CN112561118B - Municipal pipe network water flow prediction method based on GRU neural network - Google Patents

Municipal pipe network water flow prediction method based on GRU neural network Download PDF

Info

Publication number
CN112561118B
CN112561118B CN202011174683.4A CN202011174683A CN112561118B CN 112561118 B CN112561118 B CN 112561118B CN 202011174683 A CN202011174683 A CN 202011174683A CN 112561118 B CN112561118 B CN 112561118B
Authority
CN
China
Prior art keywords
data
flow
pipe network
neural network
municipal pipe
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
CN202011174683.4A
Other languages
Chinese (zh)
Other versions
CN112561118A (en
Inventor
郭毅
刘帅
王亿宝
郝二成
刘伟岩
孙云辉
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.)
Beijing Shuihui Intelligent Technology Co ltd
Beijing Enterprises Water China Investment Co Ltd
Original Assignee
Beijing Shuihui Intelligent Technology Co ltd
Beijing Enterprises Water China Investment 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 Beijing Shuihui Intelligent Technology Co ltd, Beijing Enterprises Water China Investment Co Ltd filed Critical Beijing Shuihui Intelligent Technology Co ltd
Priority to CN202011174683.4A priority Critical patent/CN112561118B/en
Publication of CN112561118A publication Critical patent/CN112561118A/en
Application granted granted Critical
Publication of CN112561118B publication Critical patent/CN112561118B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—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
    • 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
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0635—Risk analysis of enterprise or organisation activities
    • 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
    • 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
    • G06Q50/10—Services
    • G06Q50/26—Government or public services
    • 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
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A20/00—Water conservation; Efficient water supply; Efficient water use
    • Y02A20/152—Water filtration

Landscapes

  • Business, Economics & Management (AREA)
  • Engineering & Computer Science (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Strategic Management (AREA)
  • General Physics & Mathematics (AREA)
  • Tourism & Hospitality (AREA)
  • Health & Medical Sciences (AREA)
  • General Business, Economics & Management (AREA)
  • Marketing (AREA)
  • General Health & Medical Sciences (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Development Economics (AREA)
  • Game Theory and Decision Science (AREA)
  • Educational Administration (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Primary Health Care (AREA)
  • Data Mining & Analysis (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • Evolutionary Computation (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Public Health (AREA)
  • Water Supply & Treatment (AREA)
  • Feedback Control In General (AREA)

Abstract

The invention relates to a municipal pipe network water flow prediction method based on a GRU neural network. The method comprises the following steps: a water plant central control system is used as a data source (pump pit liquid level, flow meter entering the plant); cleaning the data through a data preprocessing module, and aligning the factory-entering flow data and the pump pit liquid level data by using the cleaned data to prepare a neat and error-free data source for subsequent processing; after the data are processed, pushing the processed data to a municipal pipe network data calculation module, and estimating the flow of the municipal pipe network by the data through corresponding transformation by the module; then, the estimated data is pushed to an autoregressive correlation coefficient module to calculate an autoregressive order to obtain the input dimension of the GRU neural network, and the GRU neural network is generated; and finally, carrying out prediction analysis on the municipal pipe network flow through a GRU neural network. The visual display device has the characteristics of reasonable design, simplicity and convenience in operation, strong visualization, reliable and stable performance and capability of being widely popularized and used.

Description

Municipal pipe network water flow prediction method based on GRU neural network
Technical Field
The invention relates to a municipal sewage pipe network flow prediction technology, in particular to a municipal pipe network flow prediction method based on a GRU neural network.
Background
After ten items of water are published, social and government sustainable development puts higher requirements on urban sewage treatment plants: the quality of the effluent is continuously improved, and the requirements of water pollution control and sustainable recycling of water resources are met; energy conservation, consumption reduction, carbon emission control and low-carbon operation realization; and recycling, developing and utilizing resources and energy. The sewage treatment industry has insufficient overall automation and informatization, and the biological treatment process of the process core is a complex system with multiple factors and large lag; meanwhile, the problems of large fluctuation of the quality and quantity of inlet water, asynchronous fluctuation of each index and the like are faced, so that the traditional method is difficult to realize fine management.
At present, informatization, the Internet of things and big data are one of the most popular research fields and strategic emerging industries in the world. The method is expected to realize technical breakthrough and innovation in the fields of environmental protection, engineering technology, homeland security and the like. The application of data science and technology in the field of water affairs can be regarded as a brand-new start for establishing a 'datamation, precision and intellectualization' water environment comprehensive treatment mode.
The effect of elevator pump is that the sewage that comes from municipal pipe network in the pump pit pours into sewage treatment plant into, and the aperture of elevator pump has not only decided sewage treatment plant daily output but also has decided the liquid level of sewage in the pump pit. Optimal control of the lift pump not only maximizes the daily output of the sewage plant, but also ensures that the pump pit level operates at a higher level (without causing overflow). Municipal pipe network water that comes gets into the pump hole from the water inlet, when the liquid level exceeded minimum liquid level, the intake pump was opened, drew water from the pump house and gets into the sewage factory, and when the liquid level was less than minimum liquid level, the intake pump was shut down. Obviously the liquid level has decided the intake pump to the lift of water plant house steward in the pump pit, and intake pump power consumption is directly proportional to the lift, and the higher power consumptive big more that the lift height is higher promptly. Conversely, a higher safety level is more likely to cause flooding. In order to realize optimal control and intelligent lifting, the key is to predict the sewage flow of the municipal pipe network.
Disclosure of Invention
Aiming at the characteristic of unstable liquid level control of the prior pump pit, the invention aims to: the stable operation of the control of the lift pump is improved, the energy saving effect of the lift pump and the overflow risk reducing effect can be achieved by maintaining the liquid level of the pump pit at a certain level, and in order to achieve the purpose, the invention adopts the following technical scheme:
the method comprises the steps that a water plant central control system is used as a source of two data, namely pump pit liquid level and a flow meter entering a plant, the data are cleaned through a data preprocessing module, abnormal data such as recording errors and breakpoints are removed, the data quality is improved to be in a usable state, then the cleaned data align incoming flow data and pump pit liquid level data to prepare a neat and error-free data source for subsequent processing, the data are pushed to a municipal pipe network data computing module after being processed, the municipal pipe network flow is estimated through corresponding transformation of the data through the data computing module, then the estimated data are pushed to an autoregressive correlation coefficient module to compute an autoregressive order to obtain an input dimension of a GRU neural network, and the GRU neural network is generated; and finally, carrying out prediction analysis on the municipal pipe network flow through a GRU neural network. The specific operation steps are as follows:
1. collecting the flow of a main pipe of the water plant, the liquid level of a pump pit and the bottom area of the pump pit through a central control system of the water plant;
2. preprocessing the collected data, adopting a method of replacing the average value of the first 10 data for missing data or abnormal data, and if the missing value is data in the first 10 data, sequentially and gradually pushing the starting point of the data backwards until the 10 continuous data are abnormal;
3. carrying out sliding average on the data of the total pipe flow and the liquid level with the time window length of n;
4. aligning the data according to a specific time window;
5. the municipal pipe network flow is calculated by utilizing the processed water plant header pipe flow and pump pit liquid level data:
Figure BDA0002748371380000021
wherein f is o Is the flow rate of the water plant main pipe (entering the plant), f s The flow rate of the municipal pipe network is determined as l is the liquid level of the pump pit and s is the basal area of the pump pit, and meanwhile, the flow rate of the water plant main pipe and the flow rate of the municipal pipe network are assumed to be linearly changed within one minute, and the flow rate unit is m 3 /h。
According to the above formula, the left and right sides are changed to have:
Figure BDA0002748371380000022
according to the formula (2), when f s (t-1) when the flow rate f of the pipe network is known, the flow rate f of the pipe network is determined at any moment s (t) can be calculated. The key to applying equation (2) is the acquisition of the value of the pipe network traffic at the initial moment. When the time difference between the current time and the later time is smaller, the change of the flow (whether the flow of the pipe network or the flow of the water plant main pipe) at the time and the later time is smaller, the current sampling time interval is small enough, and the flow at the time and the later time can be considered to be the same. Thus, there are:
Figure BDA0002748371380000023
in this pair f s (1) Taking:
Figure BDA0002748371380000031
thus obtaining a series of estimated values f of the municipal pipe network sewage flow according to the formulas (2) and (4) s (t), obviously f s (t) is a time series.
6. Calculating the autoregressive correlation coefficient of the municipal pipe network flow sequence obtained in the first 5 steps, taking the number of coefficients within the range of 70% of the maximum autocorrelation coefficient value as a regression order, and defining the k-order covariance as follows:
Figure BDA0002748371380000032
where n is the time series length, k is 0,1,2, …, n-1 is the dependent time step, f s (t) is the value at time t of the time series,
Figure BDA0002748371380000033
is f s Then the k-th order autocorrelation function is:
Figure BDA0002748371380000034
7. establishing a neural network by using a GRU neural network unit, wherein the GRU unit is as follows:
Figure BDA0002748371380000035
wherein
z t ＝σ(W z X t +U z h t-1 ) (8)
Figure BDA0002748371380000036
r t ＝σ(W r X t +U r h t-1 ) (10)
Where tanh is the hyperbolic tangent function, h t Is the hidden state at the current time, h t-1 For hidden states transmitted at the last moment, X t Is the state at the current time, z t To control the gating of updates, r t Sigma is a sigmoid function for reset gating, and can transform data into a value in the range of 0-1 by the function to serve as a gating signal z t . After the gating signal is obtained, reset gating is first used to obtain the data r after "reset t ⊙h t-1 Then r is further reduced t ⊙h t-1 And input X t Splicing, and then scaling the data to the range of-1 to 1 through a tanh activation function to obtain
Figure BDA0002748371380000037
U，U r ，U z ，W，W r ，W z The parameter to be trained is initialized in a random initialization manner, which represents the product of corresponding elements of the matrix. Randomly dividing original data into training sets and testing sets according to a uniform distribution at a ratio of 7:3, and adopting 128 training setsAnd training the data by a batch random gradient method, calculating the error of the model on the test set every 15 times of iteration, stopping training when the error of the model on the test set is worse than the last result, and using the parameter trained by the last model as the parameter of the model.
Drawings
FIG. 1 is a calculation flow chart of a municipal pipe network water flow prediction method based on a GRU neural network.
FIG. 2 is a graph showing the comparison of the original header pipe flow data and the flow data after the moving average of the municipal pipe network water flow prediction method based on the GRU neural network.
FIG. 3 is a municipal pipe network flow data autocorrelation function coefficient graph presumed by a municipal pipe network water flow prediction method based on a GRU neural network.
FIG. 4 is a graph comparing training results and actual values of a municipal pipe network water flow prediction method based on a GRU neural network.
FIG. 5 is a graph comparing test results with actual values of a municipal pipe network water flow prediction method based on a GRU neural network.
Detailed Description
The following describes a municipal pipe network water flow prediction method based on a GRU neural network in detail with reference to the accompanying fig. 1 to 5, and the performance of the method is described with reference to a specific embodiment.
The method comprises the steps that a water plant central control system is used as a source of two data, namely pump pit liquid level and an incoming flow meter, the data are cleaned through a data preprocessing module, abnormal data such as recording errors and breakpoints are removed, the data quality is improved to be in a usable state, then the cleaned data are used for aligning incoming flow data and pump pit liquid level data to prepare a neat data source for subsequent processing, the data are pushed to a municipal pipe network data computing module after being processed, the municipal pipe network data computing module estimates municipal pipe network flow through corresponding transformation, then the estimated data are pushed to an autoregressive correlation coefficient module to compute an autoregressive order to obtain input dimensionality of a GRU neural network, and the GRU neural network is generated; and finally, carrying out prediction analysis on the municipal pipe network flow through a GRU neural network. The specific operation steps are as follows:
1. collecting the flow of a main pipe of the water plant, the liquid level of a pump pit and the bottom area s of the pump pit through a central control system of the water plant;
2. preprocessing the collected data, adopting a method of replacing the average value of the first 10 data for missing data or abnormal data, and if the missing value is data in the first 10 data, sequentially and gradually pushing the starting point of the data backwards until the 10 continuous data are abnormal;
3. carrying out sliding average on the data of the total pipe flow and the liquid level with the time window length of 5;
4. the data is aligned according to a specific time window, which in this embodiment is 1min, and the purpose of aligning two different data is as follows:
TABLE 1 data not processed according to time windows
Figure BDA0002748371380000051
TABLE 2 processed data according to time window
Figure BDA0002748371380000052
It can be seen from table 1 that there are two data in the total pipe flow data from 5 th 48 th to 5 th 59 th), and the liquid level has only one data, so the average of the two data is the total pipe flow of 5 th 48 th, and the liquid level of 5 th 48 th is the unique value, meanwhile, there is no data in the total pipe flow 5 th 49 th, the 49 th data when the total pipe flow and the pump pit liquid level 5 are removed keep data alignment, generally, the method of removing data is used in the last data, and the first 10 window averages are used for replacing the middle missing values, which aims to ensure that the continuity of the time series data and the front-back dependency are not damaged.
5. The municipal pipe network flow is calculated by utilizing the processed water plant header pipe flow and pump pit liquid level data:
Figure BDA0002748371380000053
wherein f is o Is the water plant main pipe (entering plant) flow rate, f s The flow rate of the municipal pipe network is determined as l is the liquid level of the pump pit and s is the basal area of the pump pit, and meanwhile, the flow rate of the water plant main pipe and the flow rate of the municipal pipe network are assumed to be linearly changed within one minute, and the flow rate unit is m 3 /h。
According to the above formula, the left and right sides are changed to have:
Figure BDA0002748371380000061
according to the formula (2), when f s (t-1) when the flow rate f of the pipe network is known, the flow rate f of the pipe network is determined at any moment s (t) can be calculated. The key to applying equation (2) is the acquisition of the value of the pipe network traffic at the initial moment. When the time interval of the current later moment is smaller, the flow (whether the flow of the pipe network or the flow of the water plant main pipe) of the previous moment and the next moment is changed less, and when the time interval is small enough, the flow of the previous moment and the next moment can be considered to be the same. Thus, there are:
Figure BDA0002748371380000062
in this pair f s (1) Taking:
Figure BDA0002748371380000063
therefore, a series of estimated values f of the municipal pipe network sewage flow can be obtained according to the formulas (12) and (14) s (t), obviously f s (t) is a time series.
6. Calculating the autoregressive correlation coefficient of the municipal pipe network flow sequence obtained in the first 5 steps, and taking the number of coefficients within the range of 70% of the maximum autocorrelation coefficient value as a regression order of 27, so that the input data dimensionality of the neural network is 27:
Figure BDA0002748371380000064
where n is the time series length, k is 0,1,2, …, n-1 is the dependent time step, f s (t) is the value at time t of the time series,
Figure BDA0002748371380000065
is f s Then the k-th order autocorrelation function is:
Figure BDA0002748371380000066
7. constructing a neural network by using the units of the GRU neural network:
Figure BDA0002748371380000067
wherein
z t ＝σ(W z X t +U z h t-1 ) (18)
Figure BDA0002748371380000071
r t ＝σ(W r X t +U r h t-1 ) (20)
Where tanh is the hyperbolic tangent function, h t Hidden state at the present moment, h t-1 For hidden states transmitted at the last moment, X t Is the state at the current time, z t To control the gating of updates, r t Sigma is a sigmoid function for reset gating, and can transform data into a value ranging from 0 to 1 through the function so as to serve as a gating signal z t . After the gating signal is obtained, the data r after reset is obtained by first using reset gating t ⊙h t-1 Then r is further reduced t ⊙h t-1 And input X t Splicing, and zooming the data to-1 by a tanh activation functionWithin the range to obtain
Figure BDA0002748371380000072
U，U r ，U z ，W，W r ，W z The parameter to be trained is initialized in a random initialization manner, which represents the product of corresponding elements of the matrix. The neural network training process includes randomly dividing original data into a training set and a testing set according to the proportion of 7:3 and the uniform distribution, training by a batch random gradient method of 128 data based on the training set, iteratively calculating the error of the model on the testing set every 15 times, stopping training when the error of the model on the testing set is worse than the last result, and using the parameter of the last model training as the parameter of the model.

Claims (4)

1. A municipal pipe network water amount prediction method based on a GRU neural network is characterized by comprising the following steps: the method comprises the following steps of taking a water plant central control system as a source of two data, namely pump pit liquid level and pump pit flow meter, cleaning the data through a data preprocessing module, aligning the flow data entering the plant and the pump pit liquid level data with the cleaned data to prepare a neat data source for subsequent processing, processing the data, pushing the data to a municipal pipe network data calculating module, estimating the flow of the municipal pipe network through corresponding transformation by the module, pushing the estimated data to an autoregressive correlation coefficient module to calculate an autoregressive order to obtain an input dimension of a GRU neural network, and generating the GRU neural network; finally, the municipal pipe network flow is subjected to predictive analysis through a GRU neural network;
estimating the municipal pipe network flow through corresponding change of the processed data, and realizing the method according to the following formula
Figure FDA0003692297670000011
Wherein f is o Is the flow of the water plant main pipe f s The flow rate of the municipal pipe network is represented by l as the liquid level of the pump pit, s is the bottom area of the pump pit, and the unit of the flow rate is m 3 /h；
According to the above formula, the left and right sides are changed to have:
Figure FDA0003692297670000012
according to the formula (2), when f s (t-1) when the flow rate f of the pipe network is known, the flow rate f of the pipe network is determined at any moment s (t) can be calculated;
Figure FDA0003692297670000013
in this pair f s (1) Taking:
Figure FDA0003692297670000014
obtaining a series of estimated values f of the municipal pipe network sewage flow according to the formulas (2) and (4) s (t), obviously f s (t) is a time series;
Figure FDA0003692297670000015
and calculating the obtained auto-regression correlation coefficient of the municipal pipe network flow sequence, wherein the number of coefficients within 70% of the maximum auto-regression correlation coefficient value is taken as a regression order, and the k-order covariance is defined as follows:
Figure FDA0003692297670000021
where n is the time series length, k is 0,1,2 s (t) is the value at time t of the time series,
Figure FDA0003692297670000022
is f s Then the autocorrelation function is:
Figure FDA0003692297670000023
2. the method for predicting the water quantity of the municipal pipe network based on the GRU neural network as claimed in claim 1, wherein: the data cleaning method is a method for replacing missing abnormal values by adopting the average value of the adjacent first 10 data based on the original data of the central control system of the sewage plant.
3. The method for predicting the water quantity of the municipal pipe network based on the GRU neural network as claimed in claim 1, wherein the method comprises the following steps: the pump pit liquid level and the factory entering flow data represent the data of the period of time by taking the average value of the data within 1 minute of fixed time, and then the data are subjected to denoising treatment by respectively taking the moving average of the two data of the pump pit liquid level and the factory entering flow data in 5 time windows.
4. The method for predicting the water quantity of the municipal pipe network based on the GRU neural network as claimed in claim 1, wherein the method comprises the following steps: the unit of the GRU neural network is:
Figure FDA0003692297670000024
wherein
z t ＝σ(W z X t +U z h t-1 )，
r t ＝σ(W r X t +U r h t-1 )，
Figure FDA0003692297670000025
Where tanh is the hyperbolic tangent function, h t Is the hidden state at the current time, h t-1 For hidden states transmitted at the last moment, X t Is the state at the current time, z t To control the gating of updates, r t For reset gating, σ is a sigmoid function by which data is transformed into a value in the range of 0-1, acting as a gating signal z t (ii) a After the gating signal is obtained, reset gating is first used to obtain the data r after "reset t ⊙h t-1 Then r is further reduced t ⊙h t-1 And input X t Into W r ，W z Initializing the parameter to be trained in a random initialization mode, wherein the parameter represents the product of corresponding elements of the matrix;
the neural network training process includes randomly dividing original data into training set and testing set in the ratio of 7 to 3, training in batch random gradient method based on the training set, calculating the error of the model in the testing set every 15 times, stopping training when the error of the model in the testing set is lower than the last result, and using the parameter of the last training as the parameter of the model.
CN202011174683.4A 2020-10-29 2020-10-29 Municipal pipe network water flow prediction method based on GRU neural network Active CN112561118B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011174683.4A CN112561118B (en) 2020-10-29 2020-10-29 Municipal pipe network water flow prediction method based on GRU neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011174683.4A CN112561118B (en) 2020-10-29 2020-10-29 Municipal pipe network water flow prediction method based on GRU neural network

Publications (2)

Publication Number Publication Date
CN112561118A CN112561118A (en) 2021-03-26
CN112561118B true CN112561118B (en) 2022-09-02

Family

ID=75042299

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011174683.4A Active CN112561118B (en) 2020-10-29 2020-10-29 Municipal pipe network water flow prediction method based on GRU neural network

Country Status (1)

Country Link
CN (1) CN112561118B (en)

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR102310490B1 (en) * 2018-04-27 2021-10-08 한국과학기술원 The design of GRU-based cell structure robust to missing value and noise of time-series data in recurrent neural network
CN110991713B (en) * 2019-11-21 2022-04-01 杭州电子科技大学 Irregular area flow prediction method based on multi-graph convolution sum GRU

Also Published As

Publication number Publication date
CN112561118A (en) 2021-03-26

Similar Documents

Publication Publication Date Title
Zhao et al. Enhancing robustness of monthly streamflow forecasting model using gated recurrent unit based on improved grey wolf optimizer
CN105391083B (en) Wind power interval short term prediction method based on variation mode decomposition and Method Using Relevance Vector Machine
Li et al. DTDR–ALSTM: Extracting dynamic time-delays to reconstruct multivariate data for improving attention-based LSTM industrial time series prediction models
CN109492265B (en) Wastewater effluent index prediction method based on dynamic nonlinear PLS soft measurement method
CN110981021B (en) Intelligent dosing system and method for advanced wastewater oxidation treatment based on fuzzy BP neural network
CN112396234A (en) User side load probability prediction method based on time domain convolutional neural network
Cheng et al. Robust adaptive boosted canonical correlation analysis for quality-relevant process monitoring of wastewater treatment
Wang et al. A full-view management method based on artificial neural networks for energy and material-savings in wastewater treatment plants
Deng et al. Incipient fault detection for dynamic chemical processes based on enhanced CVDA integrated with probability information and fault-sensitive features
Liu et al. A novel deep learning ensemble model based on two-stage feature selection and intelligent optimization for water quality prediction
CN115099135A (en) Improved artificial neural network multi-type operation power consumption prediction method
CN112561118B (en) Municipal pipe network water flow prediction method based on GRU neural network
CN109033524A (en) A kind of chemical process concentration variable On-line Estimation method based on robust mixed model
Peng et al. Monitoring of wastewater treatment process based on multi-stage variational autoencoder
CN117094431A (en) DWTfar meteorological data time sequence prediction method and equipment for multi-scale entropy gating
Huang et al. Elman neural network considering dynamic time delay estimation for short-term forecasting of offshore wind power
Cao et al. Deep learning based soft sensor for microbial wastewater treatment efficiency prediction
CN116108961A (en) Drought prediction method and system based on ARIMA-SVM model
CN115905821A (en) Urban sewage treatment process state monitoring method based on multi-stage dynamic fuzzy width learning
Kang et al. Research on forecasting method for effluent ammonia nitrogen concentration based on GRA-TCN
Yoo et al. Application of fuzzy partial least squares (FPLS) modeling nonlinear biological processes
CN114462208A (en) Effluent BOD online soft measurement method based on self-organizing RBFNN
CN112862190A (en) Water quality time sequence prediction method based on improved Seq2Seq frame
CN113851184A (en) Artificial intelligence-based faecal escherichia coli population number prediction method and device
Li et al. Application of deep learning model based on transfer learning in activated sludge process for wastewater purification

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