CN117390948A - Multi-head attention long-short-term memory neural network based water chilling unit monitoring method - Google Patents
Multi-head attention long-short-term memory neural network based water chilling unit monitoring method Download PDFInfo
- Publication number
- CN117390948A CN117390948A CN202311006004.6A CN202311006004A CN117390948A CN 117390948 A CN117390948 A CN 117390948A CN 202311006004 A CN202311006004 A CN 202311006004A CN 117390948 A CN117390948 A CN 117390948A
- Authority
- CN
- China
- Prior art keywords
- water
- head
- chilling unit
- output
- neural network
- 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
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 title claims abstract description 83
- 230000015654 memory Effects 0.000 title claims abstract description 32
- 238000000034 method Methods 0.000 title claims abstract description 30
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 23
- 238000012544 monitoring process Methods 0.000 title claims abstract description 23
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 230000036541 health Effects 0.000 claims description 26
- 239000011159 matrix material Substances 0.000 claims description 12
- 230000007246 mechanism Effects 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000011156 evaluation Methods 0.000 claims description 6
- 238000005057 refrigeration Methods 0.000 claims description 3
- 230000002159 abnormal effect Effects 0.000 claims description 2
- 230000006870 function Effects 0.000 claims description 2
- 238000013507 mapping Methods 0.000 claims description 2
- 238000005259 measurement Methods 0.000 claims description 2
- 238000010606 normalization Methods 0.000 claims description 2
- 230000008569 process Effects 0.000 claims description 2
- 230000009466 transformation Effects 0.000 claims description 2
- 238000012546 transfer Methods 0.000 claims 1
- 230000007787 long-term memory Effects 0.000 abstract description 3
- 238000001816 cooling Methods 0.000 abstract 1
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000013145 classification model Methods 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 238000005265 energy consumption Methods 0.000 description 2
- 239000003507 refrigerant Substances 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
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
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M99/00—Subject matter not provided for in other groups of this subclass
- G01M99/005—Testing of complete machines, e.g. washing-machines or mobile phones
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M99/00—Subject matter not provided for in other groups of this subclass
- G01M99/008—Subject matter not provided for in other groups of this subclass by doing functionality tests
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Medical Informatics (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the technical field of water chilling unit monitoring, and discloses a multi-head attention long-short-time memory neural network-based water chilling unit monitoring method, which comprises the following steps: s1: the data preprocessing of the chiller is carried out, the sensor data sampled at each moment comprises but is not limited to condenser water inlet and outlet temperature, evaporator water inlet and outlet temperature, common water inlet and outlet temperature, building water inlet and outlet temperature and condenser water flow, firstly, the data preprocessing is carried out on multidimensional input dataThe method eliminates null value and invalid characteristic data, divides the data of the water chilling unit into three parts,、、representing input sequences related to power consumption, cooling capacity, and coefficient of performance, respectively. According to the multi-head attention long-short-term memory neural network-based water chiller monitoring method, the time sequence of the multi-section water chiller is processed and output to be in a hidden state by utilizing a plurality of LSTM, so that the long-term memory characteristics can be effectively memorized, and the unimportant characteristics can be forgotten.
Description
Technical Field
The invention relates to the technical field of water chilling unit monitoring, in particular to a water chilling unit monitoring method based on a multi-head attention long-short-time memory neural network.
Background
The water chiller is one of the main energy consumption sources of the data center, if the water chiller fails, the energy consumption of the data center is increased, and the safe and reliable operation of the data center cannot be guaranteed.
For example, the Chinese invention with publication number of CN109446187B discloses a complex equipment health state monitoring method based on an attention mechanism and a neural network, which mainly comprises the following steps: acquiring multi-sensing data of complex equipment; performing feature selection to obtain effective measurement data; preprocessing to obtain a plurality of slice samples; establishing a neural network classification model integrating an attention mechanism and a deep neural network; inputting the slice sample and the label corresponding to the slice sample into a neural network classification model, and off-line training the neural network classification model; the method has the advantages that the local characteristics and time sequence information in the data are fully mined by considering the data characteristics of the multi-sensor signals, the prediction accuracy is high, the applicability is wide, the method can be widely applied to various complex equipment, and the defects still exist;
the application document considers the data characteristics of the multi-sensor signals, fully mines the local characteristics and time sequence information in the data, has high prediction precision and wide applicability, can be widely applied to various complex equipment, but in practical application, the running state of a water chilling unit is unknown, and once the internal failure occurs, the water chilling unit has certain hysteresis, a user cannot be informed of the failure condition in time, and huge economy and other losses can be brought to the user, so that the research on the health monitoring method of the water chilling unit has certain necessity and importance, and the method for monitoring the water chilling unit based on the multi-head attention long-short-time memory neural network is proposed to solve the problems.
Disclosure of Invention
(one) solving the technical problems
Aiming at the defects of the prior art, the invention provides a multi-head attention long-short-time memory neural network-based water chiller monitoring method, which has the advantages of being capable of specifically monitoring the health condition of the water chiller and the like, and solves the problem that the health condition is inconvenient to specifically monitor.
(II) technical scheme
In order to achieve the purpose of specifically monitoring the health condition of the water chilling unit, the invention provides the following technical scheme: a method for monitoring a neural network water chiller based on long-short-term memory of multi-head attention comprises the following steps:
s1: the method comprises the steps of preprocessing chiller data, wherein data sampled at each moment comprise a plurality of chiller characteristic data such as chiller condenser scaling, condenser water flow, non-condensable gas of refrigerant, evaporator water flow, refrigerant capacity, indoor and outdoor temperature, refrigerating capacity output, power consumption, coefficient of performance (COP) and the like, eliminating null value, noise and the like through a data preprocessing method, dividing the chiller data into three parts, and dividing the chiller data into X parts 1 、X 2 、X 3 Respectively representing input sequences related to power consumption, refrigerating capacity and performance coefficient;
s2: respectively by X 1 、X 2 、X 3 As an input sequence of the LSTM, the process of delivering the hidden layer of the LSTM at the moment is as follows: f (f) t ＝sigmoid(W f [h t-1 ,x t ]+b f )，i t ＝sigmoid(W i [h t-1 ,x t ]+b i )，a t ＝tanh(W a [h t-1 ,x t ]+b a )，c t ＝c t-1 e f t +i t e a t ，o t ＝sigmoid(W o [h t-1 ,x t ]+b o )，h t ＝o t e tanh(c t )；
Wherein f t ，i t ，o t Vector values of the forgetting gate, the input gate and the output gate at the time t are respectively W f ，W i ，W o ，W a And b f ，b i ，b o ，b a Coefficient matrix and bias matrix for each gating cell and candidate memory state, a t Is a candidate memory state, c t ，c t-1 And h t-1 The state of the memory unit and the state of the hidden layer at the current moment and the previous moment are obtained, and e represents the corresponding multiplication of elements.
S3: output value of LSTM Delivering the attention to multiple heads, and respectively carrying out 3-head attention calculation, namely H, by a multiple head attention mechanism 1 And power consumption, H 2 And refrigerating capacity, H 3 And the attention calculation between the performance coefficients, each head outputs weights between a plurality of characteristics and respective predicted quantities respectively, and the output weights are spliced and integrated,
the multi-head attention calculation formula is as follows:
MultiHead(Q,K,V)＝Concat(head 1 ,head 2 ,head 3 )W o
wherein Q, K, V represent a query vector, a key vector, and a value vector, W, respectively o The output transformation matrix is represented as such,head i represents the output of the ith header, d k Is the dimension of the key vector, K T Is the transpose of the key vector K, ">Respectively trainable weight parameters;
s4: the multi-head attention mechanism outputs a power consumption predicted value, a refrigerating capacity predicted value and a performance coefficient predicted value, judges the health condition of the water chilling unit through a health evaluation criterion,
the health assessment criteria are as follows:wherein (1)>And->Respectively the predicted values of power consumption, refrigerating capacity and performance coefficient;
when epsilon is smaller than or equal to a threshold χ, the water chilling unit works normally; when epsilon is larger than a threshold value χ, the abnormal operation of the water chilling unit is indicated, a user is reminded of checking the situation in time, and χ is a manually set value and can be set according to experience.
Preferably, in the step S2 f t ，i t ，o t Vector values of the forgetting gate, the input gate and the output gate at the time t are respectively W f ，W i ，W o ，W a And b f ，b i ，b o ，b a Coefficient matrix and bias matrix for each gating cell and candidate memory state, a t Is a candidate memory state, c t ，c t-1 And h t-1 The state of the memory unit and the state of the hidden layer at the current moment and the previous moment are obtained, and e represents the corresponding multiplication of elements.
Preferably, in the step S3Is the ith head pairInputting coefficient matrix for linear mapping, d k Is the dimension of the fitness vector, softmax is the normalization function.
Preferably, in the step S4Respectively, a power consumption predicted value, a refrigerating capacity predicted value and a coefficient of performance predicted value, < >>And->The standard refrigerating capacity and the coefficient of performance of the water chilling unit are respectively.
(III) beneficial effects
Compared with the prior art, the invention provides a multi-head attention long-short-time memory neural network-based water chiller monitoring method, which has the following beneficial effects:
1. according to the multi-head attention long-short-term memory neural network-based water chiller monitoring method, the time sequence of the multi-section water chiller is processed and output to be in a hidden state by utilizing a plurality of LSTM, so that the long-term memory characteristics can be effectively memorized, and the unimportant characteristics can be forgotten.
2. According to the multi-head attention long-short-term memory neural network-based water chiller monitoring method, the attention value is calculated from three different health indexes (power consumption, refrigerating capacity and performance coefficient) of the water chiller respectively through a multi-head attention mechanism, and important characteristics affecting the three health indexes are enhanced, so that the prediction performance is improved.
3. According to the multi-head attention long-short-term memory neural network-based water chiller monitoring method, the difference between the predicted value and the true value is jointly calculated from three different health dimensions of the water chiller through establishing the water chiller health evaluation criterion, so that the health condition of the water chiller can be comprehensively evaluated, a user can be better guided to know the working condition, and the smaller the data of the health evaluation criterion is, the healthier the water chiller is indicated.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a graph showing the comparison of the predicted and actual refrigeration capacity values according to the present invention;
FIG. 3 is a graph showing the comparison of predicted and actual values of power consumption according to the present invention;
FIG. 4 is a comparison of the predicted and actual COP values of the present invention.
Detailed Description
The technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The beneficial effects of the invention are as follows: according to the multi-head attention long-short-term memory neural network based water chilling unit monitoring method, the time series of the multi-section water chilling unit is processed by utilizing a plurality of LSTMs to output hidden states, long-term memory characteristics can be effectively memorized, unimportant characteristics are forgotten, attention values are calculated from three different health indexes (power consumption, refrigerating capacity and performance coefficient) of the water chilling unit through a multi-head attention mechanism, important characteristics affecting the three health indexes are strengthened, so that prediction performance is improved, the health condition of the water chilling unit is more comprehensively evaluated through establishing a health evaluation criterion of the water chilling unit, the health condition of the water chilling unit can be better guided to know working conditions, the smaller the health evaluation criterion is, the more healthy the water chilling unit is represented, namely, hidden information closely related to the health indexes of the water chilling unit in various time sequence data is focused by the multi-head attention, the weight affecting larger parameters is strengthened by the long-short-time memory neural network, multi-dimensional sensing data of the water chilling unit is taken as input data, the health index closely related characteristics of the water chilling unit are found, and the prediction accuracy of the health indexes of the water chilling unit is improved.
The present embodiment uses input data X 1 、X 2 、X 3 The system is 19-dimensional sensor data, and comprises condenser water inlet and outlet temperature, evaporator water inlet and outlet temperature, common heat exchanger water inlet and outlet temperature, building water inlet and outlet temperature, condenser water flow, evaporator water flow, steam supply small-sized electronic valve, steam supply large-sized electronic valve, electronic three-way mixing valve, condenser valve position, city water supply valve position, city water inlet and outlet temperature and hot water inlet and outlet temperature, output data are respectively refrigerating capacity, power consumption and COP, the predicted value and the actual value of model training are respectively compared by fig. 2, 3 and 4, and the fact that the predicted refrigerating capacity, power consumption and COP of the model training are relatively accurate can be seen from the figure, so that the current running state of the water chilling unit is good.
Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (4)
1. The method for monitoring the neural network water chiller based on the long-short-term memory of the multi-head attention is characterized by comprising the following steps of:
s1: the method comprises the steps of preprocessing water chilling unit data, wherein sensor data sampled at each moment comprise, but are not limited to, condenser water inlet and outlet temperature, evaporator water inlet and outlet temperature, common water inlet and outlet temperature, building water inlet and outlet temperature, condenser water flow, evaporator water flow, condenser electronic valve water loop, electronic three-way mixing valve, urban water supply electronic valve, hot water loop electronic valve, evaporator water loop, steam supply small electronic valve, steam supply large electronic valve, watt sensor measuring instantaneous power, evaporator water flow rate, refrigeration capacity output and power consumption, selecting as much as possible of water chilling unit characteristic data as input of a prediction model, taking refrigeration capacity output, power consumption and energy efficiency ratio (COP) as output of a prediction model, firstly, carrying out a data preprocessing method on multidimensional input data, eliminating null value and invalid characteristic data, dividing the data of the water chilling unit into three parts, and X 1 、X 2 、X 3 Respectively representing input sequences related to power consumption, refrigerating capacity and performance coefficient;
s2: respectively by X 1 、X 2 、X 3 As an input sequence of the LSTM, the hidden layer transfer process of the LSTM at the time t is as follows: f (f) t ＝sigmoid(W f [h t-1 ,x t ]+b f )，i t ＝sigmoid(W i [h t-1 ,x t ]+b i )，a t ＝tanh(W a [h t-1 ,x t ]+b a )，c t ＝c t- 1 e f t +i t e a t ，o t ＝sigmoid(W o [h t-1 ,x t ]+b o )，h t ＝o t e tanh(c t )；
Wherein x is t Is the input sequence at time t, f t ，i t ，o t Vector values of the forgetting gate, the input gate and the output gate at the time t are respectively W f ，W i ，W o ，W a And b f ，b i ，b o ，b a Coefficient matrix and bias matrix for each gating cell and candidate memory state, a t Is a candidate memory state, c t ，c t-1 And h t-1 Is the state of the memory unit and the state of the hidden layer at the current moment and the previous moment, h t The memory state is output at the moment t, and e represents the corresponding multiplication of elements;
s3: output value of LSTMDelivering the attention to multiple heads, and respectively carrying out 3-head attention calculation, namely H, by a multiple head attention mechanism 1 And power consumption, H 2 And refrigerating capacity, H 3 Attention calculation with coefficient of performance, wherein +.>Is the memory state of LSTM output at the j-th head t moment, each head outputs a plurality of characteristics and each headPre-measuring weights among the measurements, splicing and integrating the output weights,
the multi-head attention calculation formula is as follows:
MultiHead(Q,K,V)＝Concat(head 1 ,head 2 ,head 3 )W o
wherein Q, K, V represent a query vector, a key vector, and a value vector, W, respectively o The output transformation matrix is represented as such,head i represents the output of the ith header, d k Is the dimension of the key vector, K T Is the transpose of the key vector K, ">Respectively trainable weight parameters;
s4: the multi-head attention mechanism outputs a power consumption predicted value, a refrigerating capacity predicted value and a performance coefficient predicted value, judges the health condition of the water chilling unit through a health evaluation criterion,
the health assessment criteria are as follows: respectively a power consumption predicted value, a refrigerating capacity predicted value and a coefficient of performance predicted value;
when epsilon is smaller than or equal to a threshold χ, the water chilling unit works normally; when epsilon is larger than a threshold value χ, the abnormal operation of the water chilling unit is indicated, a user is reminded of checking the situation in time, and χ is a manually set value and can be set according to experience.
2. The multi-head attention-based system of claim 1The method for monitoring the neural network water chiller based on the long-time memory of force is characterized by comprising the following steps of: f in the step S2 t ，i t ，o t Vector values of the forgetting gate, the input gate and the output gate at the time t are respectively W f ，W i ，W o ，W a And b f ，b i ，b o ，b a Coefficient matrix and bias matrix for each gating cell and candidate memory state, a t Is a candidate memory state, c t ，c t-1 And h t ，h t-1 The state of the memory unit and the state of the hidden layer at the current moment and the previous moment are obtained, and e represents the corresponding multiplication of elements.
3. The multi-head attention long-short-term memory neural network based water chiller monitoring method according to claim 1, wherein the method comprises the following steps of: in the step S3Is the coefficient matrix of the i-th head for linear mapping of the input, d k Is the dimension of the fitness vector, softmax is the normalization function.
4. The multi-head attention long-short-term memory neural network based water chiller monitoring method according to claim 1, wherein the method comprises the following steps of: in the step S4Respectively a power consumption predicted value, a refrigerating capacity predicted value and a coefficient of performance predicted value,and->The standard refrigerating capacity and the coefficient of performance of the water chilling unit are respectively.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311006004.6A CN117390948A (en) | 2023-08-10 | 2023-08-10 | Multi-head attention long-short-term memory neural network based water chilling unit monitoring method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311006004.6A CN117390948A (en) | 2023-08-10 | 2023-08-10 | Multi-head attention long-short-term memory neural network based water chilling unit monitoring method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117390948A true CN117390948A (en) | 2024-01-12 |
Family
ID=89465529
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311006004.6A Pending CN117390948A (en) | 2023-08-10 | 2023-08-10 | Multi-head attention long-short-term memory neural network based water chilling unit monitoring method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117390948A (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108592352A (en) * | 2018-05-09 | 2018-09-28 | 国家计算机网络与信息安全管理中心 | The prediction technique of water cooling air conditioning system operating status health degree, apparatus and system |
| CN112163426A (en) * | 2020-09-30 | 2021-01-01 | 中国矿业大学 | Relationship extraction method based on combination of attention mechanism and graph long-time memory neural network |
| CN112990258A (en) * | 2021-02-01 | 2021-06-18 | 山东建筑大学 | Fault diagnosis method and system for water chilling unit |
| CN115345344A (en) * | 2022-06-30 | 2022-11-15 | 浙江工业大学 | Sewage treatment plant effluent quality prediction method based on attention mechanism and long-short term memory network |
| CN115508119A (en) * | 2022-09-23 | 2022-12-23 | 河南科技大学 | Residual error-knowledge-data combined driving fault composite diagnosis method and system |
| CN115689014A (en) * | 2022-10-31 | 2023-02-03 | 北京工业大学 | Water quality index prediction method based on bidirectional long-and-short-term memory neural network and time attention mechanism |
| WO2023045278A1 (en) * | 2021-09-27 | 2023-03-30 | 西安交通大学 | Data dual-drive method, apparatus, and device for predicting power grid failure during typhoon |
| CN115932947A (en) * | 2022-12-05 | 2023-04-07 | 吉林大学 | Microseism P wave S wave separation method based on CNN-LSTM and multi-head attention mechanism |
-
2023
- 2023-08-10 CN CN202311006004.6A patent/CN117390948A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108592352A (en) * | 2018-05-09 | 2018-09-28 | 国家计算机网络与信息安全管理中心 | The prediction technique of water cooling air conditioning system operating status health degree, apparatus and system |
| CN112163426A (en) * | 2020-09-30 | 2021-01-01 | 中国矿业大学 | Relationship extraction method based on combination of attention mechanism and graph long-time memory neural network |
| CN112990258A (en) * | 2021-02-01 | 2021-06-18 | 山东建筑大学 | Fault diagnosis method and system for water chilling unit |
| WO2023045278A1 (en) * | 2021-09-27 | 2023-03-30 | 西安交通大学 | Data dual-drive method, apparatus, and device for predicting power grid failure during typhoon |
| CN115345344A (en) * | 2022-06-30 | 2022-11-15 | 浙江工业大学 | Sewage treatment plant effluent quality prediction method based on attention mechanism and long-short term memory network |
| CN115508119A (en) * | 2022-09-23 | 2022-12-23 | 河南科技大学 | Residual error-knowledge-data combined driving fault composite diagnosis method and system |
| CN115689014A (en) * | 2022-10-31 | 2023-02-03 | 北京工业大学 | Water quality index prediction method based on bidirectional long-and-short-term memory neural network and time attention mechanism |
| CN115932947A (en) * | 2022-12-05 | 2023-04-07 | 吉林大学 | Microseism P wave S wave separation method based on CNN-LSTM and multi-head attention mechanism |
Non-Patent Citations (1)
| Title |
|---|
| 高凯 等: "基于多头注意力的 CNN-LSTM的换道意图预测", 《机械工程学报》, 30 November 2022 (2022-11-30), pages 369 - 378 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110555479B (en) | Fault feature learning and classifying method based on 1DCNN and GRU fusion | |
| Li et al. | An explainable one-dimensional convolutional neural networks based fault diagnosis method for building heating, ventilation and air conditioning systems | |
| CN114484731B (en) | Central air conditioner fault diagnosis method and device based on stacking fusion algorithm | |
| Han et al. | PCA-SVM-based automated fault detection and diagnosis (AFDD) for vapor-compression refrigeration systems | |
| Hydeman et al. | Development and testing of a reformulated regression-based electric chiller model/discussion | |
| CN113792762A (en) | Water chilling unit fault diagnosis method, system and medium based on Bayesian optimization LightGBM | |
| CN113435593B (en) | Refrigeration equipment frosting prediction method based on sensor time sequence data analysis | |
| Li et al. | Diagnosis for multiple faults of chiller using ELM-KNN model enhanced by multi-label learning and specific feature combinations | |
| CN112884012A (en) | Building energy consumption prediction method based on support vector machine principle | |
| Li et al. | Distance measures in building informatics: An in-depth assessment through typical tasks in building energy management | |
| CN109711049B (en) | Hybrid subway air conditioner water chilling unit energy efficiency estimation method | |
| CN113591406B (en) | Heat pipe cooling reactor measuring point optimal arrangement and fault diagnosis method and system | |
| CN113432247B (en) | Water chilling unit energy consumption prediction method and system based on graph neural network and storage medium | |
| Liang et al. | The impact of improved PCA method based on anomaly detection on chiller sensor fault detection | |
| CN117390948A (en) | Multi-head attention long-short-term memory neural network based water chilling unit monitoring method | |
| CN114117852A (en) | Regional heat load rolling prediction method based on finite difference working domain division | |
| Li et al. | Improved convolutional neural network chiller early fault diagnosis by gradient-based feature-level model interpretation and feature learning | |
| CN114626655A (en) | Multi-standard comprehensive evaluation method for regional comprehensive energy system | |
| CN112747418A (en) | Air conditioner and cloud server | |
| CN116933025A (en) | Transformer top layer oil temperature prediction method based on VMD and DBO-LSTM-AT | |
| CN114678962B (en) | Distributed array temperature measurement abnormal data transmission monitoring system based on power internet of things | |
| Hu et al. | Prediction of heat transfer of vacuum glass based on intelligent algorithm modeling | |
| CN113357842B (en) | CO (carbon monoxide) 2 Transcritical parallel compression refrigeration system and control method | |
| CN114707684A (en) | Improved LSTM-based raw tobacco stack internal temperature prediction algorithm | |
| CN117079736B (en) | Gas concentration prediction method and system for intelligent gas sensing |
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 |