CN112613651B - Industrial steam end consumption prediction model building and prediction method and system - Google Patents

Industrial steam end consumption prediction model building and prediction method and system Download PDF

Info

Publication number
CN112613651B
CN112613651B CN202011485863.4A CN202011485863A CN112613651B CN 112613651 B CN112613651 B CN 112613651B CN 202011485863 A CN202011485863 A CN 202011485863A CN 112613651 B CN112613651 B CN 112613651B
Authority
CN
China
Prior art keywords
data
time
consumption
steam
steam consumption
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
CN202011485863.4A
Other languages
Chinese (zh)
Other versions
CN112613651A (en
Inventor
阳赛
王栋
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.)
Shanghai Allsense Technology Co ltd
Original Assignee
Shanghai Allsense Technology 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 Shanghai Allsense Technology Co ltd filed Critical Shanghai Allsense Technology Co ltd
Priority to CN202011485863.4A priority Critical patent/CN112613651B/en
Publication of CN112613651A publication Critical patent/CN112613651A/en
Application granted granted Critical
Publication of CN112613651B publication Critical patent/CN112613651B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24—Querying
    • G06F16/245—Query processing
    • G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
    • G06F16/2465—Query processing support for facilitating data mining operations in structured databases
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24—Querying
    • G06F16/245—Query processing
    • G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
    • G06F16/2474—Sequence data queries, e.g. querying versioned data
    • 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
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P80/00—Climate change mitigation technologies for sector-wide applications
    • Y02P80/10—Efficient use of energy, e.g. using compressed air or pressurized fluid as energy carrier

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Economics (AREA)
  • Computational Linguistics (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • General Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Databases & Information Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Human Resources & Organizations (AREA)
  • Strategic Management (AREA)
  • Marketing (AREA)
  • Biomedical Technology (AREA)
  • Artificial Intelligence (AREA)
  • General Business, Economics & Management (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Fuzzy Systems (AREA)
  • Biophysics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Tourism & Hospitality (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Operations Research (AREA)
  • Primary Health Care (AREA)
  • Water Supply & Treatment (AREA)
  • Public Health (AREA)
  • Quality & Reliability (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Game Theory and Decision Science (AREA)
  • Development Economics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Testing And Monitoring For Control Systems (AREA)

Abstract

The invention discloses a method and a system for establishing and predicting an industrial steam end consumption prediction model, which are characterized in that firstly, data are collected and preprocessed; then constructing a data set according to the processed data and the time characteristics, taking y 't1 in the data set as the output of the neural network model, and taking other values except y' t1 as the input to train the neural network model to obtain a prediction model; and for the future moment needing to be predicted, the data acquired in the earlier stage belongs to the prediction model, and the steam consumption of the future moment can be obtained. The data acquisition thought fully considers the time characteristics of the data, effectively predicts the steam consumption in a future period, can regulate and control the boiler in advance, and achieves the aim of optimizing the operation of the boiler; meanwhile, through active regulation and control, the phenomena of energy waste and insufficient steam supply can be effectively avoided, and energy conservation and emission reduction are realized; the problems of untimely feedback and slow regulation caused by a passive regulation means of the boiler are solved.

Description

Industrial steam end consumption prediction model building and prediction method and system
Technical Field
The invention belongs to the technical field of energy prediction, and relates to an industrial steam end consumption prediction model building and predicting method and system.
Background
The traditional industrial steam quantity prediction mainly focuses on a production end, the steam generation quantity is predicted by analyzing the correlation between the historical working condition data of the boiler and the steam quantity produced, the steam quantity required by the tail end (namely a consumption end) is little in attention, and in fact, the prediction of the industrial steam of the consumption end has important significance for safe and economic operation of a thermal power plant, but the data quality of the consumption end is poorer, the unpredictability is stronger and the technical requirement is higher compared with the steam data of the production end.
The traditional boiler control method is mainly based on PID feedback control, belongs to passive regulation and control, and can be caused by sudden increase of steam demand, high combustion regulation time stagnation and load follow-up delay, influence steam quality, insufficient steam supply and excessive steam supply on the other hand. If the steam consumption needed in the future can be predicted in advance, the load distribution among the boilers can be performed in advance, the optimal operation of the boilers is realized, and the problem of untimely feedback is solved.
The steam consumption is influenced by various factors, including natural factors such as weather, important dates such as holidays, production characteristics of the consumer end industry, operation habits of factory operators and the like, and the influencing factors are complex, so that it is difficult to establish a specific mathematical model to predict the variation trend of the consumption. The traditional prediction technology is mainly based on a time sequence method of statistical theory analysis, is suitable for the condition that consumption has obvious change rule, has a simple structure, but has a short plate in the aspect of feature extraction, has a general prediction effect, is only suitable for processing the problem of small samples like a common SVM, and has slow iteration convergence. Modern prediction methods mainly apply achievements related to the field of artificial intelligence, and have better universality and practicability in complex systems. Steam consumption has obvious time sequence, so a long-term and short-term memory network (LSTM) is widely applied to energy demand prediction, but the time sequence characteristics of consumption change cannot be fully extracted by simply using an LSTM model, and the model prediction precision is low. If the change trend of the future steam consumption can be accurately predicted, the passive regulation can be changed into active regulation, and the energy waste is avoided.
Disclosure of Invention
In order to solve the defects in the prior art, the invention provides an industrial steam end consumption prediction model building and predicting method and system, and solves the problem of low precision of the existing prediction method.
In order to solve the technical problems, the invention adopts the following technical scheme:
A method for modeling prediction of industrial steam end consumption, comprising:
Step 1, acquiring steam consumption data in an hour unit to obtain an hour-level data sequence;
step 2, setting an abnormal value in the data sequence as a missing value NaN, and filling the missing value in the data sequence to obtain preprocessed data;
Step 3, constructing a data set L according to the processed data and the time characteristics, wherein the step comprises the following steps:
Step 3.1, constructing a data set { X' T } according to the processed data,
Wherein y 't1 is the steam consumption at time t, x' t7 is the steam consumption at time t-1, x 't6 is the steam consumption at time t-2, x' t5 is the steam consumption at time t-3, x 't4 is the steam consumption at time t-4, x' t3 is the steam consumption at time t-5, x 't2 is the steam consumption at time t-24, and x' t1 is the difference between the steam consumption at time t-1 and the steam consumption at time t-2;
Step 3.2, constructing a time feature data set { O' T },
Wherein o' t5 is the position of time t in the day, and is an integer of [0,23 ]; o' t1 is the position of the day of week at time t, and the range is an integer of 0, 6; o' t2 is the position of the day of time t in one month, and the range is an integer of [0,30 ]; o 't3 is the position of the week of time t in one month, the range is an integer of [0,4], o' t4 is the position of the month of time t in one year, and the range is an integer of [0,11 ]; o 't6 is whether the national legal holiday is a national legal holiday o' t6 is 1, otherwise is 0; o' t7 is the weather temperature at time t;
Step 3.3, performing one-hot encoding processing on the O' t1～o't6 in the time characteristic data set in step 3.2 to obtain { O t };
Step 3.4, transversely splicing { O t}、o't7、{X'T } to form a dataset L';
step 3.5, deleting the data of the row containing the missing value in the data set L' to obtain the data set L;
And 4, taking y 't1 in the dataset L as output of the neural network model, taking values except y' t1 as input to train the neural network model, and obtaining a prediction model.
Specifically, the method for processing the outlier in the step 2 is as follows: first, a value larger than a set threshold in the data is set as a missing value, and then an abnormal value is set as a missing value according to the 3σ criterion.
Specifically, the missing value filling method in the step 2 is as follows: if two or more values are continuously deleted, the processing is not performed, and the deleted values are reserved; otherwise, taking the average value of 2-5 values before and after the missing value by adopting a moving average window method as the missing value to fill.
Preferably, in the step 3.5, after deleting the data set L' containing the data of the row where the missing value is located, the data set is normalized, and the data is normalized to [0,1].
Specifically, the structure of the neural network model comprises a GRU layer, a Dropout layer, 2 LSTM layers and 2 Dense layers which are sequentially connected.
The invention also discloses a prediction model building system of the industrial steam end consumption, which comprises the following steps:
the data acquisition module is used for acquiring steam consumption data in an hour unit to obtain a data sequence;
the data preprocessing module is used for setting the abnormal value in the data sequence as a missing value and filling the missing value in the data sequence;
A data calculation module for calculating the difference between the steam consumption amounts at adjacent times and representing the position of the time t in one day, one week, one month or one year with different numbers;
and the model training module is used for training the neural network model according to the numerical value obtained by the data calculation module to obtain a prediction model.
Specifically, the structure of the neural network model comprises a GRU layer, a Dropout layer, 2 LSTM layers and 2 Dense layers which are sequentially connected.
The invention also discloses a method for predicting the consumption of the tail end of the industrial steam, which comprises the following steps:
step 1, acquiring steam consumption data at different moments in an hour unit;
Step 2, processing the steam consumption data according to the step 2; these data values are then calculated {ot'1,o't2,o't3,ot'4,ot'5,o't6,o't7,x't1,x't2,x't3,xt'4,x't5,xt'6,x't7};
And 3, encoding and splicing the o ' t1～ot'7 and the x ' t1～x't7 obtained in the step 2, and inputting the encoded and spliced o ' t1～ot'7 and x ' t1～x't7 into a prediction model obtained in the claim 1 to obtain the steam consumption y ' t1 to be predicted.
The invention also discloses a prediction system of the industrial steam end consumption, which comprises a data acquisition module, a data processing module and a prediction module;
the data acquisition module is used for acquiring steam consumption data at different moments in an hour unit;
The data processing module is used for processing the steam consumption data of the data acquisition module; calculating {ot'1,o't2,o't3,ot'4,ot'5,o't6,o't7,x't1,x't2,x't3,xt'4,x't5,xt'6,x't7} data values, and encoding and splicing the data;
the prediction module is used for inputting the data value obtained by the data processing module into the prediction model to predict the steam consumption at the future moment.
Compared with the prior art, the invention has the beneficial effects that:
(1) The data acquisition thought fully considers the time characteristics of the data, and the model trained by the data has high detection precision;
(2) According to the invention, the industrial steam consumption is predicted based on the cyclic neural network model, the steam consumption in a future period is effectively predicted according to the modeling of the historical steam consumption, and the boiler can be regulated and controlled in advance, so that the purpose of optimizing the operation of the boiler is achieved; meanwhile, through active regulation and control, the phenomena of energy waste and insufficient steam supply can be effectively avoided, and energy conservation and emission reduction are realized; the problems of untimely feedback and slow regulation caused by a passive regulation means of the boiler are solved.
Drawings
FIG. 1 is a diagram of the process of obtaining a predictive model and the control concept of a boiler in accordance with the present invention.
FIG. 2 is a graph of the prediction results of the future 24-hour steam consumption by the prediction model in the embodiment of the present invention.
Detailed Description
The term "time feature" as used herein refers to the location of a time of day, week, month, year and is represented by a different number.
The following specific embodiments of the present application are provided, and it should be noted that the present application is not limited to the following specific embodiments, and all equivalent changes made on the basis of the technical solution of the present application fall within the protection scope of the present application.
Example 1
The embodiment discloses a method for establishing a prediction model of industrial steam end consumption, which comprises the following steps:
Step 1, preparing a sample data set: acquiring steam consumption data in an hour unit to obtain an hour-level data sequence; the steam consumption data in this step is historical steam consumption data collected by the system.
In this embodiment, historical steam consumption data of a thermal power plant of a salt-sun is collected, the steam consumption end of the thermal power plant comprises 7 enterprises, steam consumption data in a period from 1 day to 30 days of 4 months to 1 month of 2019 are extracted from a database of a monitoring system, and an hour-level historical steam consumption data sequence is formed, and total of 1349 historical data is obtained.
Step 2, data preprocessing: if the data is missing, the missing value needs to be filled, and then the abnormal value of the whole data after the missing value is filled is processed, so that the preprocessed data is obtained.
The abnormal value detection method comprises the following steps: firstly, setting a value larger than a set threshold value in data as a deletion value NaN, and screening out data which is not in the range of [0,60] in the embodiment;
Then, the outliers are processed according to the 3 sigma criterion, in particular:
Using Bessel formula Calculate standard deviation sigma, wherein,/>Obtaining a mean value for the steam data sequence, wherein n is the number of the data sequences; if a certain consumption data x i exists/>X i is considered to be an outlier.
And treating all the detected abnormal values as a missing value NaN according to a missing value treatment method.
As a preferable aspect of the present embodiment, the missing value filling method is as follows: firstly, detecting the number of the missing values NaN, and if the group of data has two or more values which are continuously missing, not processing, wherein the missing values are reserved as NaN; otherwise, taking the average value of 2-5 values before and after the missing value by adopting a moving average window method as the missing value to fill. In this example, 3 values before and after the deletion value are taken, for example, the deletion value of sequence {x1,x2,x3,x4,x5,NaN,x7,x8,x9,x10}, is filled with (x 3+x4+x5+x7+x8+x9)/6.
Step 3, constructing a data set according to the processed data and the time characteristics, wherein the embodiment specifically comprises the following steps:
Step 3.1, constructing a data set { X' T } according to the processed data,
Wherein y 't1 is the steam consumption at time t, x' t7 is the steam consumption at time t-1, x 't6 is the steam consumption at time t-2, x' t5 is the steam consumption at time t-3, x 't4 is the steam consumption at time t-4, x' t3 is the steam consumption at time t-5, x 't2 is the steam consumption at time t-24, and x' t1 is the difference between the steam consumption at time t-1 and the steam consumption at time t-2;
Step 3.2, generating a series of feature construction data sets { O' T } according to the time (such as time t) to be predicted,
Wherein, o' t5 is the position of the time t in one day and is an integer of [0,23], in this embodiment, 0 represents the 0 point in one day, and other numbers follow in time sequence; o' t1 is the position of the day of week at time t, the range is an integer of [0,6], in this embodiment 0 represents Monday, and other numbers are forward; o' t2 is the position of the day of time t in one month, the range is an integer of [0,30], in the embodiment, 0 represents one number in one month, and other numbers are forward; o' t3 is the position of the week in one month at the time t, the range is an integer of [0,4], 0 represents the first week in one month, and other numbers are forward; o' t4 is the position of the month in one year at the time t, the range is an integer of [0,11], 0 represents the first month in one year, and other numbers are forward; o 't6 is whether the national legal holiday is a national legal holiday o' t6 is 1, otherwise is 0; o' t7 is the weather temperature at time t.
Step 3.3, performing one-hot encoding processing on the time characteristic data set O' t1～o't6 in the step 3.2 to obtain { O t };
And 3.4, transversely splicing { O t}、o't7 and { X 'T } to form a data set L'.
Step 3.5, deleting the data of the row containing the missing value NaN in the L' to obtain a data set L;
Preferably, the step 3.5 further includes: and carrying out normalization processing on the data after deleting the row containing the deletion value NaN, wherein the data is normalized to be [0,1].
In this embodiment, the normalization formula is:
Wherein, X norm is the normalized value, X i is the value before normalization, X min is the minimum value of the column of X i in the dataset L, and X max is the maximum value of the column of X i in the dataset L.
The sample data set obtained after the above processing contains 1315 pieces of data. The sample set was divided into a training set, a validation set and a test set, comprising 850, 300 and 165 pieces of data, respectively.
And 4, taking y 't1 in the normalized dataset in the step 3 as output of the neural network model, and taking other values except y' t1 as input to train the neural network model.
The neural network structure constructed in this embodiment is composed of the following 6 layers in order:
GRU layer: the number of units is 100 and the activation function is relu functions: f (x) =max (0, x), x is input data, and GRU is a variant of LSTM, and is simpler than LSTM layer structure and higher in training efficiency;
Dropout layer: during training, each nerve unit is discarded with a probability of 0.1, so that the situation that some intermediate outputs only depend on some nerve cells on a given training set is avoided, the overfitting of the training set is caused, some nerve cells are randomly turned off, and more nerve cells can participate in the final output.
2 LSTM layers: the number of units is 64 and the activation function is the tanh function: x is input data;
2 Dense layers: the number of units is 32 and 1, respectively;
The loss function is MAE: y real is an actual value, and y pred is a predicted value. The Adam optimizer is adopted, the learning rate is 10 -3, the iteration round number is 360, and the batch size is 72.
The model is trained by using 850 training sets and 300 verification set data of the embodiment, and a trained prediction model is obtained.
Example 2
The embodiment discloses a prediction method of industrial steam end consumption, which predicts the steam consumption at a certain moment in the future by adopting the model obtained in the embodiment 1, and the input end of the prediction model needs to know the steam consumption of the day before the prediction day according to the data set L, so that the method can predict the steam consumption of the next day. The prediction method of the embodiment specifically comprises the following steps:
Step 1, the data in this example is 165 pieces of data of the test set of example 1. Collecting data and processing abnormal values and missing values of the data according to the method from step 1 to step 2 in the embodiment 1; these data values are calculated {ot'1,o't2,o't3,ot'4,ot'5,o't6,o't7,x't1,x't2,x't3,xt'4,x't5,xt'6,x't7}.
And 2, inputting the data obtained in the step 1 into the obtained prediction model of the embodiment 1 after the one-hot coding and the data splicing of the steps 3.3 to 3.5 of the embodiment 1 to obtain the steam consumption y' t1 needing to be predicted. In this embodiment, preferably, after the data is spliced, normalization processing is performed on the data, so as to improve accuracy of the model.
In this embodiment, the value of the steam consumption obtained by using the prediction model is compared with the data actually collected in the test set, and as shown in fig. 1, a comparison chart of the predicted value and the actual value of the steam consumption of 24 hours in the future is obtained by using the model in this embodiment. The average absolute error MAE of the predicted and actual values is 2.4.
Example 3
The embodiment discloses a prediction model building system of industrial steam end consumption, which comprises a data acquisition module, a data preprocessing module, a data calculation module and a model training module.
The data acquisition module is used for acquiring steam consumption data in an hour unit to obtain a data sequence, and a DCS (distributed control system) of the boiler in actual production can store the steam consumption in real time; the data preprocessing module is used for setting an abnormal value in the data sequence as a missing value NaN and filling the missing value in the data sequence; specific filling and removal processes are detailed in step 2 of example 1; the data calculation module is used for calculating the difference between the steam consumption at the adjacent time, such as the difference between the steam consumption at the time t-1 and the steam consumption at the time t-2 in the step 3, and the position of the time t in one day, one week, one month or one year is represented by different numbers, and the specific expression form is shown in the step 3 of the embodiment 1; the model training module is used for training the neural network model according to the numerical value obtained by the data calculation module to obtain a prediction model. The neural network model in the model training module of this embodiment has a structure including a GRU layer, a Dropout layer, 2 LSTM layers, and 2 Dense layers connected in sequence.
Example 4
The embodiment discloses a prediction system of industrial steam end consumption, which comprises a data acquisition module, a data processing module and a prediction module, wherein,
The data acquisition module is used for acquiring steam consumption data at different moments in an hour unit, and the acquisition means are the same as those of the embodiment 3; the data processing module is used for preprocessing the steam consumption data of the data acquisition module, processing abnormal values and filling missing values, and simultaneously calculating {ot'1,o't2,o't3,ot'4,ot'5,o't6,o't7,x't1,x't2,x't3,xt'4,x't5,xt'6,x't7} data values and processing of the data values, such as one-hot coding, data splicing and normalization processing; the prediction module is used for inputting the data value obtained by the data processing module into the prediction model and outputting the steam consumption at a certain moment in the future.
The method and the system can actively regulate and control the boiler in advance according to the predicted steam consumption, and the phenomenon of energy waste or insufficient steam supply is avoided as shown in the figure 1.

Claims (9)

1. A method for modeling prediction of industrial steam end consumption, comprising:
Step 1, acquiring steam consumption data in an hour unit to obtain an hour-level data sequence;
Step 2, setting an abnormal value in the data sequence as a missing value, and then filling the missing value in the data sequence to obtain preprocessed data;
Step 3, constructing a data set L according to the processed data and the time characteristics, wherein the step comprises the following steps:
Step 3.1, constructing a data set { X' T } according to the processed data,
Wherein y 't1 is the steam consumption at time t, x' t7 is the steam consumption at time t-1, x 't6 is the steam consumption at time t-2, x' t5 is the steam consumption at time t-3, x 't4 is the steam consumption at time t-4, x' t3 is the steam consumption at time t-5, x 't2 is the steam consumption at time t-24, and x' t1 is the difference between the steam consumption at time t-1 and the steam consumption at time t-2;
Step 3.2, constructing a time feature data set { O' T },
Wherein o' t5 is the position of time t in the day, and is an integer of [0,23 ]; o' t1 is the position of the day of week at time t, and the range is an integer of 0, 6; o' t2 is the position of the day of time t in one month, and the range is an integer of [0,30 ]; o 't3 is the position of the week of time t in one month, the range is an integer of [0,4], o' t4 is the position of the month of time t in one year, and the range is an integer of [0,11 ]; o 't6 is whether the national legal holiday is a national legal holiday o' t6 is 1, otherwise is 0; o' t7 is the weather temperature at time t;
Step 3.3, performing one-hot encoding processing on the O' t1～o't6 in the time characteristic data set in step 3.2 to obtain { O t };
Step 3.4, transversely splicing { O t}、o't7、{X'T } to form a dataset L';
step 3.5, deleting the data of the row containing the missing value in the data set L' to obtain the data set L;
And 4, taking y 't1 in the dataset L as output of the neural network model, taking values except y' t1 as input to train the neural network model, and obtaining a prediction model.
2. The method for modeling industrial steam end consumption prediction according to claim 1, wherein the processing method of the outlier in step 2 is as follows: first, a value larger than a set threshold in the data is set as a missing value, and then an abnormal value is set as a missing value according to the 3σ criterion.
3. The method for modeling industrial steam end consumption prediction according to claim 1, wherein the missing value filling method in step 2 is as follows: if two or more values are continuously deleted, the processing is not performed, and the deleted values are reserved; otherwise, taking the average value of 2-5 values before and after the missing value by adopting a moving average window method as the missing value to fill.
4. The method for modeling industrial steam end consumption prediction according to claim 1, wherein in step 3.5, after deleting the data in the data set L' containing the row where the missing value is located, the data set is further normalized, and the data is normalized to [0,1].
5. The method for constructing a prediction model of industrial steam end consumption according to claim 1, wherein the neural network model comprises a GRU layer, a Dropout layer, 2 LSTM layers and 2 Dense layers which are sequentially connected.
6. A predictive modeling system for end-of-industrial steam consumption, comprising:
the data acquisition module is used for acquiring steam consumption data in an hour unit to obtain a data sequence;
the data preprocessing module is used for setting the abnormal value in the data sequence as a missing value NaN and filling the missing value in the data sequence;
A data calculation module for calculating the difference between the steam consumption amounts at adjacent times, and representing the positions of the time t in one day, one week, one month and one year with different numbers;
and the model training module is used for training the neural network model according to the numerical value obtained by the data calculation module to obtain a prediction model.
7. The industrial steam end-of-life prediction model building system of claim 6, wherein the neural network model structure comprises a GRU layer, a Dropout layer, 2 LSTM layers, and 2 Dense layers connected in sequence.
8. A method for predicting industrial steam end consumption, comprising:
step 1, acquiring steam consumption data at different moments in an hour unit;
Step 2, processing steam consumption data according to step 2 described in any one of claims 1 to 3; then calculating {o't1,o't2,o't3,o't4,o't5,o't6,o't7,x't1,x't2,x't3,x't4,x't5,x't6,x't7} these data values according to step 3 of any one of claims 1 to 3;
and 3, encoding and splicing the o ' t1～o't7 and the x ' t1～x't7 obtained in the step 2, and inputting the encoded and spliced o ' t1～o't7 and x ' t1～x't7 into a prediction model obtained in the claim 1 to obtain the steam consumption y ' t1 to be predicted.
9. The prediction system of the industrial steam end consumption is characterized by comprising a data acquisition module, a data processing module and a prediction module;
the data acquisition module is used for acquiring steam consumption data at different moments in an hour unit;
The data processing module is used for processing the steam consumption data of the data acquisition module according to the step 2 of any one of claims 1 to 3, then calculating {o't1,o't2,o't3,o't4,o't5,o't6,o't7,x't1,x't2,x't3,x't4,x't5,x't6,x't7} data values according to the step 3 of any one of claims 1 to 3, and encoding and splicing the data;
the prediction module is used for inputting the data value obtained by the data processing module into the prediction model to predict the steam consumption at the future moment.
CN202011485863.4A 2020-12-16 2020-12-16 Industrial steam end consumption prediction model building and prediction method and system Active CN112613651B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011485863.4A CN112613651B (en) 2020-12-16 2020-12-16 Industrial steam end consumption prediction model building and prediction method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011485863.4A CN112613651B (en) 2020-12-16 2020-12-16 Industrial steam end consumption prediction model building and prediction method and system

Publications (2)

Publication Number Publication Date
CN112613651A CN112613651A (en) 2021-04-06
CN112613651B true CN112613651B (en) 2024-05-03

Family

ID=75239524

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011485863.4A Active CN112613651B (en) 2020-12-16 2020-12-16 Industrial steam end consumption prediction model building and prediction method and system

Country Status (1)

Country Link
CN (1) CN112613651B (en)

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2009169930A (en) * 2007-12-21 2009-07-30 Fuji Electric Systems Co Ltd Energy demand predicting device
CN104866725A (en) * 2015-06-01 2015-08-26 河南中烟工业有限责任公司 Method for predicting steam consumption of cigarette enterprise on the basis of least squares support vector machine
CN109670629A (en) * 2018-11-16 2019-04-23 浙江蓝卓工业互联网信息技术有限公司 Coal-burning boiler thermal efficiency forecast method based on shot and long term Memory Neural Networks
CN110363289A (en) * 2019-07-17 2019-10-22 山东浪潮人工智能研究院有限公司 A kind of industrial steam amount prediction technique and device based on machine learning
CN111091232A (en) * 2019-11-25 2020-05-01 黑龙江电力调度实业有限公司 Power load prediction method considering power demand change trend
CN111428906A (en) * 2020-02-17 2020-07-17 浙江大学 Industrial boiler steam quantity prediction method based on image transformation
CN111815054A (en) * 2020-03-31 2020-10-23 浙江大学 Industrial steam heat supply network short-term load prediction method based on big data
CN112016734A (en) * 2020-04-07 2020-12-01 沈阳工业大学 Stack type self-coding multi-model load prediction method and system based on LSTM

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2009169930A (en) * 2007-12-21 2009-07-30 Fuji Electric Systems Co Ltd Energy demand predicting device
CN104866725A (en) * 2015-06-01 2015-08-26 河南中烟工业有限责任公司 Method for predicting steam consumption of cigarette enterprise on the basis of least squares support vector machine
CN109670629A (en) * 2018-11-16 2019-04-23 浙江蓝卓工业互联网信息技术有限公司 Coal-burning boiler thermal efficiency forecast method based on shot and long term Memory Neural Networks
CN110363289A (en) * 2019-07-17 2019-10-22 山东浪潮人工智能研究院有限公司 A kind of industrial steam amount prediction technique and device based on machine learning
CN111091232A (en) * 2019-11-25 2020-05-01 黑龙江电力调度实业有限公司 Power load prediction method considering power demand change trend
CN111428906A (en) * 2020-02-17 2020-07-17 浙江大学 Industrial boiler steam quantity prediction method based on image transformation
CN111815054A (en) * 2020-03-31 2020-10-23 浙江大学 Industrial steam heat supply network short-term load prediction method based on big data
CN112016734A (en) * 2020-04-07 2020-12-01 沈阳工业大学 Stack type self-coding multi-model load prediction method and system based on LSTM

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
基于造纸企业月度数据的行业能耗及节能潜力分析;叶诺等;中国造纸;20160831;第35卷(第8期);35-39 *

Also Published As

Publication number Publication date
CN112613651A (en) 2021-04-06

Similar Documents

Publication Publication Date Title
CN111626506B (en) Regional photovoltaic power probability prediction method based on federal learning and cooperative regulation and control system thereof
CN114022021A (en) Online unit load prediction method based on ensemble learning
CN113107626B (en) Load prediction method of combined cycle generator set based on multivariable LSTM
CN111639783A (en) Line loss prediction method and system based on LSTM neural network
CN111461463A (en) Short-term load prediction method, system and equipment based on TCN-BP
CN112085285A (en) Bus load prediction method and device, computer equipment and storage medium
CN113902207A (en) Short-term load prediction method based on TCN-LSTM
CN115542429A (en) XGboost-based ozone quality prediction method and system
CN114218872A (en) Method for predicting remaining service life based on DBN-LSTM semi-supervised joint model
CN115222106A (en) User day-ahead load prediction method of self-adaptive model
CN114021848A (en) Generating capacity demand prediction method based on LSTM deep learning
CN115755219A (en) Flood forecast error real-time correction method and system based on STGCN
CN116014722A (en) Sub-solar photovoltaic power generation prediction method and system based on seasonal decomposition and convolution network
CN114239385A (en) Intelligent decision making system and method for warehouse resource allocation
CN112613651B (en) Industrial steam end consumption prediction model building and prediction method and system
CN117494906A (en) Natural gas daily load prediction method based on multivariate time series
Zhou et al. Short-term wind power prediction based on EMD-LSTM
CN112232570A (en) Forward active total electric quantity prediction method and device and readable storage medium
CN115081681B (en) Wind power prediction method based on propset algorithm
Prema et al. LSTM based Deep Learning model for accurate wind speed prediction
CN114862032B (en) XGBoost-LSTM-based power grid load prediction method and device
CN116090635A (en) Meteorological-driven new energy generation power prediction method
CN116154875A (en) Photovoltaic power station active power optimization distribution method based on TCN and error function
CN116739130A (en) Multi-time scale load prediction method of TCN-BiLSTM network
CN115470970A (en) Electric quantity prediction method and system based on prophet-LightGBM combined model

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