CN113159395A - Deep learning-based sewage treatment plant water inflow prediction method and system - Google Patents
Deep learning-based sewage treatment plant water inflow prediction method and system Download PDFInfo
- Publication number
- CN113159395A CN113159395A CN202110347452.7A CN202110347452A CN113159395A CN 113159395 A CN113159395 A CN 113159395A CN 202110347452 A CN202110347452 A CN 202110347452A CN 113159395 A CN113159395 A CN 113159395A
- Authority
- CN
- China
- Prior art keywords
- model
- normalized
- output
- lstm
- sewage treatment
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 239000010865 sewage Substances 0.000 title claims abstract description 46
- 238000000034 method Methods 0.000 title claims abstract description 34
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 title claims abstract description 32
- 238000013135 deep learning Methods 0.000 title claims abstract description 24
- 238000012549 training Methods 0.000 claims abstract description 60
- 230000001419 dependent effect Effects 0.000 claims abstract description 32
- 238000010606 normalization Methods 0.000 claims abstract description 27
- 238000012545 processing Methods 0.000 claims abstract description 16
- 238000012360 testing method Methods 0.000 claims description 18
- 238000012795 verification Methods 0.000 claims description 16
- 238000011156 evaluation Methods 0.000 claims description 13
- 230000008569 process Effects 0.000 claims description 10
- 210000002569 neuron Anatomy 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 2
- 230000006870 function Effects 0.000 description 12
- 238000010586 diagram Methods 0.000 description 10
- 230000004913 activation Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000005259 measurement Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000002360 preparation method Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000015654 memory Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 230000036962 time dependent Effects 0.000 description 1
- 238000004065 wastewater treatment Methods 0.000 description 1
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
- G06N3/044—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
- G06N3/048—Activation functions
-
- 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
- G06N3/08—Learning methods
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Strategic Management (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Human Resources & Organizations (AREA)
- Tourism & Hospitality (AREA)
- Economics (AREA)
- Marketing (AREA)
- Development Economics (AREA)
- General Business, Economics & Management (AREA)
- Educational Administration (AREA)
- Primary Health Care (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Testing And Monitoring For Control Systems (AREA)
- Activated Sludge Processes (AREA)
Abstract
The invention discloses a sewage treatment plant inflow prediction method based on deep learning, which comprises the following steps: acquiring historical data through an acquisition device, wherein the historical data comprises the water inflow rate of a corresponding sewage treatment plant over the years; carrying out normalization processing on the historical data to obtain a normalized model, and then training the normalized model to obtain a trained normalized model; inputting the intake independent variable as a trained normalized model, taking the intake dependent variable as actual output, and performing supervised learning on the trained normalized model; performing inverse normalization processing on the output dependent variable to obtain predicted output, and optimizing a feedback system by comparing the difference value of the actual output and the predicted output so as to predict the inflow water flow; the invention aims to solve the problem of accuracy of water inflow prediction of a sewage treatment plant.
Description
Technical Field
The invention relates to the research field of on-line prediction of water inflow, in particular to a deep learning-based method and a deep learning-based system for predicting the water inflow of a sewage treatment plant.
Background
In the wastewater treatment process, the inflow rate has direct requirements on the treatment capacity of a sewage treatment plant, and the control of effluent indexes is also closely influenced.
At present, the water inflow in the sewage treatment process is based on the real-time monitoring of hardware equipment, and has no prediction function, so that higher requirements are provided for the real-time treatment capacity of a sewage treatment plant and the preparation of various consumed articles for sewage treatment. The invention provides a method for predicting inflow based on a long-short term memory network regression (LSTM) soft measurement model.
The water inlet flow of a sewage treatment plant is time-dependent sequence data, and in the conventional prediction method for the sequence data, the conventional prediction method has no online learning function, and when a model is established, the model parameters are fixed, and if new processing data needs to be learned, the model needs to be established again. In addition, the conventional prediction method is mainly suitable for linear system prediction. Machine learning has good applicability to nonlinear system prediction of complex networks, but requires feature engineering that relies on empirical processing.
Disclosure of Invention
The invention mainly aims to overcome the defects of the prior art and provide a method and a system for predicting the water inflow of a sewage treatment plant based on deep learning, aiming at optimizing the real-time treatment capacity of the sewage treatment plant.
The deep learning has good applicability to time series nonlinear system prediction and has automatic characteristic engineering capability, long and short term memory network regression therein has long sequence data information extraction capability and abstract capability to short sequence characteristics, and meanwhile, as the input sequence becomes long, online learning capability can be realized by updating the state in real time.
The invention aims to provide a sewage treatment plant inflow prediction method based on deep learning.
The invention also provides a sewage treatment plant inflow prediction system based on deep learning.
The first purpose of the invention is realized by the following technical scheme:
a sewage treatment plant inflow prediction method based on deep learning is characterized by comprising the following steps:
acquiring historical data through an acquisition device, wherein the historical data comprises the water inflow rate of a corresponding sewage treatment plant over the years;
carrying out normalization processing on the historical data to obtain a normalized model, and then training the normalized model to obtain a trained normalized model;
inputting the intake independent variable as a trained normalized model, taking the intake dependent variable as actual output, and performing supervised learning on the trained normalized model;
and performing inverse normalization processing on the output dependent variable to obtain predicted output, and comparing the difference value between the actual output and the predicted output to optimize the feedback system so as to predict the inflow rate.
Furthermore, the water inflow independent variable is a water inflow value of the first K time units, and the water inflow dependent variable is a water inflow value of the current time unit.
Further, the normalizing the historical data to obtain a normalized model, and then training the normalized model to obtain a trained normalized model specifically include: the historical data is divided into three parts: a training set, a verification set and a test set; normalizing the training set to obtain a normalized training set, wherein the normalization processing specifically comprises the following steps:
let a certain variable sequence xiIs x1,…,xnWherein the maximum value is recorded as xmaxMinimum value of xminThen the normalized sequence is:
wherein, after normalizationThe range of the sequence value is 0-1, and x is storedmaxAnd xminA value of (d);
constructing and reserving a normalization model, wherein the normalization model comprises an input module, an LSTM module, a full-connection module, a regression module and a training end judgment module;
normalizing the verification set by using a normalization model to obtain a normalized verification set; inputting a normalized training set and a normalized verification set to train the normalized model, obtaining the trained normalized model and storing the normalized model, namely storing the LSTM model; and normalizing the test set by using the trained normalization model to obtain a normalized test set, and inputting the normalized test set into the trained normalization model to evaluate the accuracy index.
Further, the historical data is divided into three parts: training set, verification set and test set, wherein the division is as follows 8:1:1 proportion is divided by using a uniform random sampling mode, namely 80% of training set, 10% of verification set and 10% of testing set.
Further, the training of the normalized model to obtain the trained normalized model specifically includes:
1) initializing the model parameters;
2) the input data is transmitted forwards through the LSTM layer and the full connection layer to obtain an output value;
3) calculating an error between an output value of the training model and a target value;
4) when the error is larger than the expected value, the error is transmitted back to the model, and the errors of the full-connection layer and the LSTM layer are sequentially obtained, wherein the error of each layer is the total error of the model; when the error is equal to or less than the expected value, the training is finished;
5) and updating the weight of the model according to the obtained error.
Further, the intake independent variable is used as an input of the normalized model after training, the intake dependent variable is used as an actual output, and the normalized model after training is supervised and learned, specifically: recording actual output, predicting the next output, performing online learning on the normalized model through the actual output, obtaining new model parameters through the online learning, and updating the model.
Further, the inverse normalization processing is performed on the output dependent variable to obtain a prediction output, and the feedback system is optimized by comparing the difference between the actual output and the prediction output, so as to predict the inflow water flow, specifically: performing inverse normalization processing on the output dependent variable to obtain a dependent variable predicted value, and obtaining a standard error and determining a correlation coefficient through the dependent variable predicted value and the dependent variable actual value;
the standard error is calculated as follows:
where RMSE is the standard error, yiAs dependent variable predictors, xiIs the actual value of the dependent variable;
the determination of the correlation coefficient is calculated as follows:
wherein R is2In order to determine the correlation coefficient(s),
is a sequence xiIs measured.
The second purpose of the invention is realized by the following technical scheme:
a sewage treatment plant inflow prediction system based on deep learning comprises an LSTM model and a feedback system;
the LSTM model comprises an input module, an LSTM module, a full connection layer module, a regression module, a training end judgment module and a model evaluation module; the input module is used for data input of the model, the LSTM module contains M LSTM layers, the full-connection layer module includes a plurality of full-connection layers, the LSTM module contains M LSTM layers and carries out various feature combinations to the variable of input, the full-connection layer module includes that the feature that a plurality of full-connection layer LSTM layer was drawed carries out the secret reflection, the regression layer uses loss function MSE to carry out loss calculation to the model, the number of iterations that the training was ended and is judged the module and judge through setting up judges and stop the model training, the model evaluation module is appraised the performance of model through root mean square error and correlation.
The feedback system described includes the predicted output of the LSTM, the actual measured value, and the decision to retrain the model; judging whether to retrain the LSTM model by comparing the difference between the predicted output of the LSTM and the real measured value and judging whether to exceed the designed threshold;
the working process is as follows: and inputting the normalized historical data into an LSTM model for model training to obtain the LSTM model, evaluating the LSTM model by a model evaluation module, updating the state of the LSTM model in real time according to the evaluation, predicting the output of the LSTM model by a prediction system, and feeding the output result back to a control system of a sewage treatment plant.
Further, the first full connection layer of the full connection layer module comprises 100-300 connection neurons; since the regression dependent variable is a predicted value of the current inflow rate, the number of outputs of the fully-connected layers is 1.
Further, the training end judgment module performs control through the maximum number of rounds set by initialization.
Further, the input module is used for inputting models, namely independent variable input; the activation function adopted by the activation layer is a Leaky ReLU function.
Further, the LSTM module employs an LSTM layer containing 100 hidden units.
Further, the loss function employed by the regression module is Mean Square Error (MSE).
Compared with the prior art, the invention has the following advantages and beneficial effects:
1. the invention provides an application of a prediction method of water inflow of a sewage treatment plant, wherein the LSTM synthesizes the characteristics of a water inflow sequence of the sewage treatment plant, so that regression prediction has a better effect.
2. The invention provides the prediction of the water inflow rate of the sewage treatment plant, and has an optimization function on the real-time treatment capacity of the sewage treatment plant and the preparation of various consumed articles for sewage treatment.
Drawings
FIG. 1 is a flow chart of a deep learning-based method for predicting the inflow of a sewage treatment plant according to the present invention;
FIG. 2 is a flow chart of a method for predicting water inflow according to an embodiment of the present invention;
FIG. 3 is a general block diagram of a method for forecasting influent water flow in accordance with an embodiment of the present invention;
FIG. 4 is an overall structure diagram of a training process according to the embodiment of the present invention;
FIG. 5 is a schematic diagram of a training architecture in accordance with an embodiment of the present invention;
FIG. 6 is a detailed structural diagram of a training process according to the embodiment of the present invention;
FIG. 7 is a schematic diagram illustrating the result evaluation of the predicted inflow rate of the training model according to the embodiment of the present invention;
fig. 8 is a structural diagram of a water inlet flow prediction system of a sewage treatment plant according to an embodiment of the invention.
Detailed Description
The present invention will be described in further detail with reference to examples and drawings, but the present invention is not limited thereto.
Example (b):
a method for predicting the inflow of a sewage treatment plant based on deep learning is disclosed, as shown in FIG. 1, and comprises the following steps:
acquiring historical data through an acquisition device, wherein the historical data comprises the water inflow rate of a corresponding sewage treatment plant over the years;
carrying out normalization processing on the historical data to obtain a normalized model, and then training the normalized model to obtain a trained normalized model;
inputting the intake independent variable as a trained normalized model, taking the intake dependent variable as actual output, and performing supervised learning on the trained normalized model;
and performing inverse normalization processing on the output dependent variable to obtain predicted output, and comparing the difference value between the actual output and the predicted output to optimize the feedback system so as to predict the inflow rate.
The specific process is shown in fig. 2, and fig. 3 is a general framework diagram of the inflow water prediction method, and the specific contents are as follows:
step 101, dividing historical data, wherein the historical data comprises inflow water flow corresponding to the past year of a sewage treatment plant.
The measurement data is taken from a sewage treatment plant in Guangzhou, Guangdong province, and shows 4300 hours of sample points for 1 year at 2 hour intervals.
4300 sample points are divided into a training set, a verification set and a test set according to the proportion of 8:1:1, wherein the sample points of the training set, the verification set and the test set are 3440, 430 and 430.
And 102, normalizing, namely normalizing the training set in the independent variable front 11 step water inflow values and the dependent variable current water inflow values respectively.
The normalization method is that a certain variable sequence is assumed to be x1,…,xnNote that the maximum and minimum values are respectively:
xmaxand xminThen the normalized sequence is:
wherein the normalized sequence value ranges from 0 to 1, storing xmaxAnd xminThe value of (c).
Step 103, inputting a module, wherein the input is input of a model, the format of input data is 1 × 11 × 1, namely the data width, height and channel number are 1, 11 and 1 respectively, the input data comprises a normalized training set and a verification set, the training set is used for model fitting, and the verification set is used for adjusting hyper-parameters of the model and primarily evaluating the capability of the model. Fig. 4 is an overall structure diagram of the training process. FIG. 5 is a schematic diagram of a training architecture; FIG. 6 is a detailed structural diagram of a training process;
step 104, LSTM module, the embodiment of the present invention adopts an LSTM layer including 100 hidden units.
And 105, a full-connection module, wherein the embodiment of the invention adopts two full-connection layers, the first full-connection layer comprises 200 connection neurons, and the output number of the full-connection layer is 1 because the regression dependent variable is the predicted value of the current water inflow.
Step 106, a regression module, wherein the loss function adopted in the embodiment of the present invention is Mean Square Error (MSE).
Step 107, the training end judgment module controls the training by the initialized maximum round number (MaxEpochs), which is 200 in the embodiment of the present invention. Training is completed and the model is saved, otherwise, the step 403 is executed.
Step 108, normalizing the test set, wherein the maximum value and the minimum value of normalization are x stored in step 402maxAnd xmin. The purpose of using the training set to normalize the model parameters is to map the test set through the training set model, ensure that data beyond the range of the training set can be well recovered, and improve the generalization capability of the model.
Step 109, storing the LSTM model, where the model is the trained weight coefficient of each layer, and obtaining the normalized dependent variable, i.e. the current intake flow value, by inputting the normalized independent variable by the step intake flow value of 11.
And then carrying out inverse normalization on the output dependent variable to obtain a predicted numerical value of the inflow water flow.
Step 110, model evaluation module, in the embodiments of the present invention, standard error (also called root mean square error, RMSE for short) and correlation coefficient (R) determination are used2 score)。
The standard error is calculated by the formula:
wherein the calculation formula for determining the correlation coefficient is as follows:
here yi,xiRespectively a predicted value and an actual value,
is a sequence xiIs measured.
By setting the training parameters, the comparison of the test results of the LSTM model to the actual values of the test set is shown in fig. 7, where the predicted inlet water flow values are all below the factory set threshold values.
Further, the prediction system for the inflow of the sewage treatment plant according to the embodiment of the present invention can be expressed as a flowchart of fig. 8, which includes the following steps:
and step 701, inputting the system, including the inflow water flow value of the first 11 steps.
And step 702, feeding back the actual measured value of the inflow of the current system and the prediction comparison.
And 703, predicting the next output quantity of the system by the prediction system of the system, wherein the prediction system predicts the next output quantity of the system according to the input of the step 701, and simultaneously online learns the prediction system according to the current output quantity of the step 702 and updates a prediction system model.
A sewage treatment plant inflow prediction system based on deep learning is shown in FIG. 8 and comprises an LSTM model and a feedback system;
the LSTM model comprises an input module, an LSTM module, a full connection layer module, a regression module, a training end judgment module and a model evaluation module; the input module is used for data input of the model, the LSTM module contains M LSTM layers, the full-connection layer module includes a plurality of full-connection layers, the LSTM module contains M LSTM layers and carries out various feature combinations to the variable of input, the full-connection layer module includes that the feature that a plurality of full-connection layer LSTM layer was drawed carries out the secret reflection, the regression layer uses loss function MSE to carry out loss calculation to the model, the number of iterations that the training was ended and is judged the module and judge through setting up judges and stop the model training, the model evaluation module is appraised the performance of model through root mean square error and correlation.
The working process is as follows: and inputting the normalized historical data into an LSTM model for model training to obtain the LSTM model, evaluating the LSTM model by a model evaluation module, updating the state of the LSTM model in real time according to the evaluation, predicting the output of the LSTM model by a prediction system, and feeding the output result back to a control system of a sewage treatment plant.
Further, the first full connection layer of the full connection layer module comprises 100-300 connection neurons; since the regression dependent variable is a predicted value of the current inflow rate, the number of outputs of the fully-connected layers is 1.
Further, the training end judgment module performs control through the maximum number of rounds set by initialization.
Further, the input module is used for inputting models, namely independent variable input; the activation function adopted by the activation layer is a Leaky ReLU function.
Further, the LSTM module employs an LSTM layer containing 100 hidden units.
Further, the loss function employed by the regression module is Mean Square Error (MSE).
The above embodiments are preferred embodiments of the present invention, but the present invention is not limited to the above embodiments, and any other changes, modifications, substitutions, combinations, and simplifications which do not depart from the spirit and principle of the present invention should be construed as equivalents thereof, and all such changes, modifications, substitutions, combinations, and simplifications are intended to be included in the scope of the present invention.
Claims (10)
1. A sewage treatment plant inflow prediction method based on deep learning is characterized by comprising the following steps:
acquiring historical data through an acquisition device, wherein the historical data comprises the water inflow rate of a corresponding sewage treatment plant over the years;
carrying out normalization processing on the historical data to obtain a normalized model, and then training the normalized model to obtain a trained normalized model;
inputting the intake independent variable as a trained normalized model, taking the intake dependent variable as actual output, and performing supervised learning on the trained normalized model;
and performing inverse normalization processing on the output dependent variable to obtain predicted output, and comparing the difference value between the actual output and the predicted output to optimize the feedback system so as to predict the inflow rate.
2. The deep learning based sewage treatment plant inflow prediction method according to claim 1, wherein the inflow independent variable is an inflow value of the first K time units, and the inflow dependent variable is an inflow value of the current time unit.
3. The deep learning-based sewage treatment plant inflow prediction method according to claim 1, wherein the historical data is normalized to obtain a normalized model, and the normalized model is trained to obtain a trained normalized model, specifically: the historical data is divided into three parts: a training set, a verification set and a test set; normalizing the training set to obtain a normalized training set, wherein the normalization processing specifically comprises the following steps:
let a certain variable sequence xiIs x1,…,xnWherein the maximum value is recorded as xmaxMinimum value of xminThen the normalized sequence is:
wherein the range of the normalized sequence value is 0-1, and x is storedmaxAnd xminA value of (d);
constructing and reserving a normalization model, wherein the normalization model comprises an input module, an LSTM module, a full-connection module, a regression module and a training end judgment module;
normalizing the verification set by using a normalization model to obtain a normalized verification set; inputting a normalized training set and a normalized verification set to train the normalized model, obtaining the trained normalized model and storing the normalized model, namely storing the LSTM model; and normalizing the test set by using the trained normalization model to obtain a normalized test set, and inputting the normalized test set into the trained normalization model to evaluate the accuracy index.
4. The deep learning based sewage treatment plant inflow prediction method according to claim 3, wherein the historical data is divided into three parts: training set, verification set and test set, wherein the division is as follows 8:1:1 proportion is divided by using a uniform random sampling mode, namely 80% of training set, 10% of verification set and 10% of testing set.
5. The deep learning-based sewage treatment plant inflow prediction method according to claim 3, wherein the normalized model is trained to obtain a trained normalized model, and specifically comprises:
1) initializing the model parameters;
2) the input data is transmitted forwards through the LSTM layer and the full connection layer to obtain an output value;
3) calculating an error between an output value of the training model and a target value;
4) when the error is larger than the expected value, the error is transmitted back to the model, and the errors of the full-connection layer and the LSTM layer are sequentially obtained, wherein the error of each layer is the total error of the model; when the error is equal to or less than the expected value, the training is finished;
5) and updating the weight of the model according to the obtained error.
6. The deep learning-based sewage treatment plant inflow prediction method according to claim 1, wherein the intake independent variable is input as a normalized model after training, the intake dependent variable is output as an actual output, and the normalized model after training is supervised and learned, specifically: recording actual output, predicting the next output, performing online learning on the normalized model through the actual output, obtaining new model parameters through the online learning, and updating the model.
7. The deep learning-based sewage treatment plant inflow prediction method according to claim 1, wherein the output dependent variable is subjected to inverse normalization processing to obtain a prediction output, and a feedback system is optimized by comparing a difference between an actual output and the prediction output to predict the inflow, specifically: performing inverse normalization processing on the output dependent variable to obtain a dependent variable predicted value, and obtaining a standard error and determining a correlation coefficient through the dependent variable predicted value and the dependent variable actual value;
the standard error is calculated as follows:
where RMSE is the standard error, yiAs dependent variable predictors, xiIs the actual value of the dependent variable;
the determination of the correlation coefficient is calculated as follows:
wherein R is2In order to determine the correlation coefficient(s),
is a sequence xiIs measured.
8. A sewage treatment plant inflow prediction system based on deep learning is characterized by comprising an LSTM model and a feedback system;
the LSTM model comprises an input module, an LSTM module, a full connection layer module, a regression module, a training end judgment module and a model evaluation module; the input module is used for inputting data of the model, the LSTM module comprises M LSTM layers, and the full-connection layer module comprises a plurality of full-connection layers;
the feedback system described includes the predicted output of the LSTM, the actual measured value, and the decision to retrain the model; judging whether to retrain the LSTM model by comparing the difference between the predicted output of the LSTM and the real measured value and judging whether to exceed the designed threshold;
the working process is as follows: and inputting the normalized historical data into an LSTM model for model training to obtain the LSTM model, evaluating the LSTM model by a model evaluation module, updating the state of the LSTM model in real time according to the evaluation, predicting the output of the LSTM model by a prediction system, and feeding the output result back to a control system of a sewage treatment plant.
9. The deep learning based sewage treatment plant inflow prediction system of claim 8, wherein the first fully-connected layer of the fully-connected layer module comprises 100-300 connecting neurons.
10. The deep learning based sewage treatment plant inflow prediction system of claim 8, wherein the training end determination module is controlled by initializing a set maximum number of rounds.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110347452.7A CN113159395A (en) | 2021-03-31 | 2021-03-31 | Deep learning-based sewage treatment plant water inflow prediction method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110347452.7A CN113159395A (en) | 2021-03-31 | 2021-03-31 | Deep learning-based sewage treatment plant water inflow prediction method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113159395A true CN113159395A (en) | 2021-07-23 |
Family
ID=76886120
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110347452.7A Pending CN113159395A (en) | 2021-03-31 | 2021-03-31 | Deep learning-based sewage treatment plant water inflow prediction method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113159395A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113837364A (en) * | 2021-09-17 | 2021-12-24 | 华南师范大学 | Sewage treatment soft measurement method and system based on residual error network and attention mechanism |
| CN114169730A (en) * | 2021-11-30 | 2022-03-11 | 温州科技职业学院 | Machine learning-based urban garbage classification work evaluation method and system |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105389980A (en) * | 2015-11-09 | 2016-03-09 | 上海交通大学 | Short-time traffic flow prediction method based on long-time and short-time memory recurrent neural network |
-
2021
- 2021-03-31 CN CN202110347452.7A patent/CN113159395A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105389980A (en) * | 2015-11-09 | 2016-03-09 | 上海交通大学 | Short-time traffic flow prediction method based on long-time and short-time memory recurrent neural network |
Non-Patent Citations (1)
| Title |
|---|
| 郑博一: "人工鱼群神经网络应用于城市污水处理厂进水量预测研究", 《环境科学与管理》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113837364A (en) * | 2021-09-17 | 2021-12-24 | 华南师范大学 | Sewage treatment soft measurement method and system based on residual error network and attention mechanism |
| CN113837364B (en) * | 2021-09-17 | 2023-07-11 | 华南师范大学 | Sewage treatment soft measurement method and system based on residual network and attention mechanism |
| CN114169730A (en) * | 2021-11-30 | 2022-03-11 | 温州科技职业学院 | Machine learning-based urban garbage classification work evaluation method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110705743B (en) | New energy consumption electric quantity prediction method based on long-term and short-term memory neural network | |
| CN110889085A (en) | Intelligent wastewater monitoring method and system based on complex network multiple online regression | |
| CN110766212B (en) | Ultra-short-term photovoltaic power prediction method for historical data missing electric field | |
| CN112990556A (en) | User power consumption prediction method based on Prophet-LSTM model | |
| CN112733417B (en) | Abnormal load data detection and correction method and system based on model optimization | |
| CN109919356B (en) | BP neural network-based interval water demand prediction method | |
| CN113554466B (en) | Short-term electricity consumption prediction model construction method, prediction method and device | |
| CN110824914B (en) | Intelligent wastewater treatment monitoring method based on PCA-LSTM network | |
| CN115374995A (en) | Distributed photovoltaic and small wind power station power prediction method | |
| CN113449919B (en) | Power consumption prediction method and system based on feature and trend perception | |
| CN113159395A (en) | Deep learning-based sewage treatment plant water inflow prediction method and system | |
| CN114169416B (en) | Short-term load prediction method based on migration learning under small sample set | |
| CN114357670A (en) | Power distribution network power consumption data abnormity early warning method based on BLS and self-encoder | |
| CN112766548A (en) | Order completion time prediction method based on GASA-BP neural network | |
| CN115329930A (en) | Flood process probability forecasting method based on mixed deep learning model | |
| CN116303786A (en) | Block chain financial big data management system based on multidimensional data fusion algorithm | |
| CN115982141A (en) | Characteristic optimization method for time series data prediction | |
| CN113762591B (en) | Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning | |
| CN118017482A (en) | Flexible climbing capacity demand analysis method based on prediction error feature extraction | |
| CN109359741A (en) | A kind of wastewater treatment influent quality timing variations intelligent Forecasting | |
| CN115563848A (en) | Distributed photovoltaic total radiation prediction method and system based on deep learning | |
| CN117578441A (en) | Method for improving power grid load prediction precision based on neural network | |
| CN113033898A (en) | Electrical load prediction method and system based on K-means clustering and BI-LSTM neural network | |
| CN111863153A (en) | Method for predicting total amount of suspended solids in wastewater based on data mining | |
| CN116565850A (en) | Wind power ultra-short-term prediction method based on QR-BLSTM |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210723 |
|
| RJ01 | Rejection of invention patent application after publication |