CN113867306A - Fault detection method and system for air conditioning system of subway station hall - Google Patents
Fault detection method and system for air conditioning system of subway station hall Download PDFInfo
- Publication number
- CN113867306A CN113867306A CN202110874186.3A CN202110874186A CN113867306A CN 113867306 A CN113867306 A CN 113867306A CN 202110874186 A CN202110874186 A CN 202110874186A CN 113867306 A CN113867306 A CN 113867306A
- Authority
- CN
- China
- Prior art keywords
- rnn
- value
- condition
- data
- module
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 66
- 238000004378 air conditioning Methods 0.000 title claims abstract description 17
- 238000000034 method Methods 0.000 claims abstract description 19
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 17
- 238000012545 processing Methods 0.000 claims abstract description 14
- 238000004364 calculation method Methods 0.000 claims abstract description 13
- 230000006870 function Effects 0.000 claims abstract description 13
- 238000005457 optimization Methods 0.000 claims abstract description 11
- 230000006872 improvement Effects 0.000 claims abstract description 6
- 238000010606 normalization Methods 0.000 claims description 10
- 239000011159 matrix material Substances 0.000 claims description 7
- 238000013528 artificial neural network Methods 0.000 claims description 6
- 230000000694 effects Effects 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 238000011156 evaluation Methods 0.000 claims description 5
- 238000012549 training Methods 0.000 claims description 4
- 238000007405 data analysis Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 230000000306 recurrent effect Effects 0.000 claims description 3
- 230000005284 excitation Effects 0.000 claims description 2
- 230000009467 reduction Effects 0.000 claims description 2
- 230000003044 adaptive effect Effects 0.000 claims 1
- 230000004044 response Effects 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000004088 simulation Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000008034 disappearance Effects 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0259—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterized by the response to fault detection
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
A method and a system for detecting faults of an air conditioning system in a subway station hall comprise the following steps: collecting operation data of a subway air conditioner; inputting algorithm parameter values, and initializing and standardizing each value; the latest sequence gradient and the learning rate value are saved through updating iteration; data reverse processing is carried out by taking whether convergence is carried out or not as a judgment condition, and if the convergence condition is not met, improvement is returned; using the optimized parameter value for a forward current output value y (k) in RNN calculation; and calculating an error function step by step, taking the condition that the termination condition reaches the maximum iteration times or reaches the corresponding error requirement as a judgment condition, returning to the weight iteration updating module to recalculate the output value at the current moment if the termination condition does not meet the requirement, and finally outputting the result. The parameters in the existing RNN are optimized, and the magnitude of the sequence gradient parameters and the learning rate transmitted by the previous item are optimized through a self-adaptive optimization algorithm, so that the accuracy and the rapidity in fault detection are improved to a certain extent.
Description
Technical Field
The invention belongs to the technical field of control science and engineering, and particularly relates to a method and a system for detecting faults of an air conditioning system of a subway station hall.
Background
The air conditioning system in the subway station hall has the advantages that as the special underground form of a station is a large and narrow closed structure, equipment continuously operates and space continuously illuminates under the influence of passenger flow and traveling density, so that the failure rate is increased compared with that of a common air conditioning system; piston air generated when a train runs in a tunnel enters a station platform to generate a piston effect, and the air enters the station hall through a vertical shaft to cause interference on the reading of a sensor and bring a blockage of false detection on fault detection; and in consideration of energy-saving effect, except for a train traction system, the air conditioner is a large energy consumption item capable of actively saving energy in the subway. Therefore, the problem of air conditioner fault detection in a subway station hall is researched.
And on the basis of comparing the characteristics of the control loops of the subway air-conditioning system and the traditional air-conditioning system and summarizing and analyzing the control performance of the control loops, establishing an optimized fault model by taking the maximum fault detection rate and the minimum false detection rate as targets. In the selection of the fault detection method, the representation of whether the fault occurs is represented by the error between a predicted value and an actual value of the neural network through comparison between the predicted value and the actual value; in consideration of the time interval of starting and stopping between stations in the subway and the correlation dependency between data, a Recurrent Neural Network (RNN) with a memory layer function is selected.
The existing RNN method has the problems of gradient explosion or disappearance caused by unstable gradients, low fault detection rate and incapability of eliminating misdiagnosis caused by false errors generated by the instantaneity of subway train operation stations due to piston wind.
Disclosure of Invention
The invention aims to provide a method and a system for detecting faults of an air conditioning system in a subway station hall, so as to solve the problems.
In order to achieve the purpose, the invention adopts the following technical scheme:
a fault detection method for an air conditioning system in a subway station hall comprises the following steps:
collecting operation data of a subway air conditioner, carrying out normalization denoising processing on the operation data, and then establishing a fault model based on an RNN algorithm of self-adaptive optimization sequence gradient and learning rate;
inputting algorithm parameter values, and initializing and standardizing each value;
the latest sequence gradient and the learning rate value are saved through updating iteration;
data reverse processing is carried out by taking whether convergence is carried out or not as a judgment condition, and if the convergence condition is not met, improvement is returned;
using the optimized parameter value for a forward current output value y (k) in RNN calculation;
and calculating an error function step by step, taking the condition that the termination condition reaches the maximum iteration times or reaches the corresponding error requirement as a judgment condition, returning to the weight iteration updating module to recalculate the output value at the current moment if the termination condition does not meet the requirement, and finally outputting the result.
Further, the gradient matrix in the RNN is composed of hierarchically distributed nodes, where a node at a higher level is a parent node, a node at a lower level is called a child node, the endmost child node is usually an output node, and the properties of the node are the same as those of the nodes in the tree;
at the node at the ith layer, the calculation equation of the system state is as follows:
h(i)andthe system state of the node and all its parents; in the case where there are a plurality of parent nodes,the system state is combined into a matrix, X is the data input of the node, and if the node has no input, the calculation is not carried out; f is an excitation function or an encapsulated feedforward neural network, which corresponds to a gating algorithm and some depth algorithms; the U, W, b weight coefficients, the weights of all nodes of the recurrent neural network are shared.
Further, the collected operational data: selecting 6-dimensional indexes influencing the air supply temperature value to form a matrix array as network input quantity; the predicted value of the air supply temperature is used as the output quantity.
Further, the values are initialized and normalized: the array is transposed, normalized, and the 1500 groups of data are divided into 5 × 5 × 60 groups.
Further, an error function is calculated step by step: and judging whether the precision is met or the iteration number is reached, the met output sequence error and weight are reduced, the unsatisfied random gradient is reduced, a self-adaptive optimization reduction coefficient is output, and the error is reversely transferred to update the weights U, W and b.
Further, the establishment of the RNN algorithm for adaptively optimizing the sequence gradient and the learning rate comprises the following steps: setting an RNN structure: the input quantity is a 6-dimensional independent variable array, the middle hidden memory layer is a 10-dimensional network with memory, and the output quantity is 1-dimensional air supply temperature; APG-RNN network training; and (3) fitting and predicting: predicting the air supply temperature at the next moment or a future moment through fitting calculation, and pre-detecting faults; carrying out inverse normalization processing on the data: and normalizing the data in advance for higher readability in the data analysis process, performing inverse normalization, and comparing the predicted value with the actual value.
Further, after the comparison is finished, selecting an evaluation index root mean square error MSE and a correlation coefficient R, and analyzing the prediction effect to obtain a corresponding input and output function relation.
Further, a subway station room air conditioning system fault detection system includes
The system comprises an acquisition module, a fault model establishing module and a fault model establishing module, wherein the acquisition module is used for acquiring operation data of a subway air conditioner, performing normalization denoising processing on the operation data and then performing fault model establishing on the basis of an RNN algorithm of self-adaptive optimization sequence gradient and learning rate;
the preprocessing module is used for inputting algorithm parameter values and initializing and standardizing all the values;
the storage module is used for storing the latest sequence gradient and the learning rate value through updating iteration;
the data reverse processing module is used for taking whether convergence is adopted as a judgment condition, and returning improvement if the convergence condition is not met;
the output module is used for using the optimized parameter value for a forward current output value y (k) in RNN calculation;
and the iteration module is used for calculating the error function step by step, taking the condition that the termination condition reaches the maximum iteration times or reaches the corresponding error requirement as a judgment condition, returning to the weight iteration updating module to recalculate the output value at the current moment if the termination condition does not meet the requirement, and finally outputting the result.
Compared with the prior art, the invention has the following technical effects:
the invention optimizes the parameters in the prior RNN, optimizes the magnitude of the sequence gradient parameters and the learning rate transmitted by the previous item through a self-adaptive optimization algorithm, wherein the gradient parameters are specifically the coefficients in the gradient descending process are subjected to self-adaptive processing, so that the accuracy and the rapidity in the sensor fault detection are improved to a certain degree, and the problem of misdiagnosis caused by false errors generated at the subway train operation station time interval due to piston wind to a certain degree can be solved.
Drawings
FIG. 1 is a flow chart of a method calculation;
FIG. 2 is a schematic view of a subject system;
FIG. 3 is a flowchart of an AGP-RNN based fault detection optimization;
FIG. 4 illustrates the addition of different levels of deviation faults to the supply air temperature;
FIG. 5 sensor 5%, 10% offset fault detection comparison;
FIG. 6 comparison of 15%, 20% deviation fault detection for the sensor;
FIG. 7 sensor 25%, 30% offset fault detection comparison;
FIG. 8 blast temperature incorporates varying degrees of drift fault;
FIG. 95% sensor drift fault detection comparison;
FIG. 1010% sensor drift fault detection comparison;
FIG. 1115% sensor drift fault detection comparison;
FIG. 1220% sensor drift fault detection comparison;
FIG. 1325% sensor drift fault detection comparison;
FIG. 1430% sensor drift fault detection comparison.
Detailed Description
The invention is further described below with reference to the accompanying drawings:
the process design of the fault detection of the sensor of the subway station hall based on the APG-RNN is as follows:
(1) selecting input variables and output variables: selecting 6-dimensional indexes influencing the air supply temperature value to form a matrix array as network input quantity; the predicted value of the air supply temperature is used as the output quantity.
(2) Preprocessing data: the array is transposed and normalized, and 1500 groups of data are divided into 5 × 5 × 60 groups, so as to avoid factors such as excessive data size and difficulty in processing.
(3) Setting an RNN structure: the input quantity is 6-dimensional independent variable array, the middle hidden memory layer is a 10-dimensional network with memory, and the output quantity is 1-dimensional air supply temperature.
(4) APG-RNN network training: the training process is shown in fig. 3.
(5) And (3) fitting and predicting: and predicting the air supply temperature at the next moment or a future moment through fitting calculation, and pre-detecting the fault.
(6) Carrying out inverse normalization processing on the data: in order to enable the readability to be strong in the data analysis process, the data are normalized first, and reverse normalization is conducted at the position so that the predicted value and the actual value can be compared.
(7) And (3) error analysis: and selecting an evaluation index root mean square error MSE and a correlation coefficient R, and analyzing the prediction effect.
(8) And obtaining a corresponding input and output function relation.
The summary flow is shown in fig. 3.
The fault detection experiment simulation and result analysis are as follows:
on the basis of the research of the method and the analysis of the detection flow, 1500 groups of data are selected to obtain a comparison curve of a predicted value and an actual value of the air supply temperature, so that the accuracy of the model is determined, and then the other 500 groups of data are detected and 250 groups of deviation faults and drift faults are added respectively.
Detecting and analyzing the deviation fault of the sensor:
500 groups of normal operation data are selected to carry out fault detection simulation experiments, 250 groups of deviation faults with different degrees are introduced from the time 251 to obtain fault data curves with various degrees of 5% -30%. (see attached FIG. 4)
(1) 5% and 10% sensor bias fault detection (see FIG. 5)
Firstly, when 5% deviation fault occurs, the mean square error of air supply temperature prediction based on RNN is 7.94, and the relative error is 0.065; the mean square error of APG-RNN is 1.37, and the relative error is 0.048. It can be seen that the detection capability is still improved along with the increase of the iteration number, and the APG-RNN is improved by 10% at most than the RNN, but the fault detection requirement cannot be met generally.
When 10% deviation fault occurs, the mean square error of the air supply temperature prediction based on RNN is 3.74, and the relative error is 0.015; the mean square error of APG-RNN is 0.95, and the relative error is 0.013. At the moment, the fault degree is increased by 5% on the original basis, and the detection capability is improved by 20% relative to 5% of the same type of faults.
(2) 15% and 20% sensor bias fault detection (see FIG. 6)
Firstly, when 15% deviation fault occurs, the RNN air supply temperature prediction mean square error is 10.92, the relative error is 0.061, when the fault degree is increased, the difference of the detection capability is the same relative to the same type of tiny fault, and the mean square error value changes along with the change of the fault degree; compared with the 10% fault degree detection rate, the APG-RNN fault degree detection rate is improved by 50% to the maximum extent, the mean square error is 0.26, and the relative error is 0.015. The prediction of APG-RNN may meet the fault detection requirements in 15% of fault cases.
When 20% deviation fault occurs, the maximum error value of RNN air supply temperature prediction is 1 ℃, the mean square error is 13.23, and the relative error is 0.029, wherein the mean square error value also changes along with the change of the fault degree; compared with the 15% fault degree detection rate, the APG-RNN fault degree detection rate is improved by 10% to the maximum extent, the mean square error is 0.19, and the relative error is 0.003.
(3) 25% and 30% sensor bias fault detection (see FIG. 7)
Firstly, when 25% deviation fault occurs, the RNN mean square error is 3.26, the relative error is 0.028, and the detection capability is improved by 30% compared with 20% fault of the same type when the fault degree is 25%; the mean square error of APG-RNN is 0.65, and the relative error is 0.003. The failure detection rate of the prediction of the APG-RNN can reach more than 95% under the condition of 25% failure.
When 30% deviation fault occurs, the correlation between the air supply temperature predicted value based on RNN and the air supply temperature predicted value based on APG-RNN and the actual fault value is large, the fault detection requirement can be met, the detection rate can reach more than 97%, and the difference between the two is mainly reflected in response time, the mean square error of RNN is 1.28, and the relative error is 0.002; the APG-RNN mean square error is 0.75, and the relative error is 0.001.
The results are tabulated below:
sensor bias fault detection evaluation
Detecting and analyzing the drift fault of the sensor:
500 groups of normal operation data are selected to carry out fault detection simulation experiments, 250 groups of drift faults with different degrees are introduced from the time 251 to obtain 5% -30% of each fault data curve. (see the attached figure 8)
(1) 5% sensor drift fault detection (see FIG. 9)
When 5% of drift faults occur, the deviation between the air supply temperature predicted value based on the RNN and the air supply temperature predicted value based on the APG-RNN and the actual fault value is large, the mean square error of the RNN is 1.21, and the relative error is 0.035; the mean square error of APG-RNN is 0.51, and the relative error is 0.029. The detection capability of the APG-RNN is improved by 20% compared with the RNN, but the fault detection requirement cannot be met totally.
(2) 10% sensor drift fault detection (see fig. 10)
When 10% of drift faults occur, the RNN mean square error is 0.14, and the relative error is 0.013; the mean square error of APG-RNN is 0.06, and the relative error is 0.007. The detection capability of the APG-RNN is improved by 9 percent compared with that of the RNN. However, the drift fault has a small value, and the fault detection requirement cannot be met totally.
(3) 15% sensor drift fault detection (see fig. 11)
When 15% of drift faults occur, the RNN mean square error is 0.11, and the relative error is 0.006; the mean square error of APG-RNN is 0.04, and the relative error is 0.005. Compared with RNN, the detection capability of APG-RNN is improved by 11%, the deviation of the RNN predicted value is generally 0.3 ℃, and the deviation of the APG-RNN predicted value is generally 0.2 ℃.
(4) 20% sensor drift fault detection (see fig. 12)
When 20% of drift faults occur, the RNN mean square error is 0.104, and the relative error is 0.005; the mean square error of APG-RNN is 0.01, and the relative error is 0.003. Compared with RNN, the detection capability of APG-RNN is improved by 28%, the deviation of the RNN predicted value is generally 0.2 ℃, the fault detection rate of APG-RNN can reach more than 95%, and after a drift fault is introduced, the response is fast, and the safe operation guarantee of the system is realized.
(5) 25% sensor drift fault detection (see FIG. 13)
When 25% of drift faults occur, the RNN mean square error is 0.103, and the relative error is 0.002; the mean square error of APG-RNN is 0.021, and the relative error is 0.001. The detection capability of the APG-RNN is improved by 18% compared with that of the RNN, the response capability of the RNN due to gradient instability under the same fault degree still has problems, and the overall fault detection rate of the APG-RNN can reach more than 96%.
(6) 30% sensor drift fault detection (see FIG. 14)
When 30% of drift faults occur, the correlation between the air supply temperature predicted value based on the RNN and the air supply temperature predicted value based on the APG-RNN and an actual fault value is large, the fault detection requirements can be met, the detection rate can reach more than 97%, and the difference between the two is mainly reflected in response time, the mean square error of the RNN is 0.09, and the relative error is 0.001; APG-RNN mean square error is 0.006, and relative error is 0.001.
The results are tabulated below:
sensor drift fault detection evaluation
The experimental result of comparison in the embodiment of the application shows that the improved AGP-RNN is better than the original RNN in the aspects of model stability and accuracy, and the misdiagnosis caused by false errors generated by the time interval of subway train operation stations due to piston wind is eliminated to a certain extent.
The optimization method provided by the application effectively analyzes the sensor fault detection of the subway station hall, and can provide reference for the design of the underground space air-conditioning control scheme. The foregoing is directed to embodiments of the present application only, and it is understood that those skilled in the art can make further research and improvements without departing from the present principles, and that several professionals are now fully researching a single object, so that the detection capability of a certain point or a certain range can be obviously improved.
Claims (8)
1. A fault detection method for an air conditioning system in a subway station hall is characterized by comprising the following steps:
collecting operation data of a subway air conditioner, carrying out normalization denoising processing on the operation data, and then establishing a fault model based on an RNN algorithm of self-adaptive optimization sequence gradient and learning rate;
inputting algorithm parameter values, and initializing and standardizing each value;
the latest sequence gradient and the learning rate value are saved through updating iteration;
data reverse processing is carried out by taking whether convergence is carried out or not as a judgment condition, and if the convergence condition is not met, improvement is returned;
using the optimized parameter value for a forward current output value y (k) in RNN calculation;
and calculating an error function step by step, taking the condition that the termination condition reaches the maximum iteration times or reaches the corresponding error requirement as a judgment condition, returning to the weight iteration updating module to recalculate the output value at the current moment if the termination condition does not meet the requirement, and finally outputting the result.
2. The method as claimed in claim 1, wherein the gradient matrix in the RNN is composed of hierarchically distributed nodes, wherein the nodes at a higher level are parent nodes, the nodes at a lower level are called child nodes, the endmost child node is usually an output node, and the properties of the nodes are the same as those of the nodes in the tree;
at the node at the ith layer, the calculation equation of the system state is as follows:
h(i)andthe system state of the node and all its parents; in the case where there are a plurality of parent nodes,the system state is combined into a matrix, X is the data input of the node, and if the node has no input, the calculation is not carried out; f is an excitation function or an encapsulated feedforward neural network, which corresponds to a gating algorithm and some depth algorithms; the U, W, b weight coefficients, the weights of all nodes of the recurrent neural network are shared.
3. The method for detecting faults of air conditioning systems in the ground station halls according to claim 1, wherein the collected operation data comprises: selecting 6-dimensional indexes influencing the air supply temperature value to form a matrix array as network input quantity; the predicted value of the air supply temperature is used as the output quantity.
4. The method of claim 1, wherein the values are initialized and normalized by: the array is transposed, normalized, and the 1500 groups of data are divided into 5 × 5 × 60 groups.
5. The method of claim 2, wherein the error function is calculated step by step as: and judging whether the precision is met or the iteration number is reached, the met output sequence error and weight are reduced, the unsatisfied random gradient is reduced, a self-adaptive optimization reduction coefficient is output, and the error is reversely transferred to update the weights U, W and b.
6. The method for detecting faults of air conditioning systems in the subway station halls as claimed in claim 1, wherein the adaptive optimization sequence gradient and learning rate RNN algorithm is established by: setting an RNN structure: the input quantity is a 6-dimensional independent variable array, the middle hidden memory layer is a 10-dimensional network with memory, and the output quantity is 1-dimensional air supply temperature; APG-RNN network training; and (3) fitting and predicting: predicting the air supply temperature at the next moment or a future moment through fitting calculation, and pre-detecting faults; carrying out inverse normalization processing on the data: and normalizing the data in advance for higher readability in the data analysis process, performing inverse normalization, and comparing the predicted value with the actual value.
7. The method for detecting the fault of the air conditioning system in the subway station hall as claimed in claim 6, wherein after the comparison is finished, an evaluation index root mean square error MSE and a correlation coefficient R are selected, and a prediction effect is analyzed to obtain a corresponding input and output function relation.
8. A system for detecting faults of air conditioning systems in a subway station hall is characterized by comprising
The system comprises an acquisition module, a fault model establishing module and a fault model establishing module, wherein the acquisition module is used for acquiring operation data of a subway air conditioner, performing normalization denoising processing on the operation data and then performing fault model establishing on the basis of an RNN algorithm of self-adaptive optimization sequence gradient and learning rate;
the preprocessing module is used for inputting algorithm parameter values and initializing and standardizing all the values;
the storage module is used for storing the latest sequence gradient and the learning rate value through updating iteration;
the data reverse processing module is used for taking whether convergence is adopted as a judgment condition, and returning improvement if the convergence condition is not met;
the output module is used for using the optimized parameter value for a forward current output value y (k) in RNN calculation;
and the iteration module is used for calculating the error function step by step, taking the condition that the termination condition reaches the maximum iteration times or reaches the corresponding error requirement as a judgment condition, returning to the weight iteration updating module to recalculate the output value at the current moment if the termination condition does not meet the requirement, and finally outputting the result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110874186.3A CN113867306A (en) | 2021-07-30 | 2021-07-30 | Fault detection method and system for air conditioning system of subway station hall |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110874186.3A CN113867306A (en) | 2021-07-30 | 2021-07-30 | Fault detection method and system for air conditioning system of subway station hall |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113867306A true CN113867306A (en) | 2021-12-31 |
Family
ID=78990226
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110874186.3A Pending CN113867306A (en) | 2021-07-30 | 2021-07-30 | Fault detection method and system for air conditioning system of subway station hall |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113867306A (en) |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108334059A (en) * | 2018-02-26 | 2018-07-27 | 上海理工大学 | Based on particle cluster algorithm Optimized BP Neural Network Symbolic fault diagnosis method |
CN109931678A (en) * | 2019-03-13 | 2019-06-25 | 中国计量大学 | Air-conditioning fault diagnosis method based on deep learning LSTM |
CN110689171A (en) * | 2019-09-05 | 2020-01-14 | 哈尔滨工程大学 | Turbine health state prediction method based on E-LSTM |
KR102182226B1 (en) * | 2020-07-22 | 2020-11-24 | 경북대학교 산학협력단 | Failure Detection-Diagnosis System and Method using Thereof |
CN112036581A (en) * | 2019-05-15 | 2020-12-04 | 上海杰之能软件科技有限公司 | Performance detection method and device of vehicle air conditioning system, storage medium and terminal |
CN112269367A (en) * | 2020-10-13 | 2021-01-26 | 西安建筑科技大学 | Fault detection method and system based on EEMD-combined neural network air processing system |
KR102279351B1 (en) * | 2020-11-27 | 2021-07-20 | 성한 주식회사 | A fault diagnosis method of based ai on optimization of air conditioner |
-
2021
- 2021-07-30 CN CN202110874186.3A patent/CN113867306A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108334059A (en) * | 2018-02-26 | 2018-07-27 | 上海理工大学 | Based on particle cluster algorithm Optimized BP Neural Network Symbolic fault diagnosis method |
CN109931678A (en) * | 2019-03-13 | 2019-06-25 | 中国计量大学 | Air-conditioning fault diagnosis method based on deep learning LSTM |
CN112036581A (en) * | 2019-05-15 | 2020-12-04 | 上海杰之能软件科技有限公司 | Performance detection method and device of vehicle air conditioning system, storage medium and terminal |
CN110689171A (en) * | 2019-09-05 | 2020-01-14 | 哈尔滨工程大学 | Turbine health state prediction method based on E-LSTM |
KR102182226B1 (en) * | 2020-07-22 | 2020-11-24 | 경북대학교 산학협력단 | Failure Detection-Diagnosis System and Method using Thereof |
CN112269367A (en) * | 2020-10-13 | 2021-01-26 | 西安建筑科技大学 | Fault detection method and system based on EEMD-combined neural network air processing system |
KR102279351B1 (en) * | 2020-11-27 | 2021-07-20 | 성한 주식회사 | A fault diagnosis method of based ai on optimization of air conditioner |
Non-Patent Citations (1)
Title |
---|
张善兴;何金凝;杜志敏;何军;晋欣桥;: "基于遗传算法-误差反传神经网络的变风量空调***空调箱故障诊断", 制冷技术, no. 05, 15 October 2019 (2019-10-15) * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110441065B (en) | Gas turbine on-line detection method and device based on LSTM | |
CN107944648B (en) | Large ship speed and oil consumption rate prediction method | |
CN111815033A (en) | Offshore wind power prediction method based on RCNN and meteorological time sequence characteristics | |
CN109063355A (en) | Near-optimal method based on particle group optimizing Yu Kriging model | |
CN111814956B (en) | Multi-task learning air quality prediction method based on multi-dimensional secondary feature extraction | |
CN111598224B (en) | Air conditioner cold load prediction method for optimizing BP neural network based on simulated annealing algorithm | |
CN113031555B (en) | Intelligent purification system for harmful gas in environment of livestock and poultry house | |
CN109255726A (en) | A kind of ultra-short term wind power prediction method of Hybrid Intelligent Technology | |
CN115016276B (en) | Intelligent water content adjustment and environment parameter Internet of things big data system | |
CN113947182B (en) | Traffic flow prediction model construction method based on dual-stage stacked graph convolution network | |
CN112149883A (en) | Photovoltaic power prediction method based on FWA-BP neural network | |
CN110083125A (en) | A kind of machine tool thermal error modeling method based on deep learning | |
CN112766603A (en) | Traffic flow prediction method, system, computer device and storage medium | |
CN111624522A (en) | Ant colony optimization-based RBF neural network control transformer fault diagnosis method | |
CN114548591A (en) | Time sequence data prediction method and system based on hybrid deep learning model and Stacking | |
CN114397043A (en) | Multi-point temperature intelligent detection system | |
CN115751441A (en) | Heat supply system heating station heat regulation method and system based on secondary side flow | |
CN115907122A (en) | Regional electric vehicle charging load prediction method | |
CN109492816B (en) | Coal and gas outburst dynamic prediction method based on hybrid intelligence | |
CN116894180B (en) | Product manufacturing quality prediction method based on different composition attention network | |
CN117668743A (en) | Time sequence data prediction method of association time-space relation | |
CN117313795A (en) | Intelligent building energy consumption prediction method based on improved DBO-LSTM | |
CN117113086A (en) | Energy storage unit load prediction method, system, electronic equipment and medium | |
CN113867306A (en) | Fault detection method and system for air conditioning system of subway station hall | |
CN114415503A (en) | Temperature big data thing networking detects and intelligent control system |
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 |