CN112307650B - Multi-step prediction method for ultra-supercritical boiler heating surface pipe wall overtemperature early warning - Google Patents

Multi-step prediction method for ultra-supercritical boiler heating surface pipe wall overtemperature early warning Download PDF

Info

Publication number
CN112307650B
CN112307650B CN202011352342.1A CN202011352342A CN112307650B CN 112307650 B CN112307650 B CN 112307650B CN 202011352342 A CN202011352342 A CN 202011352342A CN 112307650 B CN112307650 B CN 112307650B
Authority
CN
China
Prior art keywords
data
heating surface
input
time series
wall temperature
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.)
Active
Application number
CN202011352342.1A
Other languages
Chinese (zh)
Other versions
CN112307650A (en
Inventor
楼玉民
钟葳
吴燕玲
鲍听
赵宁宁
严小华
赵炜炜
李望
黄一君
雪小峰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang Energy Group Research Institute Co Ltd
Original Assignee
Zhejiang Energy Group Research Institute Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Zhejiang Energy Group Research Institute Co Ltd filed Critical Zhejiang Energy Group Research Institute Co Ltd
Priority to CN202011352342.1A priority Critical patent/CN112307650B/en
Publication of CN112307650A publication Critical patent/CN112307650A/en
Application granted granted Critical
Publication of CN112307650B publication Critical patent/CN112307650B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/08—Thermal analysis or thermal optimisation

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Data Mining & Analysis (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Computational Linguistics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • Control Of Steam Boilers And Waste-Gas Boilers (AREA)

Abstract

The invention relates to a multi-step prediction method for ultra-temperature early warning of a pipe wall of a heating surface of an ultra-supercritical boiler, which comprises the following steps: step 1, training data are collected in a training stage, a time series network model is constructed after the collected training data are preprocessed, and off-line training is carried out; and 2, acquiring input data in the use stage as the input of the time series network model to obtain the prediction result of the metal wall temperature data. The invention has the beneficial effects that: the temperature of the tube wall of the heating surface can be predicted in advance, and guidance is provided for overtemperature control; longer time sequence characteristic information can be captured, and prediction precision is improved; by carrying out sinusoidal coding on the time, the transition is smooth every day, and the time is more consistent with the actual operation condition; the problem that the sensor cannot work for a long time when directly measuring can be solved, and the wall temperature prediction at different positions can be realized; the latest data can be conveniently added, the model can be retrained, and the anti-interference performance is good.

Description

Multi-step prediction method for ultra-supercritical boiler heating surface pipe wall overtemperature early warning
Technical Field
The invention belongs to the technical field of boiler safety, and particularly relates to a multi-step prediction method for ultra-temperature early warning of a heating surface pipe wall of an ultra-supercritical boiler.
Background
With the continuous improvement of the power demand and the energy-saving and emission-reducing requirements, the steam temperature and the steam pressure of the boiler are continuously improved, and the ultra-supercritical boiler and the superheater work in a severe environment, so that the overtemperature pipe explosion is more easily caused.
The temperature of the pipe wall of the heating surface of the ultra-supercritical boiler is high, the flow of flue gas is large, the temperature measuring sensor cannot measure the temperature of the metal pipe wall for a long time due to poor working conditions, the temperature change of the pipe wall has the characteristics of hysteresis, nonlinearity, nonstationness and strong coupling, and the wall temperature change trend of a traditional linear model and a non-time sequence model is difficult to accurately predict in advance; therefore, a soft measurement monitoring method is urgently needed.
At present, research on monitoring of the temperature of the tube wall of the heating surface of the ultra-supercritical boiler mainly focuses on monitoring the temperature at the current moment in real time, and a multi-step advance prediction method is lacked. In patent No. 201010183756.6, a series of thermocouple sensors are arranged on pipelines to collect wall temperature signals of the heating surface of the boiler, so that the layout of the sensors is optimized; however, the inside of the ultra-supercritical boiler has a bad environment and high temperature, and the sensor cannot work for a long time and only can monitor the current wall temperature. In the patent with the patent number of 201611123426.1, flue gas and steam water variables are collected through a boiler DCS system or an SIS system, and the metal wall temperature is calculated through an empirical formula. Therefore, the establishment of a multi-step advanced monitoring method for the tube wall temperature of the heating surface of the ultra-supercritical boiler is particularly important.
Disclosure of Invention
The invention aims to overcome the defects in the prior art and provides a multi-step prediction method for ultra-temperature early warning of the pipe wall of a heating surface of an ultra-supercritical boiler.
The multi-step prediction method for the ultra-supercritical boiler heating surface pipe wall overtemperature early warning comprises the following steps:
step 1, collecting training data in a training stage, preprocessing the collected training data, then constructing a time series network model, and finally performing offline training on the time series network model;
step 1.1, data are exported from a DCS (distributed control system) data acquisition system or an SIS (SIS) data acquisition system of a boiler, temperature measurement sensors are additionally arranged in regions which are easy to overtemperature to measure the metal wall temperature data of a heating surface, and the metal wall temperature data of the heating surface is used as a target to be predicted;
step 1.2, preprocessing the training data and the heating surface metal wall temperature data collected in the step 1, using the heating surface metal wall temperature data as the output of a time series network, and using the preprocessed training data as the input of the time series network; the data preprocessing mode is feature coding; the characteristic codes are soot blowing variable category characteristic codes and time characteristic codes; the soot blowing characteristic variable value is 1 when soot blowing operation is performed, the soot blowing characteristic variable value is 0 when soot blowing operation is not performed, the time characteristic is the acquisition time t of each piece of data, and the coding mode is sin (t/24 x pi); ensuring that the transition between the 23 point and the 0 point on the next day is smooth; the feature dimensionality reduction adopts a PCA dimensionality reduction method, and N features [ x ] are reserved after feature dimensionality reduction1,x2,...xN]；
Step 1.3, constructing a time series network model, wherein the time series network model comprises an input layer, a multi-layer decreasing LSTM network, a full-connection neural network and an output layer; the input layer is connected with a multi-layer decreasing LSTM network; in the multi-layer decreasing LSTM network, layers are connected in series and decreased, and the output of the last layer of LSTM network is used as the input of the fully-connected neural network; the output of the full-connection neural network is connected with an output layer;
step 1.4, dividing the data preprocessed in the step 1.2 into a training set and a verification set, taking K historical moment data samples as input of a time series network, taking M steps of metal wall temperature data at P metal wall temperature positions to be predicted of a heating surface as output of the time series network, performing off-line training of a time series network model, and finally storing the trained model weight;
step 2, collecting input data in a use stage, preprocessing the collected input data, and then inputting the preprocessed input data as a time series network model to obtain a prediction result of metal wall temperature data;
step 2.1, acquiring input data of a use stage, acquiring the input data by a boiler DCS (distributed control System) data acquisition system or an SIS (SIS) data acquisition system, preprocessing the acquired input data in the same preprocessing mode as that in the step 1.2, and taking the preprocessed data as the input of a time series network model;
2.2, extracting a hysteresis influence factor in a multi-layer serial decreasing LSTM mode, gradually discarding time characteristics at a longer distance, and increasing the importance of the historical time closer to the predicted time; the time series network model outputs the prediction results of M steps of metal wall temperature data at P metal wall temperature positions to be predicted of the heating surface, and P different metal wall temperature positions to be predicted are correspondingly connected with different fully-connected neural networks, so that the wall temperature results at different positions are separated earlier, and the interference between different positions is avoided.
Preferably, in step 1.3: inputting the characteristics corresponding to the K historical moment data samples in sequence to form an input layer, and placing the moment sample with the farthest time in the input layer at the first time; the output of the upper layer of the LSTM network in the multi-layer decreasing LSTM network is used as the input of the next layer of the LSTM network after the first output is removed; p fully-connected neural networks are connected in parallel, P is the number of the metal wall temperature positions of the heating surface to be predicted, the number of the outputs of each fully-connected neural network is M, and M is the predicted time step number at the same position of the heating surface; and the output layer is the future M-step wall temperature values of P heating surface metal wall temperature positions to be predicted.
Preferably, each layer of LSTM network in the step 1.3 consists of a forgetting gate, an input gate and an output gate; the iteration mode of the LSTM is as follows:
Figure GDA0002996983540000031
in the above formula, ft、it、ot、ct、htRespectively showing forgetting gate, input gate, output gate and hidden stateState quantity, state carrier,. sigma.denotes a sigmoid activation function, tanh denotes a double tangent activation function, W and U denote weight matrices, b denotes an offset matrix,. phi.denotes multiplication of corresponding elements, x denotestIs the t-th feature.
Preferably, the data derived from the DCS data acquisition system or SIS data acquisition system of the boiler in step 1.1 includes: flue gas side data, steam water side data and equipment data.
Preferably, the training loss function is used when the time series network model is trained offline in step 1.4.
Preferably, the temperature measuring sensor in step 1.1 is a thermocouple sensor with a protective tube, so that the sensor can acquire more target tags within a limited life.
The invention has the beneficial effects that:
1) according to the method, the time sequence correlation is learned through an LSTM time sequence network, the multi-step prediction of the pipe wall temperature of the heating surface is realized, the pipe wall temperature of the heating surface is predicted in advance, and guidance is provided for overtemperature control.
2) Because the characteristics of the heated surface pipe wall temperature prediction have hysteresis, the invention adopts a multilayer series connection decreasing LSTM mode to extract hysteresis influence factors, gradually discards time characteristics at longer distance, and has larger importance when the historical time is closer to the prediction time, thereby capturing longer time sequence characteristic information and improving the prediction precision.
3) The invention solves the sudden change influence of soot blowing operation on the metal wall temperature by carrying out characteristic coding on the soot blowing discrete variable; by carrying out sinusoidal coding on the time, the transition is smooth every day, and the time is more consistent with the actual operation condition.
4) If the wall temperature is directly measured by the thermocouple under the high-temperature working condition, the sensor can only work for a short time, but the invention only uses the supplementary temperature measuring sensor to measure the wall temperatures of a plurality of positions which are easy to exceed the temperature as a training target in the training stage, thereby establishing the rule of the wall temperature and the historical data characteristics, and the heating surface pipe wall temperature measuring sensor is not needed to be used in the use stage, thereby not only solving the problem that the sensor can not work for a long time when being directly measured, but also realizing the wall temperature prediction of different positions.
5) The data acquisition mode of the invention directly adopts the existing DCS or SIS of the power plant, which is convenient for the transformation and the upgrade of the DCS or SIS of the power plant, and can conveniently increase the latest data and retrain the model, thereby having good anti-interference performance.
Drawings
FIG. 1 is a flow chart of a multi-step prediction method of the present invention;
fig. 2 is a time-series network structure composition diagram of the multi-step prediction method.
Description of reference numerals: the system comprises an input layer 1, a multi-layer decreasing LSTM network 2, a fully connected neural network 3 and an output layer 4.
Detailed Description
The present invention will be further described with reference to the following examples. The following examples are set forth merely to aid in the understanding of the invention. It should be noted that, for a person skilled in the art, several modifications can be made to the invention without departing from the principle of the invention, and these modifications and modifications also fall within the protection scope of the claims of the present invention.
The invention establishes the influence relation between the metal pipe wall temperature and the time sequence input variable through the time sequence prediction network, thereby achieving the purpose of predicting the wall temperature in advance.
As an embodiment, as shown in FIG. 1, the multi-step prediction method for the ultra supercritical boiler heating surface tube wall overtemperature early warning includes a training phase and a using phase.
The training phase comprises the following steps:
step 1, training data is collected, and the following data are derived from a DCS data acquisition system or an SIS data acquisition system of a boiler: flue gas side data such as superheater inlet flue gas temperature and flue gas oxygen content, and steam-water side data such as superheater outlet header temperature and main steam pressure; primary air flow, generator power and other equipment data; measuring heating surface metal wall temperature data by additionally arranging thermocouple sensors with protective tubes at 200 positions easy to overtemperature, wherein the heating surface metal wall temperature data is used as a target to be predicted;
step 2, aiming at 14400 training data samples collected in step 1 (the time interval is 3 min)One sample) and the metal wall temperature data of the heating surface at 200 positions easy to overtemperature are subjected to preprocessing such as feature coding, missing value processing, maximum and minimum value normalization, feature dimension reduction and the like; taking the metal wall temperature data of the heating surface measured by the thermocouple sensor as the output of the time series network, and taking the data after training data preprocessing as the input of the time series network; the characteristic codes are soot blowing variable category characteristic codes and time characteristic codes; the soot blowing characteristic variable value is 1 when soot blowing operation is performed, the soot blowing characteristic variable value is 0 when soot blowing operation is not performed, the time characteristic is the acquisition time t of each piece of data, and the coding mode is sin (t/24 x pi); ensuring that the transition between the 23 point and the 0 point on the next day is smooth; the feature dimensionality reduction adopts a PCA dimensionality reduction method, and N features [ x ] are reserved after feature dimensionality reduction1,x2,...xN]，N＝512；
Step 3, as shown in fig. 2, constructing a time series network model, wherein the time series network model comprises an input layer 1, a three-layer decreasing LSTM network 2, a full-connection neural network 3 and an output layer 4; inputting the characteristics corresponding to the 4 time samples in sequence to form an input layer 1, placing the time sample with the farthest time in the input layer 1 at the first time, and connecting the input layer 1 with a multi-layer decreasing LSTM network 2; in the three-layer decreasing LSTM network 2, layers are connected in series in a decreasing manner, the output of the upper layer of LSTM network in the three-layer decreasing LSTM network 2 is used as the input of the next layer of LSTM network after the first output is removed, and the output of the last layer of LSTM network is used as the input of the fully connected neural network 3; 200 fully-connected neural networks 3 are connected in parallel; the output number of each fully-connected neural network 3 is 2, and the number of each neural network neuron in the fully-connected neural network 3 is 256; the output of the full-connection neural network 3 is connected with an output layer 4; the output layer 4 is the future 2-step wall temperature values of 200 heating surface metal wall temperature positions to be predicted;
and 4, constructing a time series network model aiming at the step 3, dividing the data preprocessed in the step 2 into a training set and a verification set according to the proportion of 7:3, taking K historical moment data samples as the input of the time series network, taking M steps of metal wall temperature data at P metal wall temperature positions to be predicted on a heating surface as the output of the time series network, selecting a root mean square error loss function to perform time series network model off-line training, and finally storing the trained model weight.
The using stage comprises the following steps:
step 1, collecting input data of a use stage, collecting the input data by a boiler DCS (distributed control System) data collection system or an SIS (SIS) data collection system, preprocessing the collected input data by adopting the same preprocessing mode as that in step 1.2, and taking the preprocessed data as the input of a time series network model;
and 2, outputting the prediction results of the metal wall temperature data in the M steps at P metal wall temperature positions to be predicted of the heating surface by the time series network model, wherein P different metal wall temperature positions to be predicted are correspondingly connected with different fully-connected neural networks, so that the wall temperature results at different positions are separated earlier, and the interference between different positions is avoided.

Claims (6)

1. A multi-step prediction method for ultra-temperature early warning of pipe walls of heating surfaces of ultra-supercritical boilers is characterized by comprising the following steps:
step 1, collecting training data in a training stage, preprocessing the collected training data, then constructing a time series network model, and finally performing offline training on the time series network model;
step 1.1, data are exported from a DCS (distributed control system) data acquisition system or an SIS (SIS) data acquisition system of a boiler, temperature measurement sensors are additionally arranged in regions which are easy to overtemperature to measure the metal wall temperature data of a heating surface, and the metal wall temperature data of the heating surface is used as a target to be predicted;
step 1.2, preprocessing the training data and the heating surface metal wall temperature data collected in the step 1, using the heating surface metal wall temperature data as the output of a time series network, and using the preprocessed training data as the input of the time series network; the data preprocessing mode is feature coding; the characteristic codes are soot blowing variable category characteristic codes and time characteristic codes; the soot blowing characteristic variable value is 1 when soot blowing operation is performed, the soot blowing characteristic variable value is 0 when soot blowing operation is not performed, the time characteristic is the acquisition time t of each piece of data, and the coding mode is sin (t/24 x pi); PCA dimension reduction method for feature dimension reductionAnd after the dimension of the feature is reduced, keeping N features [ x ]1,x2,...xN]；
Step 1.3, constructing a time series network model, wherein the time series network model comprises an input layer (1), a multi-layer decreasing LSTM network (2), a full-connection neural network (3) and an output layer (4); the input layer (1) is connected with a multi-layer decreasing LSTM network (2); in the multilayer decreasing LSTM network (2), layers are connected in series and decreased gradually, and the output of the last layer of LSTM network is used as the input of the fully-connected neural network (3); the output of the full-connection neural network (3) is connected with an output layer (4);
step 1.4, dividing the data preprocessed in the step 1.2 into a training set and a verification set, taking K historical moment data samples as input of a time series network, taking M steps of metal wall temperature data at P metal wall temperature positions to be predicted of a heating surface as output of the time series network, performing off-line training of a time series network model, and finally storing the trained model weight;
step 2, collecting input data in a use stage, preprocessing the collected input data, and then inputting the preprocessed input data as a time series network model to obtain a prediction result of metal wall temperature data;
step 2.1, acquiring input data of a use stage, acquiring the input data by a boiler DCS (distributed control System) data acquisition system or an SIS (SIS) data acquisition system, preprocessing the acquired input data in the same preprocessing mode as that in the step 1.2, and taking the preprocessed data as the input of a time series network model;
and 2.2, outputting the prediction results of the metal wall temperature data in the M steps at P metal wall temperature positions to be predicted of the heating surface by the time series network model, wherein the P different metal wall temperature positions to be predicted are correspondingly connected with different fully-connected neural networks.
2. The multi-step prediction method for the ultra-supercritical boiler heating surface pipe wall overtemperature early warning as claimed in claim 1 is characterized in that in step 1.3: inputting the characteristics corresponding to the K historical time data samples in sequence to form an input layer (1), and placing the time sample with the farthest time in the input layer (1) at the first time; the output of the upper layer of the multi-layer decreasing LSTM network (2) is used as the input of the next layer of the LSTM network after the first output is removed; p fully-connected neural networks (3) are connected in parallel, P is the number of the metal wall temperature positions of the heating surface to be predicted, the output number of each fully-connected neural network (3) is M, and M is the predicted time step number of the same position of the heating surface; the output layer (4) is the future M-step wall temperature values of P heating surface metal wall temperature positions to be predicted.
3. The multi-step prediction method for the ultra-supercritical boiler heating surface pipe wall overtemperature early warning as claimed in claim 1 is characterized in that: in the step 1.3, each layer of LSTM network consists of a forgetting gate, an input gate and an output gate; the iteration mode of the LSTM is as follows:
Figure FDA0002996983530000021
in the above formula, ft、it、ot、ct、htRespectively representing a forgetting gate, an input gate, an output gate, a hidden state quantity and a state carrier, sigma represents a sigmoid activation function, tanh represents a double tangent activation function, W and U represent weight matrices, b represents an offset matrix, a represents multiplication of corresponding elements, x represents a linear multiplication of corresponding elementstIs the t-th feature.
4. The multi-step prediction method for the ultra-supercritical boiler heating surface pipe wall overtemperature warning as claimed in claim 1 is characterized in that the data derived from the DCS data acquisition system or the SIS data acquisition system of the boiler in step 1.1 comprises: flue gas side data, steam water side data and equipment data.
5. The multi-step prediction method for the ultra-supercritical boiler heating surface pipe wall overtemperature early warning as claimed in claim 1 is characterized in that: and (4) adopting a training loss function when the time series network model is trained off line in the step 1.4.
6. The multi-step prediction method for the ultra-supercritical boiler heating surface pipe wall overtemperature early warning as claimed in claim 1 is characterized in that: in the step 1.1, the temperature measuring sensor is a thermocouple sensor with a protective tube.
CN202011352342.1A 2020-11-27 2020-11-27 Multi-step prediction method for ultra-supercritical boiler heating surface pipe wall overtemperature early warning Active CN112307650B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011352342.1A CN112307650B (en) 2020-11-27 2020-11-27 Multi-step prediction method for ultra-supercritical boiler heating surface pipe wall overtemperature early warning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011352342.1A CN112307650B (en) 2020-11-27 2020-11-27 Multi-step prediction method for ultra-supercritical boiler heating surface pipe wall overtemperature early warning

Publications (2)

Publication Number Publication Date
CN112307650A CN112307650A (en) 2021-02-02
CN112307650B true CN112307650B (en) 2021-05-11

Family

ID=74487818

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011352342.1A Active CN112307650B (en) 2020-11-27 2020-11-27 Multi-step prediction method for ultra-supercritical boiler heating surface pipe wall overtemperature early warning

Country Status (1)

Country Link
CN (1) CN112307650B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113111519B (en) * 2021-04-16 2024-07-16 南京金创有色金属科技发展有限公司 Method for judging overtemperature interval and time of tube explosion of water-cooled wall of 20G boiler
CN113359425A (en) * 2021-07-06 2021-09-07 浙江浙能技术研究院有限公司 Thermal power plant boiler main steam temperature intelligent control system based on LSTM neural network PID optimization
CN113761794B (en) * 2021-08-17 2024-05-31 浙江浙能技术研究院有限公司 Boiler soot blowing optimization method for supplementing pollution factors based on time sequence prediction algorithm
CN113806981B (en) * 2021-09-16 2024-07-05 浙江衡玖医疗器械有限责任公司 Water temperature prediction method and device for hemispherical ultrasonic imaging system and electronic equipment

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101476734A (en) * 2009-01-16 2009-07-08 陕西午禾科技有限责任公司 Intelligent ash blowing control system based on sound wave gas temperature field measuring technique
CN103672846A (en) * 2013-12-03 2014-03-26 广东电网公司电力科学研究院 Method for monitoring wall temperature of heating surface and thickness of scale cinder of boiler superheater or reheater
KR101931875B1 (en) * 2017-04-26 2018-12-21 주식회사 씨브이네트 System and method for servicing a intelligent heating
CN109583585A (en) * 2018-11-22 2019-04-05 西安热工研究院有限公司 A kind of station boiler wall temperature prediction neural network model

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20200033515A (en) * 2018-09-20 2020-03-30 한국전자통신연구원 Machine learning method/apparatus based on multiple features extraction and transfer learning and apparatus for leak detection using same
CN110531797A (en) * 2019-05-31 2019-12-03 华电国际电力股份有限公司技术服务分公司 Extra-supercritical unit high temperature superheater wall temperature prediction technique neural network based

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101476734A (en) * 2009-01-16 2009-07-08 陕西午禾科技有限责任公司 Intelligent ash blowing control system based on sound wave gas temperature field measuring technique
CN103672846A (en) * 2013-12-03 2014-03-26 广东电网公司电力科学研究院 Method for monitoring wall temperature of heating surface and thickness of scale cinder of boiler superheater or reheater
KR101931875B1 (en) * 2017-04-26 2018-12-21 주식회사 씨브이네트 System and method for servicing a intelligent heating
CN109583585A (en) * 2018-11-22 2019-04-05 西安热工研究院有限公司 A kind of station boiler wall temperature prediction neural network model

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
A Deep Learning Approach for Heating and Cooling Equipment Monitoring;Yunli Wang等;《2019 IEEE 15th International Conference on Automation Science and Engineering (CASE)》;20190826;228-234 *
基于LSTM的锅炉四管高温再热器超温预测分析;李友志等;《无线互联科技》;20190930(第18期);103-104 *

Also Published As

Publication number Publication date
CN112307650A (en) 2021-02-02

Similar Documents

Publication Publication Date Title
CN112307650B (en) Multi-step prediction method for ultra-supercritical boiler heating surface pipe wall overtemperature early warning
CN109446187B (en) Method for monitoring health state of complex equipment based on attention mechanism and neural network
CN110441065B (en) Gas turbine on-line detection method and device based on LSTM
CN111275367B (en) Regional comprehensive energy system energy efficiency state evaluation method
CN109491358B (en) Control performance monitoring method for boiler dynamic information of million-kilowatt ultra-supercritical unit
CN113837464A (en) Load prediction method of cogeneration boiler based on CNN-LSTM-Attention
CN113868966B (en) Optimization design method for structural parameters of biomass boiler economizer
CN113486950B (en) Intelligent pipe network water leakage detection method and system
Liu et al. Model fusion and multiscale feature learning for fault diagnosis of industrial processes
CN106322412A (en) Coal-fired unit convection heating surface intelligent soot blowing method based on two-dimensional optimization
Yin et al. Multi-feature-scale fusion temporal convolution networks for metal temperature forecasting of ultra-supercritical coal-fired power plant reheater tubes
CN106289754B (en) A kind of fired power generating unit component capabilities variation determination method based on Statistical Control Technique
CN115510904A (en) Boiler heating surface ash deposition monitoring method based on time sequence prediction
CN113722656B (en) Real-time health evaluation method and system for thermal generator set
CN117829342A (en) Coal-fired power plant nitrogen oxide emission prediction method based on improved random forest algorithm
CN113705888A (en) Industrial steam generation amount prediction method and system based on Pearson correlation and neural network
CN115113519A (en) Coal-gas co-combustion boiler denitration system outlet NO x Concentration early warning method
CN114817847B (en) Energy storage power station intelligent monitoring method based on multilayer feedforward neural network
CN114548701B (en) Full-measurement-point-oriented coupling structure analysis and estimation process early warning method and system
CN114897435A (en) One-stage furnace operation abnormity detection method based on instant neighbor reconstruction error
CN113756886A (en) Fault judgment method and system for steam turbine heater
Yin et al. Modal decomposition integrated model for ultra-supercritical coal-fired power plant reheater tube temperature multi-step prediction
CN111307493B (en) Knowledge-based fault diagnosis method for tower type solar molten salt heat storage system
CN117267707A (en) Reheater with temperature monitoring function
Tang et al. Deep neural network based the oxygen content of boiler flue gas

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant