CN114548636A - Deep neural network-based multidimensional unit life cycle evaluation method - Google Patents
Deep neural network-based multidimensional unit life cycle evaluation method Download PDFInfo
- Publication number
- CN114548636A CN114548636A CN202111267329.0A CN202111267329A CN114548636A CN 114548636 A CN114548636 A CN 114548636A CN 202111267329 A CN202111267329 A CN 202111267329A CN 114548636 A CN114548636 A CN 114548636A
- Authority
- CN
- China
- Prior art keywords
- neural network
- deep neural
- life cycle
- evaluation method
- unit life
- 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
- 238000011156 evaluation Methods 0.000 title claims abstract description 54
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 50
- 238000012549 training Methods 0.000 claims abstract description 30
- 238000012360 testing method Methods 0.000 claims abstract description 12
- 238000012795 verification Methods 0.000 claims abstract description 7
- 230000006870 function Effects 0.000 claims description 19
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 17
- 239000003921 oil Substances 0.000 claims description 12
- 210000002569 neuron Anatomy 0.000 claims description 7
- 239000010687 lubricating oil Substances 0.000 claims description 5
- 238000000034 method Methods 0.000 claims description 5
- 238000003062 neural network model Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 238000004804 winding Methods 0.000 claims description 3
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical group [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 claims description 2
- 230000003862 health status Effects 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000011157 data evaluation Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- 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
- G06N3/048—Activation functions
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Software Systems (AREA)
- Marketing (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Entrepreneurship & Innovation (AREA)
- Educational Administration (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Development Economics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a multidimensional unit life cycle evaluation method based on a deep neural network, which comprises the steps of carrying out category marking-label setting on all data samples by acquiring characteristic data and normalizing; dividing a sample into a test set and a training set, and training by using a deep neural network; after the model is trained, inputting health state data for verification to obtain an evaluation result; the health status data samples are periodically updated and trained and tested to update the model. The invention utilizes the deep neural network to judge the evaluation grade of the life cycle according to the input multidimensional characteristic data, thereby greatly simplifying the flow of manual evaluation and improving the working efficiency.
Description
Technical Field
The invention provides a multidimensional unit life cycle evaluation method based on a deep neural network, and belongs to the technical field of electric power system device evaluation.
Background
Electric power is vital to China, most of the current evaluation in an electric power system is based on manual evaluation, and the subjective consciousness of the manual evaluation is very strong, so that the difference of evaluation results is easily caused. Therefore, the online evaluation method and system are key to solving the problem. How to objectively, accurately and completely evaluate the comprehensive capability level of the power system, and the problem of extracting characteristic data evaluation indexes of a life cycle in all aspects and in multiple angles is urgently needed to be solved. With the rapid development of artificial intelligence, deep learning has been advanced to a plurality of industries, and the use of a deep neural network to replace a traditional optimization function can not only obtain an accurate result, but also save time. Therefore, the deep neural network is used for solving the life cycle evaluation, and the method and the system for analyzing the unit life cycle evaluation data based on the deep neural network multi-dimensional features are very important.
Disclosure of Invention
The invention aims to provide a multidimensional unit life cycle evaluation method based on a deep neural network, which is used for solving the problems in the background technology.
The invention provides a multidimensional unit life cycle evaluation method based on a deep neural network, which adopts the following technical scheme that the method comprises the following steps:
acquiring characteristic data and normalizing; performing category labeling-label setting on all data samples; dividing a sample into a test set and a training set, and training by using a deep neural network; after the model is trained, inputting health state data for verification to obtain an evaluation result; periodically updating the health state data samples, and performing training and testing to update the model;
(1) the acquiring and normalizing feature data, which includes a plurality of data of a life cycle, includes: the temperature of an upper guide bearing and an upper frame bush, the oil level and the oil temperature of a motor lubricating oil tank, the water flow, the pressure and the temperature of a water inlet and a water outlet of a cooler, the X-direction swing of the upper swing of the vibration pendulum, the temperature of each block of bush of the bearing bush of the thrust bearing, the X-direction swing of the lower swing of the vibration pendulum, the Y-direction swing, the temperature of a rotor winding, the air clearance of a unit, the vibration of a stator core, the water level of a top cover of an overflowing part and the water leakage flow of a guide vane; and then normalizing the characteristic data, wherein the temperatures of all the tiles of the upper guide bearing, the upper frame tile, the oil level and the oil temperature of the oil groove of the lubricating oil of the motor, the water flow rate of the water inlet and the water outlet of the cooler, the pressure, the temperature, the X-direction pendulum degree of the pendulum and the tile temperature of the thrust bearing have fixed weights, and the sum of the weights is 1, namely the requirement that:
wherein K is a large class characteristic number,the subclass feature number is the kth class feature;
(2) and performing category labeling-setting labels on all data samples. Acquiring all historical sample data (historical characteristics and evaluation results), classifying the historical sample data into five categories, namely excellent, good, general, poor and poor categories based on sample characteristics, and labeling each sample;
(3) and dividing the sample into a test set and a training set, and training by using a deep neural network. Performing cross classification (5 times of cross) according to 0.2 of the test set in all samples and 0.8 of the training set in all samples, and then constructing a deep neural network for training;
(4) and after the model is trained, inputting health state data for verification to obtain an evaluation result. After the model is trained, all characteristic data samples are input into the neural network model, and an evaluation result is output according to the model;
(5) periodically updating the health state data samples, and performing training and testing to update the model; when the health state data number reaches a certain number (a preset threshold value), inputting the health state data feature data into the model and continuing training the model together with the original historical feature data, thereby perfecting the evaluation system;
(6) the deep neural network comprises an input layer, two hidden layers and an output layer, wherein the input layer inputs normalized characteristic data, the two hidden layers are of a full-connection structure, the number of neurons is 150 and 100 respectively, an activation function is a tanh function and a sigmond function respectively, and the output layer is a five-dimensional vector and represents excellent, good, common, poor and very poor respectively;
(7) the Loss function of the deep neural network utilizes Mean square error Loss Mean Squared Loss (MSL), and adopts a random gradient descent method to update the gradient to optimize the Loss function, so that the Loss function is minimized.
Drawings
FIG. 1 is a schematic diagram of a unit life cycle evaluation method and system based on a deep neural network multi-dimensional feature;
FIG. 2 is a schematic diagram of a deep neural network architecture;
FIG. 3 is a flow chart of a unit life cycle evaluation method based on a deep neural network multi-dimensional feature;
FIG. 4 is a block diagram of a unit life cycle evaluation system based on a deep neural network multi-dimensional feature.
Detailed Description
The present invention is further illustrated by the following examples, which do not limit the present invention in any way, and any modifications or changes that can be easily made by a person skilled in the art to the present invention will fall within the scope of the claims of the present invention without departing from the technical solution of the present invention.
Example 1
As shown in fig. 1, the method includes: acquiring characteristic data and normalizing; performing category labeling-label setting on all data samples; dividing a sample into a test set and a training set, and training by using a deep neural network; after the model is trained, inputting health state data for verification to obtain an evaluation result; the healthier data samples are periodically sampled, trained and tested, and the model is updated.
(1) The acquiring and normalizing feature data, which includes a plurality of data of a life cycle, includes: the temperature of the upper guide bearing and the upper frame tile, the oil level and the oil temperature of a lubricating oil groove of a motor, the water flow, the pressure, the temperature, the swing upper guide X-direction swing degree, the temperature of each tile of the thrust bearing tile, the swing lower guide X-direction swing degree, the Y-direction swing degree, the temperature of a rotor winding, the air gap of a unit, the vibration of a vibration stator iron core, the water level of a top cover of an overflowing part and the water leakage flow of a guide vane. And then normalizing the characteristic data, wherein the temperature characteristics of each tile of the upper guide bearing, the tile temperature of the upper frame, the oil level and the oil temperature of the oil groove of the lubricating oil of the motor, the water flow rate of the water inlet and the water outlet of the cooler, the pressure, the temperature, the X-direction pendulum degree of the pendulum and the tile temperature of the thrust bearing have fixed weights, and the weight sum of the fixed weights is 1, namely the requirement that:
wherein K is a large class characteristic number,is the subclass feature number of the kth class feature.
(2) And performing category labeling-setting labels on all data samples. All historical sample data (historical characteristics and evaluation results) are acquired, classified based on sample characteristics, classified into five categories, namely excellent, good, general, poor and poor, and labeled for each sample.
(3) And dividing the sample into a test set and a training set, and training by using a deep neural network. And performing cross classification (5 times of cross) according to 0.2 of the test set in all samples and 0.8 of the training set in all samples, and then constructing a deep neural network for training.
(4) And after the model is trained, inputting health state data for verification to obtain an evaluation result. After the model is trained, all characteristic data samples are input into the neural network model, and the evaluation result is output according to the model.
(5) The healthier data samples are periodically sampled, trained and tested, and the model is updated. And when the health state data number reaches a certain number (a preset threshold value), inputting the health state data feature data into the model and continuing training the model together with the original historical feature data, thereby perfecting the evaluation system.
As shown in fig. 2, the deep neural network includes an input layer, two hidden layers and an output layer, wherein the input layer inputs normalized feature data, the two hidden layers are fully connected structures, the number of neurons is 150,100 respectively, the activation functions are respectively a tanh function and a sigmond function, and the output layer is a five-dimensional vector, which respectively represents excellent, good, general, poor and very poor. The Loss function of the deep neural network utilizes Mean square error Loss Mean Squared Loss (MSL), and adopts a random gradient descent method to update the gradient to optimize the Loss function, so that the Loss function is minimized.
As shown in FIG. 3, the method of the present invention begins by normalizing sample data based on its characteristics and weights; for example: the characteristic data can be set as parameters for initializing a neural network, the weight of two hidden layer neurons is initialized by Gaussian distribution with the mean value of 0 and the variance of 1, and the learning rate is set to be 0.01; and obtaining an expected output evaluation result value according to the sample label, and obtaining a historical evaluation grade according to historical data information so as to compare with the output of the neural network. Then determining an actual output result of an output layer according to the neural network model; since the weights of the initial neurons of the hidden layer are random, the deviation between the initial output result and the expected output result is large, but as the training progresses, the weights of the neurons of the hidden layer are gradually updated along with the back propagation of the neural network, and the final error is smaller than a set threshold, for example: 0.001. at this point, the neural network training is complete. And the trained model can be used for evaluation.
As shown in fig. 4, in the drawing, 1 represents an intelligent mobile terminal, which may be a mobile phone or a tablet, and 2 represents a server, which may be a mobile edge server or a large background server. The intelligent mobile terminal 1 can communicate with the server terminal 2 through a wireless communication technology (WLAN, a mobile network), and because training of the neural network needs to consume a large amount of computing resources and time cost, the training of the neural network is carried out at the server terminal 2, parameters of the model are transmitted to the intelligent mobile terminal 1 after the training is finished, and the intelligent mobile terminal 1 can directly output evaluation results according to input characteristic data of the intelligent mobile terminal 1 and by utilizing the existing model of the intelligent mobile terminal and weight parameters of neurons received from the server terminal 2, so that intelligent online evaluation is realized.
Claims (8)
1. A multidimensional unit life cycle evaluation method based on a deep neural network is characterized by comprising the following steps:
(1) acquiring characteristic data and normalizing;
(2) performing category labeling-label setting on all data samples;
(3) dividing a sample into a test set and a training set, and training by using a deep neural network;
(4) after the model is trained, inputting health state data for verification to obtain an evaluation result;
(5) the healthier data samples are periodically sampled, trained and tested, and the model is updated.
2. The deep neural network multi-dimensional unit life cycle evaluation method according to claim 1, wherein the deep neural network multi-dimensional unit life cycle evaluation method comprises the following steps:
acquiring characteristic data and normalizing, wherein the characteristic data comprises a plurality of items of data of a life cycle; the method comprises the following steps: the temperature of the upper guide bearing and an upper frame tile, the oil level and the oil temperature of a lubricating oil groove of a motor, the water flow, the pressure, the temperature, the pendulum upper guide X-direction pendulum degree, the temperature of each tile of the tile temperature of a thrust bearing tile, the pendulum lower guide X-direction pendulum degree, the Y-direction pendulum degree, the rotor winding temperature, the air gap of a unit, the vibration of a vibration stator iron core, the water level of a top cover of an overflowing part and the water leakage flow of a guide vane; all the features of the features have fixed weights, and the sum of the weights of the features is 1, namely that:
wherein K is the major characteristic number and the minor characteristic number of the kth characteristic.
3. The deep neural network multi-dimensional unit life cycle evaluation method according to claim 1, wherein the deep neural network multi-dimensional unit life cycle evaluation method comprises the following steps:
performing category labeling-label setting on all data samples; acquiring characteristics and evaluation results of all histories; the samples are classified based on their characteristics into five categories, excellent, good, general, poor and very poor, and each sample is labeled.
4. The deep neural network multi-dimensional unit life cycle evaluation method according to claim 1, wherein the deep neural network multi-dimensional unit life cycle evaluation method comprises the following steps:
the divided samples are a test set and a training set, and a deep neural network is used for training; and performing cross classification according to 0.2 of the test set in all samples and 0.8 of the training set in all samples, performing 5 times of cross, and then constructing a deep neural network for training.
5. The deep neural network multi-dimensional unit life cycle evaluation method according to claim 1, wherein the deep neural network multi-dimensional unit life cycle evaluation method comprises the following steps:
after the model is trained, inputting health state data for verification to obtain an evaluation result; after the model is trained, all characteristic data samples are input into the neural network model, and an evaluation result is output according to the model.
6. The deep neural network multi-dimensional unit life cycle evaluation method according to claim 1, wherein the deep neural network multi-dimensional unit life cycle evaluation method comprises the following steps:
the regular healthier state data samples are trained and tested, so that the model is updated; and when the health state data number reaches a preset threshold value, inputting the health state data feature data into the model and continuing training the model together with the original historical feature data, thereby perfecting the evaluation system.
7. The deep neural network multi-dimensional unit life cycle evaluation method according to claim 1, wherein the deep neural network multi-dimensional unit life cycle evaluation method comprises the following steps:
the deep neural network comprises an input layer, two hidden layers and an output layer, wherein the input layer inputs normalized characteristic data, the two hidden layers are of a full-connection structure, the number of neurons is 150 and 100 respectively, activation functions are a tanh function and a sigmond function respectively, and the output layer is a five-dimensional vector and represents excellent, good, common, poor and poor respectively.
8. The deep neural network multi-dimensional unit life cycle evaluation method based on claim 1, characterized in that:
the Loss function of the deep neural network utilizes Mean square error Loss Mean Squared Loss (MSL), and adopts a random gradient descent method to update the gradient to optimize the Loss function, so that the Loss function is minimized.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111267329.0A CN114548636A (en) | 2021-10-29 | 2021-10-29 | Deep neural network-based multidimensional unit life cycle evaluation method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111267329.0A CN114548636A (en) | 2021-10-29 | 2021-10-29 | Deep neural network-based multidimensional unit life cycle evaluation method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114548636A true CN114548636A (en) | 2022-05-27 |
Family
ID=81668796
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111267329.0A Pending CN114548636A (en) | 2021-10-29 | 2021-10-29 | Deep neural network-based multidimensional unit life cycle evaluation method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114548636A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115828699A (en) * | 2022-12-19 | 2023-03-21 | 华中科技大学 | Power semiconductor module full-life-cycle junction temperature prediction method, system and terminal |
-
2021
- 2021-10-29 CN CN202111267329.0A patent/CN114548636A/en active Pending
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115828699A (en) * | 2022-12-19 | 2023-03-21 | 华中科技大学 | Power semiconductor module full-life-cycle junction temperature prediction method, system and terminal |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107273490B (en) | Combined wrong question recommendation method based on knowledge graph | |
CN106845717B (en) | Energy efficiency evaluation method based on multi-model fusion strategy | |
CN103166830B (en) | A kind of Spam Filtering System of intelligent selection training sample and method | |
CN109145949A (en) | Non-intrusive electrical load monitoring and decomposition method and system based on integrated study | |
CN106022509B (en) | Consider the Spatial Load Forecasting For Distribution method of region and load character double differences | |
CN110929918A (en) | 10kV feeder line fault prediction method based on CNN and LightGBM | |
CN106779069A (en) | A kind of abnormal electricity consumption detection method based on neutral net | |
CN109472241A (en) | Combustion engine bearing remaining life prediction technique based on support vector regression | |
CN106321072B (en) | A kind of oil well fault diagnostic method based on pump dynagraoph | |
CN110598854A (en) | GRU model-based transformer area line loss rate prediction method | |
CN107463993A (en) | Medium-and Long-Term Runoff Forecasting method based on mutual information core principle component analysis Elman networks | |
CN111723257B (en) | User portrayal method and system based on water usage rule | |
CN108549907B (en) | Data verification method based on multi-source transfer learning | |
CN107515822B (en) | Software defect positioning method based on multiple-objection optimization | |
CN113344288B (en) | Cascade hydropower station group water level prediction method and device and computer readable storage medium | |
CN110530650A (en) | Heavy duty gas turbine performance state monitoring method based on generalized regression nerve networks Yu box map analysis | |
CN109307854A (en) | A kind of electric energy metering device comprehensive estimation method and system | |
CN106022352A (en) | Submersible piston pump fault diagnosis method based on support vector machine | |
CN114548636A (en) | Deep neural network-based multidimensional unit life cycle evaluation method | |
CN105787259A (en) | Method for analyzing influence correlation of multiple meteorological factors and load changes | |
CN111562541A (en) | Software platform for realizing electric energy meter detection data management by applying CART algorithm | |
CN116645129A (en) | Manufacturing resource recommendation method based on knowledge graph | |
Zhang et al. | Wind power prediction based on CNN-LSTM | |
CN112836876B (en) | Power distribution network line load prediction method based on deep learning | |
CN112417308A (en) | User portrait label generation method based on electric power big data |
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 |