neural network has the number of input nodes which is the same as the number h of the k values, and the output quantity obtained by the neural network is described as the service life of the motor and is a percentage/decimal number of 0 to 1 corresponding to each input data of the state parameters;
3) correcting network parameters by using an error back propagation algorithm to obtain parameterized description of the service life of the motor by the network; the neural network structure includes: the system comprises an input layer, four full-connection layers and a softmax classification layer, wherein 4h internal nodes are set in each of the four full-connection layers, and h is the number of state parameter types;
in the classification layer, the softmax classifier is preferably defined as: accumulating m possible values of all class labels as denominator, and sampling x(i)The probability of classifying into category j is:
wherein e is a natural constant, theta is a classification parameter, T represents transposition, and j is a fault type number;
4) constructing a motor loss test set: in the existing motor protector background management software, aiming at each motor which has faults and needs to be replaced, the running state parameters at any time are extracted to form a test set, the data of the test set is input into a neural network, and the service life of the motor is predicted according to the network forward propagation process.
Preferably, the parameter ik(k 8, …,12) is calculated from the following normalized equation:
i9＝maxt(max(IA+IB+IC))，
wherein, IA，IB，ICA, B, C three-phase currents, max, respectivelytThe representation is taken to be the maximum value in the history,
is the impedance angle, maxtThe representation is the minimum value of the historical time, and T represents the working temperature of the motor.
Preferably, the operation state parameter further includes a total operation start time i1Total number of start-stops i3Total thermal overload trip time i4Total locked rotor tripping time i5Total unbalanced three phase trip time i6Total energy consumption i7And rated power i of motor13K is equal to 0-7 of 1,3,4,5,6,7, 13.
Since the system records the operating state parameters of each damaged motor in different use stages, in some preferred embodiments, the operating state parameters of at least three typical operating stages (100% loss, 50% loss, 10% loss) are selected as training samples, so that the number of samples is three times greater than the number h of the types of parameters selected by the actual motor. Other embodiments of the invention can still select data of more working stages, and further increase the number of training samples, thereby enabling the network training to be more accurate and reliable.
The comprehensive technical scheme and the comprehensive effect of the invention comprise:
1. recording state parameter information and protection event parameters of the motor protector, using the state parameter information and the protection event parameters in a deep learning network, and obtaining prediction of the service life of the motor through calculation of the deep learning network, so as to guide a production manager to take necessary measures to replace a standby motor before the service life of the motor is reached, further, the potential safety hazard of abnormal shutdown of a factory process is avoided, and the economic loss and the casualty risk caused by safety accidents are reduced;
2. the state information of the motor protector is associated with the protection event record to be used for a deep learning network, the prediction of the service life of the motor is calculated in real time through the deep learning network, the defect that the service life of the motor is predicted only through the total running time and the total start-stop times of the motor in the prior art is overcome, and the accuracy of the prediction of the service life of the motor is improved;
3. the invention adopts the neural network to carry out the life correlation prediction by utilizing specific and sensitive motor operation parameters, avoids the complex modeling process, directly analyzes the data, and can reduce the error caused by incomplete consideration and neglect of secondary factors in the modeling process;
4. the four-layer fully-connected deep neural network structure is adopted, the number of nodes on each layer is four times that of the input layer, influence factors of key operation parameters on the service life of the motor can be highlighted according to different combinations of samples, the complexity of a motor operation service life prediction model is embodied, and therefore the prediction accuracy is improved.
Drawings
Fig. 1 is a schematic structural diagram of an inverse neural network in a method for predicting the life of a three-phase alternating-current asynchronous motor based on a neural network model according to an embodiment of the present invention.
Detailed Description
The invention is described in detail below with reference to the figures and the specific embodiments.
Example 1
A neural network-based multi-parameter life prediction method for a three-phase alternating current asynchronous motor comprises the following prediction steps:
1) constructing a neural network training set:
a) constructing a full loss training subset: in the existing motor protector background management software, aiming at each motor which has faults and needs to be replaced, the running state parameters during fault maintenance are extracted, and after the data are standardized, the running state parameters are constructedNeural network input vector Ik(ii) a Since these motors are when they have failed/repaired, their used life is marked as 100%;
b) constructing half of life training subsets: in the existing motor protector background management software, aiming at each motor which has faults and needs to be replaced, operating state parameters of half of the service life of the motor are extracted, and after the data are standardized, the data are merged into a neural network input vector Ik(ii) a Extracting parameters of half of the service life of the failed motor, and marking the service life of the failed motor as 50%;
c) constructing a loss training subset: in the existing motor protector background management software, aiming at each motor which has faults and needs to be replaced, the running state parameters of which the service life is 10 percent are extracted, and the data are standardized and then merged into a neural network input vector Ik(ii) a Extracting parameters when the failed motor has 10% of service life, and marking the service life as 10%;
in each of the above training sets, k equals 2 and (8, …,12), and the operation status parameter includes the total operation time i2Maximum average operating current i8Maximum operating current i9Average power factor i10Minimum power factor i11And average operating temperature i12；
2) Sending the training set into a neural network for training, wherein the input layer of the neural network has the number of input nodes which is the same as the number h of the k values, the output quantity obtained by the neural network is described as the service life of the motor and is a percentage/decimal of 0 to 1 corresponding to each input data of the state parameters, and h is 6 in the embodiment;
3) correcting network parameters by using an error back propagation algorithm to obtain parameterized description of the service life of the motor by the network; the neural network structure includes: the system comprises an input layer, four full-connection layers and a softmax classification layer, wherein each layer of the four full-connection layers is provided with 24 internal nodes, and h is the number of state parameter types;
in the classification layer, the softmax classifier is defined as: m of all classes can be markedAccumulating the energy values as denominators, and sampling x(i)The probability of classifying into category j is:
wherein e is a natural constant, theta is a classification parameter, T represents transposition, and j is a fault type number;
4) constructing a motor loss test set: in the existing motor protector background management software, aiming at each motor which has faults and needs to be replaced, the running state parameters at any time are extracted to form a test set, the data of the test set is input into a neural network, and the service life of the motor is predicted according to the network forward propagation process.
Parameter ik(k 8, …,12) is calculated from the following normalized equation:
i9＝maxt(max(IA+IB+IC))，
wherein, IA，IB，ICA, B, C three-phase currents, max, respectivelytThe representation is taken to be the maximum value in the history,
is the impedance angle, mintThe expression is to take the minimum value of the historical time, T represents the working temperature of the motor, and the state parameters are strongly related to the working time of the motor.
Example 2
In this embodiment, the method steps described in embodiment 1 are adopted to construct a neural network training set, send the neural network training set to a neural network for training, correcting the neural network, and perform testing, except that,
in step 1), as shown in fig. 1 and table 1, the operating state parameters further include a total operating start time i1Total number of start-stops i3Total thermal overload trip time i4Total locked rotor tripping time i5Total unbalanced three phase trip time i6Total energy consumption i7And rated power i of motor137, corresponding to k ═ 1, …, 13.
In step 2), the 13 normalized parameters of the 10 failed motors are sent to a neural network as a training set for training, the input layer of the neural network has the number of input nodes which is the same as the number h of the k values, the output quantity obtained by the neural network is described as the service life of the motor corresponding to each input data of the state parameters, and is a percentage/decimal of 0 to 1, and h is 13 in the embodiment;
in the step 3), network parameters are corrected by using an error back propagation algorithm to obtain parameterized description of the service life of the motor by the network; the neural network structure includes: one input layer, four full connection layers and one softmax classification layer; however, 52 internal nodes are set in each layer of four full-connection layers, and h is 13 as the number of the state parameter types;
in step 4), a motor loss test set is constructed: in the existing motor protector background management software, two motors with faults 11 and 12 are selected, the running state parameters of the two motors with faults and needing to be replaced at any time are extracted to form a test set, data of the test set are input into a neural network, the service life of the motor is predicted according to the network forward propagation process, and the test result is shown in table 2.
Table 1 shows the distribution of 13 normalized parameters when the 10 motors with faults are replaced, and from these parameters, it can be seen that some motors are damaged due to reaching the service life, some motors are damaged due to long-term abnormal working state, and the accuracy of the prediction method of the present invention is further ensured due to the diversity of samples.
TABLE 110 normalized 13 parameter distributions for a failed motor when replaced
| Motor numbering | i1 | i2 | i3 | i4 | i5 | i6 | i7 | i8 | i9 | i10 | i11 | i12 | i13 | Cause of |
| 1 | 0.31 | 0.33 | 0.54 | 0.85 | 0.82 | 0.11 | 0.94 | 0.98 | 0.98 | 0.74 | 0.71 | 0.72 | 0.5 | Long term overload |
| 2 | 0.43 | 0.95 | 0.39 | 0.74 | 0.03 | 0.02 | 0.95 | 0.87 | 0.87 | 0.68 | 0.39 | 0.49 | 0.8 | Expiration date |
| 3 | 0.33 | 0.98 | 0.78 | 0.34 | 0.34 | 0.04 | 0.91 | 0.97 | 0.97 | 0.68 | 0.54 | 0.44 | 0.5 | Expiration date |
| 4 | 0.35 | 0.35 | 0.47 | 0.77 | 0.16 | 0.06 | 0.88 | 0.99 | 0.99 | 0.78 | 0.78 | 0.23 | 0.5 | Long term overload |
| 5 | 0.99 | 0.9 | 0.98 | 0.60 | 0.14 | 0.05 | 0.68 | 0.99 | 0.96 | 0.66 | 0.76 | 0.68 | 0.2 | Frequent start and stop |
| 6 | 0.73 | 0.7 | 0.65 | 0.54 | 0.26 | 0.09 | 0.77 | 0.89 | 0.93 | 0.70 | 0.23 | 0.62 | 0.8 | Is unknown |
| 7 | 0.62 | 0.99 | 0.98 | 0.23 | 0.26 | 0.11 | 0.9 | 0.90 | 0.90 | 0.77 | 0.51 | 0.59 | 0.5 | Expiration date |
| 8 | 0.9 | 0.5 | 0.61 | 0.21 | 0.49 | 0.11 | 0.91 | 0.92 | 0.96 | 0.74 | 0.31 | 0.13 | 0.5 | Is unknown |
| 9 | 0.51 | 0.5 | 0.67 | 0.43 | 0.34 | 0.12 | 0.84 | 0.82 | 0.87 | 0.76 | 0.41 | 0.49 | 0.5 | Is unknown |
| 10 | 0.62 | 0.5 | 0.76 | 0.14 | 0.03 | 0.12 | 0.88 | 0.78 | 0.85 | 0.76 | 0.42 | 0.77 | 0.5 | Is unknown |
It can be seen from table 1 that for a motor known to be damaged, the corresponding parameters are found to be significantly larger, for example, the number 1 long-term overload motor, the total thermal overload trip time i4Reaches 0.85, and for the No. 5 motor with frequent start-stop damage, the total starting time i1And total number of starts i3Are all relatively high. However, for the motor with unknown damage cause, the damage cause cannot be intuitively judged by people, and the service life of the motor cannot be predicted accordingly.
Therefore, a neural network needs to be established for the damaged motor under the complex conditions, and the intelligent algorithm finds the combination of the key data which are easy to damage through a large amount of data learning by the prediction method, so that the service life of the motor is accurately predicted.
And taking the data in the table 1 as a neural network standardized parameter test set for training and correcting to obtain a neural network, substituting state parameters of two motor test sets to be tested for life test, and obtaining the result shown in the table 2.
TABLE 2 13 normalized parameter life distribution of two failed machines when replaced
| Motor numbering | i1 | i2 | i3 | i4 | i5 | i6 | i7 | i8 | i9 | i10 | i11 | i12 | i13 | Cause of failure |
| 11 | 0.83 | 0.91 | 0.77 | 0.34 | 0.5 | 0.1 | 0.59 | 0.83 | 0.56 | 0.32 | 0.33 | 0.58 | 0.18 | Frequent start and stop |
| 12 | 0.41 | 0.38 | 0.53 | 0.80 | 0.16 | 0.06 | 0.88 | 0.68 | 0.95 | 0.48 | 0.69 | 0.56 | 0.57 | Long term overload |
Because the motor protector background management software runs in real time, the processes of data extraction and prediction can be fused into the motor background management software in real time, and dynamic motor service life prediction is provided. For example, when a trip protection condition frequently occurs in a certain motor in a recent period of time, background management software automatically calls historical data of the motor to predict the service life, and real-time diagnosis and service life prediction are realized on the basis of a real-time monitoring function.
It will be understood that modifications and variations can be effected by a person skilled in the art in light of the above teachings and that all such modifications and variations are intended to be within the scope of the invention as defined by the appended claims. Since the system records the operating state parameters of each damaged motor in different use stages, in the preferred embodiments 1 and 2, the operating state parameters of at least three typical operating stages (100% loss, 50% loss, 10% loss) are selected as training samples, so that the number of the samples is three times greater than the number h of the types of parameters selected by the actual motor. Other embodiments of the invention can still select data of more working stages, and further increase the number of training samples, thereby enabling the network training to be more accurate and reliable.
Claims (7)
1. The multi-parameter life prediction method of the three-phase alternating current asynchronous motor based on the neural network is characterized by comprising the following prediction steps:
1) establishing a neural network training set:
a) constructing a full loss training subset: in the existing motor protector background management software, aiming at each motor which has faults and needs to be replaced, the running state parameters during fault maintenance are extracted, after the data are standardized, the neural network input vector I is constructedk(ii) a Since these motors are when they have failed/repaired, their used life is marked as 100%;
b) constructing half of life training subsets: in the existing motor protector background management software, aiming at each motor which has faults and needs to be replaced, operating state parameters of half of the service life of the motor are extracted, and after the data are standardized, the data are merged into a neural network input vector Ik(ii) a Because the parameters of the failed motor with half of the service life are extracted, the failed motor with half of the service life is usedThe hit number is 50%;
c) constructing a loss training subset: in the existing motor protector background management software, aiming at each motor which has faults and needs to be replaced, the running state parameters of which the service life is 10 percent are extracted, and the data are standardized and then merged into a neural network input vector Ik(ii) a Extracting parameters when the failed motor has 10% of service life, and marking the service life as 10%;
in each of the above training sets, k is equal to 2 and at least two of 8, …,12, and the operation status parameter includes a total operation time i2And includes a maximum average operating current i8Maximum operating current i9Average power factor i10Minimum power factor i11And average operating temperature i12At least two of;
2) sending the training set into a neural network for training, wherein the input layer of the neural network has the number of input nodes which is the same as the number h of the k values and corresponds to each input data of the state parameters, and the output quantity obtained by the neural network is described as the service life of the motor and is a percentage/decimal number of 0 to 1;
3) correcting network parameters by using an error back propagation algorithm to obtain parameterized description of the service life of the motor by the network; the neural network structure includes: the system comprises an input layer, four full-connection layers and a softmax classification layer, wherein 4h internal nodes are set in each of the four full-connection layers, and h is the number of state parameter types;
4) constructing a motor loss test set: in the existing motor protector background management software, aiming at each motor which has faults and needs to be replaced, the running state parameters at any time are extracted to form a test set, the data of the test set is input into a neural network, and the service life of the motor is predicted according to the network forward propagation process.
2. The multi-parameter life prediction method of a three-phase AC asynchronous motor according to claim 1, characterized in that parameter ikK is 8, …,12 calculated from the following normalized equation:
i9＝maxt(max(IA+IB+IC))，
wherein, IA，IB，ICA, B, C three-phase currents, max, respectivelytThe representation is taken to be the maximum value in the history,
is the impedance angle, mintThe representation is the minimum value of the historical moment, T represents the working temperature of the motor, and T represents the working time of the motor.
3. The multi-parameter life prediction method of a three-phase AC asynchronous motor according to claim 1,
the operating state parameter further comprises a total operating start time i1Total number of start-stops i3Total thermal overload trip time i4Total locked rotor tripping time i5Total unbalanced three phase trip time i6Total energy consumption i7And rated power i of motor13K is equal to 0-7 of 1,3,4,5,6,7, 13.
4. The multi-parameter life prediction method of a three-phase AC asynchronous motor according to claim 1,
in the classification layer of the step 3), the softmax classifier is defined as: accumulating m values of all class marks as denominator, and sampling x(i)The probability of classifying into category j is:
where e is a natural constant, θ is a classification parameter, T denotes transposition, and j is a fault type number.
5. Prediction method according to any of claims 1-4, characterized in that at least three typical working phases are selected, the operating state parameters of 100% loss, 50% loss and 10% loss being used as training samples.
6. The prediction method according to claim 5, wherein besides the operation state parameters of the typical working phase, other 5 times of percentage loss are selected as training samples, that is, the state parameters of each phase are selected to be 100%, 95%, …, 10% and 5% loss.
7. The prediction method of claim 5, wherein the process of extracting data and predicting is fused into the motor background management software in real time, and dynamic motor life prediction is provided;
specifically, when a certain motor frequently has a trip protection condition within a recent period of time, background management software automatically calls historical data of the motor to predict the service life, and real-time diagnosis and service life prediction are realized on the basis of a real-time monitoring function.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911354074.4A CN111008485B (en) | 2019-12-25 | 2019-12-25 | Neural network-based multi-parameter life prediction method for three-phase alternating current asynchronous motor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911354074.4A CN111008485B (en) | 2019-12-25 | 2019-12-25 | Neural network-based multi-parameter life prediction method for three-phase alternating current asynchronous motor |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111008485A CN111008485A (en) | 2020-04-14 |
| CN111008485B true CN111008485B (en) | 2021-05-11 |
Family
ID=70118127
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911354074.4A Expired - Fee Related CN111008485B (en) | 2019-12-25 | 2019-12-25 | Neural network-based multi-parameter life prediction method for three-phase alternating current asynchronous motor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111008485B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111612242A (en) * | 2020-05-15 | 2020-09-01 | 北京天工智造科技有限公司 | Motor state parameter prediction method based on LSTM deep learning model |
| CN111722106A (en) * | 2020-06-05 | 2020-09-29 | 格力电器（合肥）有限公司 | Motor operation detection system and method |
| CN113064075B (en) * | 2021-03-16 | 2023-05-05 | 电子科技大学成都学院 | Motor service life prediction method based on edge calculation and deep learning |
| CN114896873B (en) * | 2022-04-26 | 2023-04-07 | 盐城工学院 | Traveling wave ultrasonic motor service life prediction method and system based on convolutional neural network |
| CN117193149B (en) * | 2023-11-08 | 2024-02-13 | 重庆华悦生态环境工程研究院有限公司深圳分公司 | Control method and device for equipment operation state and electronic equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101630840A (en) * | 2009-08-12 | 2010-01-20 | 电子科技大学 | Intelligent control system for microgrid energy |
| CN106612094A (en) * | 2015-10-26 | 2017-05-03 | 发那科株式会社 | Machine learning apparatus and method, life prediction apparatus and motor driving apparatus |
| EP3432094A1 (en) * | 2017-07-20 | 2019-01-23 | Siemens Aktiengesellschaft | System state prediction |
| CN110175369A (en) * | 2019-04-30 | 2019-08-27 | 南京邮电大学 | A kind of gear method for predicting residual useful life based on two-dimensional convolution neural network |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090177068A1 (en) * | 2002-10-09 | 2009-07-09 | Stivoric John M | Method and apparatus for providing derived glucose information utilizing physiological and/or contextual parameters |
| CN105649978B (en) * | 2016-02-02 | 2017-09-19 | 西南石油大学 | Single-screw (single screw) pump fault diagnosis and performance testing device and method |
-
2019
- 2019-12-25 CN CN201911354074.4A patent/CN111008485B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101630840A (en) * | 2009-08-12 | 2010-01-20 | 电子科技大学 | Intelligent control system for microgrid energy |
| CN106612094A (en) * | 2015-10-26 | 2017-05-03 | 发那科株式会社 | Machine learning apparatus and method, life prediction apparatus and motor driving apparatus |
| EP3432094A1 (en) * | 2017-07-20 | 2019-01-23 | Siemens Aktiengesellschaft | System state prediction |
| CN110175369A (en) * | 2019-04-30 | 2019-08-27 | 南京邮电大学 | A kind of gear method for predicting residual useful life based on two-dimensional convolution neural network |
Non-Patent Citations (1)
| Title |
|---|
| Non-equilibrium plasma jet induced thermo-acoustic resistivity imaging for higher contrast and resolution;Liang Guo;《Scientific Reports》;20170825;第1-6页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111008485A (en) | 2020-04-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111008485B (en) | Neural network-based multi-parameter life prediction method for three-phase alternating current asynchronous motor | |
| CN103454516B (en) | Intelligent transformer substation secondary equipment health state diagnostic method | |
| CN104537487B (en) | A kind of appraisal procedure of power transmission and transforming equipment operation state risk | |
| EP2697699B1 (en) | Dynamic assessment system for high-voltage electrical components | |
| CN108073551B (en) | High-voltage switch cabinet online fault diagnosis method based on multi-Agent cooperation | |
| CN112713649B (en) | Power equipment residual life prediction method based on extreme learning machine | |
| CN109188227B (en) | Double-fed wind driven generator insulation state evaluation method and system | |
| CN106597965A (en) | Nuclear power apparatus running state monitoring system and monitoring method | |
| CN102289731A (en) | Method for maintaining state of power transmission equipment based on system risk | |
| CN103278728A (en) | Method and system for diagnosing interturn short circuit fault of generator rotor | |
| CN106408438A (en) | Method for automatically checking safety and stability control strategy based on dynamic simulation of whole process | |
| CN105260778A (en) | Power transformer fault diagnosis system based on expert database | |
| CN111929579B (en) | Generator online fault diagnosis method and device and computer device | |
| CN107247198A (en) | A kind of distribution equipment malfunction Forecasting Methodology and device | |
| CN113156058A (en) | Nuclear power station transformer light gas alarm response method, device, equipment and medium | |
| CN104834305B (en) | Distribution automation terminal remote measurement exception analysis system and method based on DMS systems | |
| Bardyk et al. | Parametric identification of fuzzy model for power transformer based on real operation data | |
| CN106802599A (en) | A kind of diagnosing fault of power transformer system based on expert database | |
| CN112183610A (en) | Transformer fault prediction method and transformer fault prediction device | |
| CN110265998B (en) | Method and device for checking strategy fixed value of stability control system | |
| CN117031201A (en) | Multi-scene topology anomaly identification method and system for power distribution network | |
| CN112182960A (en) | Power transformer state risk assessment method based on Bayesian network | |
| Tippannavar et al. | Smart transformer-An analysis of recent technologies for monitoring transformer | |
| CN103499754A (en) | Method for real time on-line monitoring of operating state of power capacitor bank | |
| Li et al. | Design and realization of transformer fault diagnostic expert system based on drools |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210511 |