CN113642230B - Multi-target complex drainage system adjustable weir intelligent control method based on machine learning - Google Patents
Multi-target complex drainage system adjustable weir intelligent control method based on machine learning Download PDFInfo
- Publication number
- CN113642230B CN113642230B CN202110780334.5A CN202110780334A CN113642230B CN 113642230 B CN113642230 B CN 113642230B CN 202110780334 A CN202110780334 A CN 202110780334A CN 113642230 B CN113642230 B CN 113642230B
- Authority
- CN
- China
- Prior art keywords
- discrimination
- drainage system
- data
- probability
- adjustable weir
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 20
- 238000010801 machine learning Methods 0.000 title claims abstract description 13
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims abstract description 21
- 238000012549 training Methods 0.000 claims abstract description 10
- 238000004364 calculation method Methods 0.000 claims abstract description 9
- 230000002349 favourable effect Effects 0.000 claims abstract description 3
- 239000011159 matrix material Substances 0.000 claims description 29
- 239000013598 vector Substances 0.000 claims description 11
- 230000002265 prevention Effects 0.000 claims description 9
- 238000009825 accumulation Methods 0.000 claims description 8
- 238000004422 calculation algorithm Methods 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 3
- 238000003672 processing method Methods 0.000 claims description 2
- 238000012850 discrimination method Methods 0.000 claims 1
- 230000006870 function Effects 0.000 abstract description 9
- 230000000694 effects Effects 0.000 abstract description 6
- 230000009286 beneficial effect Effects 0.000 abstract description 3
- 238000004458 analytical method Methods 0.000 abstract description 2
- 238000013398 bayesian method Methods 0.000 abstract 1
- 238000012544 monitoring process Methods 0.000 description 5
- 238000010606 normalization Methods 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 230000002411 adverse Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000008094 contradictory effect Effects 0.000 description 1
- 239000007788 liquid Substances 0.000 description 1
- 238000012821 model calculation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 239000010865 sewage Substances 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/08—Probabilistic or stochastic CAD
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Mathematical Analysis (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Computational Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Databases & Information Systems (AREA)
- Computing Systems (AREA)
- Algebra (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Operations Research (AREA)
- Probability & Statistics with Applications (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides an intelligent control method for an adjustable weir of a multi-target complex drainage system based on machine learning. The invention relates to a control object discrimination state element, a Markov distance between a population and a sample, a priori probability and a posterior probability function structure, and an output feedback solver takes a misjudgment rate as an output feedback solver. The invention selects key state element sensitive data such as the initial water level of a pipeline, the water level flow of a key point, the water level of a front pool of a pump station, the water level of the least favorable point and the like based on machine learning discriminant analysis and calculation, and carries out auxiliary decision on key steps of drainage system control by using four machine learning discriminant methods such as Bayesian method, linear discriminant, linear diagonal and secondary discriminant. The invention has the advantages that the required external data is simple, the real-time acquisition is easy, the training samples can be continuously expanded and accumulated, and the closer the training samples are to the whole, the better the judging effect is. The invention provides an effective method for improving the intelligent control level of the adjustable weir of the complex drainage system, and is beneficial to improving the overall operation level of the complex drainage system.
Description
Technical Field
The invention belongs to the field of municipal engineering, relates to a control method of a drainage system, and in particular relates to an intelligent control method of an adjustable weir of a complex drainage system based on multiple targets.
Background
In recent years, urban waterlogging prevention and black and odorous river management are becoming a focus of social attention, corresponding drainage system upgrading and river runoff pollution control engineering construction is continuously accelerated, and related engineering designs of old drainage system upgrading and pollution control are realized simultaneously after original drainage systems are changed. The operation control of the drainage system after transformation is more complex, and generally relates to a plurality of targets, namely flood control targets, mainly aiming at improving the safety of the drainage system and realizing the upgrading and waterlogging control; secondly, the pollution control aim is to reduce the initial rainwater pollution and the CSO emission. Because the whole framework of the complex system is that a new drainage system is formed by the newly built intercepting system and the original drainage system, the newly built intercepting system and the original drainage system are required to be combined to achieve the engineering goal.
In the drainage system engineering, a unified and contradictory complex relationship exists between two large projects, namely flood prevention and pollution control. The unification is that the two engineering targets need to be realized in the same engineering in a unified way, and the two engineering targets must be mutually combined; the "contradiction" is that the preconditions and operation modes of the two engineering targets do not match perfectly, and in some cases, there is a certain conflict between the two engineering targets. For example, under certain conditions, in order to achieve the goal of controlling pollution engineering that rainfall of a certain millimeter number does not leave the river (for example, controlling initial rainwater of 10 mm), the municipal pump is started after rainfall of the first 10mm enters the intercepting system; in order to achieve the aim of preventing and controlling waterlogging, the regulation and storage system is required to be mainly used as the regulation and storage volume of peak clipping, namely, the regulation and storage tank is not fed in at the initial moment, peak clipping is started at the moment of waiting for a rain peak, and essentially, two functions are realized through regulation and storage space, and the limitation of the regulation and storage space is the essential cause of contradiction.
The intercept point is generally selected from a more flexible control form of the adjustable weir. The regulation of the weir is of great importance for the realization of complex drainage functions. Through model calculation, if the weir descending time is too early, the regulation system may be fully closed too early, and cannot continue to participate in peak clipping, so that system water accumulation is caused. In addition, the water in the shallow pipe network is discharged into the regulation system before the runoff peak value, so that municipal pumps can not be completely started, the drainage capacity of the shallow pipe network can not be fully utilized, and if the weir descending time is too late, the inflow weir can not timely participate in the peak clipping effect. Because municipal pump drainage capacity is insufficient to cope with rain peaks, if the shallow pipe network is full at the moment and the inflow weir is not lowered at the moment, the overflow weir flow is insufficient, and the system ponding can be caused. Therefore, it is critical to perform the weir reducing operation (increase the overflow capacity) at a proper time and increase the peak clipping capacity. The difficulty in adjusting the inflow weir, namely descending the weir is the judgment of the time of descending the weir, so the invention aims at utilizing the instrument monitoring data of the drainage system to judge the time point of the operation of descending the weir in real time.
Machine learning is a type of artificial intelligence that can predict future situations based on the vast amount of data collected at present. Briefly, machine learning is the analysis of data using algorithms from which inferences or predictions are learned and made. In view of the fact that a large number of effective databases which are updated continuously are generated in the actual operation of the drainage system, the data are analyzed through machine learning, and therefore the judgment feasibility of the weir descending time is high. When the prediction model is built, the machine learning prediction result is compared with the actual result of the training data through the building of the learning process, and the higher judgment accuracy is realized through continuous adjustment.
Disclosure of Invention
The invention solves the problems in the prior art, and provides a multi-target intelligent control method for an adjustable weir of a complex drainage system based on machine learning, so as to achieve the purpose of improving the intelligent control level and the integral operation level of the adjustable weir of the complex drainage system.
The technical solution provided by the invention for realizing the purpose of the invention comprises the following steps:
an intelligent control method for an adjustable weir of a multi-target complex drainage system based on machine learning comprises the following steps:
1) Obtaining discrimination state element data of a control object:
Selecting state element data: collecting state element data of a drainage system, wherein the state element data comprises water level values and flow values of pipelines at the positions of adjustable weirs, water level of a front pool of a pump station in the drainage system and water accumulation conditions of least favorable points of areas;
Carrying out standardized processing on the state element data to eliminate the limitation of dimension and magnitude order for subsequent discrimination and solution;
the normalization method is as follows:
assume that the observation matrix for the p-dimensional vector is as follows:
Wherein the standardized formula of x ij is
Where S ij is the variance of the variable x ij,Is the average value of the variable observation values;
2) Discrimination state element control concept of control object
The control targets are as follows: on the premise of ensuring flood prevention safety, if no water is accumulated in the first year of X years, the rainwater enters the intercepting system at the initial stage of pollution control and interception; the control elements are as follows: an adjustable weir, municipal pump; the real-time data is solved through discrimination calculation, if the category belongs to true, the adjustable weir is lowered to the bottom, and flood prevention safety is ensured; triggering the municipal pump to start if water is accumulated at the unfavorable point;
(3) Distinguishing algorithm solution
A. solving by distance discrimination
Distance discrimination (Mahalanobis): the samples have the closest mahalanobis distance (mean vector, covariance matrix) to the belonging population and have a certain far mahalanobis distance to other populations, and the calculation principle is as follows:
Let G be the p-dimensional ensemble, its distributed mean vector and covariance matrix are:
Let x= (x 1,x2,...,xp)',y＝(y1,y2,...,yp)' be two samples taken from the population G, let Σ >0 (Σ be positive definite matrix), define the square mahalanobis distance between x, y as: d 2(x,y)＝(x-y)′∑-1 (x-y) defines the square mahalanobis distance of x to the total G as: d 2(x,G)＝(x-μ)′∑-1 (x- μ)
Let two p-dimensional populations G 1 and G 2, distributed mean vectors μ 1 and μ 2, respectively, and covariance matrices Σ 1＞0,Σ2 > 0, respectively. Samples of capacity n 1,n2 were taken from both populations, denoted x 11,x12,…,x1n1 and x 21,x22,…,x2n2, respectively. The existing data set for judging the position is marked as x, and the attribution of the x is judged in a trial mode, and the following judging rule is adopted
When Σ 1＝Σ2 =Σ is known, subtracting the distances d 2(x,G2) and d 2(x,G1) can be obtained
d2(x,G2)-d2(x,G1)＝(x-μ2)′∑-1(x-μ2)-(x-μ1)′∑-1(x-μ1)
The discriminant rule is represented by W (x):
wherein W (x) is a linear discriminant function of two groups of distance discriminants, and a is a discriminant coefficient.
When Σ 1＝Σ2 =Σunknown
Order the
I.e. derived from the sampleThe estimate of sigma can then be derived as an estimate of a and W (x)
W (x) in the discriminant rule denoted by W (x) is replaced withThe discrimination rule at this time can be obtained.
When Sigma 1≠Σ2 is known
Let J (x) =d 2(x,G1)-d2(x,G2
J (x) is a quadratic discriminant function with discriminant rules:
When Σ 1≠Σ2 is unknown
From a sample pairSigma 1,Σ2 estimation
Obtaining square Markov distance estimation and secondary discrimination function
The discrimination rules are as follows:
if multiple overall distance determination
Let k p-dimensional populations G 1,G2,…,Gk, distributed mean vectors μ 1,μ2,…,μk, covariance matrices Σ 1>0,Σ2>0,…,Σk >0, respectively. Samples of capacity n 1,n2,…,nk were taken from the k populations, respectively, and noted as
The existing unknown data is marked as x, the attribution of the x is judged, and the judgment rule is that
Similar to the distance discrimination of two populations, the situation discrimination is also divided
When Σ 1＝Σ2＝…＝Σk =Σ is known
d2(x,Gi)＝(x-μi)′∑-1(x-μi)
Order the
Then d 2(x,Gi)＝x′∑-1x-2(I′i+ci) i=1, 2,..k
Since there is a common quadratic term in each distance, only the linear part thereof needs to be considered.
Let W I(x)＝I′ix+ci, i=1, 2,..k
The discriminant rule is rewritten as
X is G 1, if
Let W i (x) be I linear discriminant functions, I i be the discriminant coefficients, where c i is a constant term
When Σ 1＝Σ2＝…＝Σk =Σunknown
Order the
Where i=1, 2,..k
From the samples, an estimate of μ i, Σ is derived, thus yielding estimates of I i、ci and W i (x)
Where i=1, 2,..k
When Σ 1、Σ2,…,Σk is not fully equal and is unknown
Order the
Where i=1, 2,..k
The rule is determined as
X is G 1, if
B. Linear discriminant solution
Linear discrimination, assuming that the prior distribution of each data is the p-element normal distribution with the same covariance matrix, obtaining the joint estimation of the covariance matrix by the sampleThe judgment of the data group is obtained.
C. Linear diagonal solution
At this time, the diagonal matrix is taken as an estimate of the covariance matrix.
D. secondary discrimination solution
The prior distribution of the data set is assumed to be p-element normal distribution, but covariance matrixes are not identical, and estimation of each covariance matrix is obtained at the moment.
E. bayesian discriminant solution
Bayesian discriminant method (bayes): the Bayesian discrimination describes the existing knowledge by using a priori probability, then the priori probability is corrected by the sample to obtain posterior probability, and finally the discrimination is carried out based on the posterior probability. The calculation principle is as follows:
There are k p-dimensional populations G1, G2, …, gk, with probability density functions f 1(x),f2(x),…,fk (x), respectively. Assuming that the prior probability of sample x from the overall Gi is pi (i=1, 2,3 …, k), there is p 1+p2+…+pk =1. According to bayesian theory, the posterior probability of sample x from the overall Gi is:
If the erroneous judgment cost is considered, the aggregate of all samples of judge to the ownership of Gi (i=1, 2, …, k) according to a certain judgment rule is represented by Ri, and the cost of erroneous judgment of the sample x from Gi as Gi is represented by c (j|i) (i, j=1, 2, …, k), then there is c (i|i) =0. The conditional probability of misjudging sample x from Gi as Gi is:
p (j|i) =p (x e ri|x e Gi) = c|f i (x) dx can obtain the average misjudgment cost of any judgment rule as follows:
the discrimination rule for minimizing the average misdiscrimination cost ECM is:
x is one of Gi, if
If the average misjudgment cost of the sample judgment rule Gi is smaller than that of the other population judge to the ownership of, the sample judge to the ownership of Gi is obtained.
(4) Misjudgment rate solving
Calculating the misjudgment rate: the probability that the sample originally belonging to the ith group is erroneously determined as the jth group is represented by P (j|i) (i=1, 2), and the probability that the sample originally belonging to the jth group is erroneously determined as the ith group is represented by P (i|j) (j=1, 2); the false positive probability is err=0.5p (j|i) +0.5p (i|j).
Through the misjudgment rate and the continuous expansion and accumulation of training samples in actual operation, a proper judgment method is selected to intelligently control the adjustable weir, and when the training samples are closer to the whole body, the judgment effect is better.
The invention has the beneficial effects that:
After the drainage system is modified, the contradiction exists between drainage waterlogging prevention and the realization of pollution control functions, and particularly, during a rainstorm, how to control the adjustable weir is a difficult problem of complex drainage system operation. The invention can judge whether the adjustable weir is opened or not in real time by utilizing the easily acquired real-time monitoring data and based on the machine learning method, and ensures the drainage safety of the region and assists in manual decision on the premise of realizing pollution control as much as possible. Compared with the model simulation feedback method, the invention not only has no delay in output time, but also takes the complexity of external input such as rainfall, rainfall intensity, initial state of drainage system, confluence conditions and the like into consideration, and compared with the uncertainty of the model simulation method, the accuracy of feedback output by using the 'ash box' model is greatly improved along with the accumulation of a database, and the invention is a 'growing' method. In a word, the external data needed by the invention is simple and easy to collect in real time, training samples can be continuously expanded and accumulated, and the closer the training samples are to the whole, the better the judging effect is. The invention provides an effective method for improving the intelligent control level of the adjustable weir of the complex drainage system, and is beneficial to improving the overall operation level of the complex drainage system.
Detailed Description
Further advantages and effects of the present invention will become apparent to those skilled in the art from the disclosure of the present invention, which is described by the following specific examples.
Examples
In this embodiment, taking a certain drainage system as an example, collecting a water level value and a flow value of a pipeline with an adjustable weir position of the drainage system through a liquid level meter, a flowmeter and a camera, selecting the following variables as a sample matrix (the last data is assumed to be unknown) according to the water accumulation condition of the front pool water level of a pump station and the least adverse point of a region in the drainage system, wherein the operation type 1 indicates that the weir-lowering operation should be performed under the monitoring data of the same row, and the operation type 0 indicates that the weir-lowering operation should not be performed under the monitoring data of the same row.
Table 1: drainage system inflow point flow and weir-down operation truth table
Because the dimension and the magnitude of each variable are inconsistent, some are water levels and some are flow, the data are firstly required to be standardized so as to eliminate the limitation of the dimension and the magnitude, and the follow-up statistical analysis is convenient. The standardized processing method comprises the following steps:
the observation matrix is shown in table 1, which can be written in the form of the following matrix (19×7):
Wherein the standardized formula of x ij is
Where S ij is the variance of the variable x ij,Is the average value of the variable observation values;
Table 2 can be obtained after the normalization formula is changed.
Table 2: result table after standardized processing of monitoring data sample
The control concept of the discrimination state element of the control object of the embodiment is as follows: the control target is to realize that the rainwater enters the intercepting system at the initial stage of sewage control and interception under the premise of ensuring the flood prevention safety (no ponding is encountered in the first X year). The control elements are as follows: adjustable weir, municipal pump. The real-time data is solved through discrimination calculation, if the category belongs to true, the adjustable weir is lowered to the bottom, and flood prevention safety is ensured; and if water is accumulated at the unfavorable point, triggering the municipal pump to start.
Further, the present invention is solved by using the discrimination algorithm described in the specification.
Let G be the p-dimensional ensemble, its distributed mean vector and covariance matrix are:
Can be calculated according to Table 2
Let x= (x 1,x2,...,xp)',y＝(y1,y2,...,yp)' be two samples taken from the population G, for the data in table 2, let Σ > 0 (Σ be positive definite matrix), define the square mahalanobis distance between x, y as:
d 2(x,y)＝(x-y)′∑-1 (x-y) defines the square mahalanobis distance of x to the total G as: d 2(x,G)＝(x-μ)′∑-1 (x-. Mu.);
two p-dimensional populations G 1 and G 2 are set, mean vectors of distribution are mu 1 and mu 2 respectively, and covariance matrixes are sigma 1＞0,Σ2 > 0 respectively; samples of capacity n 1,n2 were taken from both populations, denoted x 11,x12,…,x1n1 and x 21,x22,…,x2n2, respectively. The existing data set for judging the position is marked as x, and the attribution of the x is judged in a trial mode, and the following judging rules are provided:
Therefore, whether x needs to fall a weir or not can be judged according to whether x belongs to G or not, wherein the covariance matrix of different discrimination modes is estimated as follows:
Linear discrimination solution: assuming that the prior distribution of each data is the p-element normal distribution with the same covariance matrix, obtaining the joint estimation sigma of the covariance matrix by the sample at the moment, and obtaining the judgment of the data group;
Linear diagonal solution: at this time, the diagonal matrix is used as an estimate of the covariance matrix;
And (3) secondary discrimination solution: assuming that the prior distribution of the data set is p-element normal distribution, but covariance matrixes are not completely the same, and at the moment, respectively obtaining estimation of each covariance matrix;
solving by Bayesian discrimination: describing the existing knowledge by using a priori probability, correcting the priori probability by a sample to obtain a posterior probability, and judging based on the posterior probability;
TABLE 3 discrimination results table of different methods
Calculating the misjudgment rate: the probability that the sample originally belonging to the ith group is erroneously determined as the jth group is represented by P (j|i) (i=1, 2), and the probability that the sample originally belonging to the jth group is erroneously determined as the ith group is represented by P (i|j) (j=1, 2); the false positive probability is err=0.5p (j|i) +0.5p (i|j). Finally, the misjudgment rate of different algorithm control of a certain drainage system is obtained. The calculation results are shown in Table 4:
TABLE 4 discrimination tables for different methods
According to the invention, through the misjudgment rate and the continuous expansion and accumulation of the training samples in the actual operation, the optimal judgment method is selected, the intelligent control is carried out on the adjustable weir, and when the training samples are closer to the whole body, the judgment effect is better.
The foregoing has outlined rather broadly the more detailed description of the invention in order that the detailed description thereof that follows may be better understood, and in order that the present invention may be better understood. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the invention, which are all within the scope of the invention. Accordingly, the scope of protection of the present invention is to be determined by the appended claims.
Claims (1)
1. An intelligent control method for an adjustable weir of a multi-target complex drainage system based on machine learning comprises the following steps:
1) Obtaining discrimination state element data of a control object:
Selecting state element data: collecting state element data of a drainage system, wherein the state element data comprises water level values and flow values of pipelines at the positions of adjustable weirs, water level of a front pool of a pump station in the drainage system and water accumulation conditions of least favorable points of areas;
carrying out standardized processing on the state element data to eliminate the limitation of dimension and magnitude order for subsequent discrimination and solution; the standardized processing method comprises the following steps:
assume that the observation matrix for the p-dimensional vector is as follows:
Wherein the standardized formula of x ij is
Where S ij is the variance of the variable x ij,Is the average value of the variable observation values;
2) Determining discrimination state element logic for control object
The control targets are as follows: on the premise of ensuring flood prevention safety, if no water is accumulated in the first year of X years, the rainwater enters the intercepting system at the initial stage of pollution control and interception; the control elements are as follows: an adjustable weir, municipal pump; the real-time data is solved through discrimination calculation, if the category belongs to true, the adjustable weir is lowered to the bottom, and flood prevention safety is ensured; triggering the municipal pump to start if water is accumulated at the unfavorable point;
3) Solving a discrimination algorithm, wherein the discrimination algorithm comprises the following steps:
a. solving by distance discrimination
The distance discrimination method is the mahalanobis distance from the sample to the belonging population, namely the mean vector and covariance matrix, and is nearest; the mahalanobis distance to other general bodies is quite far, and the calculation principle is as follows:
Let G be the p-dimensional ensemble, its distributed mean vector and covariance matrix are:
Let x= (x 1,x2,...,xp)',y＝(y1,y2,...,yp)' be two samples taken from the population G, let Σ > 0 (Σ be positive definite matrix), define the square mahalanobis distance between x, y as: d 2(x,y)＝(x-y)′∑-1 (x-y) defines the square mahalanobis distance of x to the total G as: d 2(x,G)＝(x-μ)′∑-1 (x-. Mu.);
Two p-dimensional populations G 1 and G 2 are set, mean vectors of distribution are mu 1 and mu 2 respectively, and covariance matrixes are sigma 1＞0,Σ2 > 0 respectively; samples with the capacity of n 1,n2 are respectively extracted from the two populations, and are marked as data sets of x 11,x12,…,x1n1 and x 21,x22,…,x2n2; with existing position judgment, and are marked as x, and the attribution of the x is tried to be judged, and the following judgment rules are provided:
b. Linear discrimination solution: assuming that the prior distribution of each data is the p-element normal distribution with the same covariance matrix, obtaining the joint estimation of the covariance matrix by the sample The judgment of the data group is obtained;
c. Linear diagonal solution: at this time, the diagonal matrix is used as an estimate of the covariance matrix;
d. And (3) secondary discrimination solution: assuming that the prior distribution of the data set is p-element normal distribution, but covariance matrixes are not completely the same, and at the moment, respectively obtaining estimation of each covariance matrix;
e. Solving by Bayesian discrimination: describing the existing knowledge by using a priori probability, correcting the priori probability by a sample to obtain a posterior probability, and judging based on the posterior probability;
4) Carrying out solving and calculating on the misjudgment rate to obtain misjudgment rates controlled by different algorithms of a certain drainage system; the misjudgment rate calculation is to represent the probability that the sample originally belonging to the ith group is misjudged as the jth group by using P (j|i) (i=1, 2), and represent the probability that the sample originally belonging to the jth group is misjudged as the ith group by using P (i|j) (j=1, 2); the false positive probability is err=0.5p (j|i) +0.5p (i|j);
5) And (3) selecting a discrimination algorithm to control the adjustable weir through the misjudgment rate and the expansion accumulation of the training samples in the actual operation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110780334.5A CN113642230B (en) | 2021-07-09 | 2021-07-09 | Multi-target complex drainage system adjustable weir intelligent control method based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110780334.5A CN113642230B (en) | 2021-07-09 | 2021-07-09 | Multi-target complex drainage system adjustable weir intelligent control method based on machine learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113642230A CN113642230A (en) | 2021-11-12 |
| CN113642230B true CN113642230B (en) | 2024-05-10 |
Family
ID=78417027
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110780334.5A Active CN113642230B (en) | 2021-07-09 | 2021-07-09 | Multi-target complex drainage system adjustable weir intelligent control method based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113642230B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114968647A (en) * | 2022-08-02 | 2022-08-30 | 中海油能源发展股份有限公司采油服务分公司 | Monitoring data deep mining method based on fault diagnosis of flash steam compressor unit |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101806861A (en) * | 2010-04-19 | 2010-08-18 | 哈尔滨工业大学 | Method for judging category of failures caused by electrical contact in sealed electromagnetic relay |
| US7961956B1 (en) * | 2009-09-03 | 2011-06-14 | Thomas Cecil Minter | Adaptive fisher's linear discriminant |
| CN108920428A (en) * | 2018-06-07 | 2018-11-30 | 辽宁工程技术大学 | A kind of fuzzy distance method of discrimination based on joint Fuzzy Extension Principle |
| CN110852364A (en) * | 2019-10-31 | 2020-02-28 | 中国煤炭地质总局勘查研究总院 | Method and device for identifying water source of water burst in mine and electronic equipment |
-
2021
- 2021-07-09 CN CN202110780334.5A patent/CN113642230B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7961956B1 (en) * | 2009-09-03 | 2011-06-14 | Thomas Cecil Minter | Adaptive fisher's linear discriminant |
| CN101806861A (en) * | 2010-04-19 | 2010-08-18 | 哈尔滨工业大学 | Method for judging category of failures caused by electrical contact in sealed electromagnetic relay |
| CN108920428A (en) * | 2018-06-07 | 2018-11-30 | 辽宁工程技术大学 | A kind of fuzzy distance method of discrimination based on joint Fuzzy Extension Principle |
| CN110852364A (en) * | 2019-10-31 | 2020-02-28 | 中国煤炭地质总局勘查研究总院 | Method and device for identifying water source of water burst in mine and electronic equipment |
Non-Patent Citations (1)
| Title |
|---|
| 边坡稳定性预测的Bayes判别分析方法及应用;史秀志;周健;郑纬;胡海燕;王怀勇;;四川大学学报(工程科学版);20100520(第03期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113642230A (en) | 2021-11-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Aderyani et al. | Short-term rainfall forecasting using machine learning-based approaches of PSO-SVR, LSTM and CNN | |
| CN112989538B (en) | Control method and control device for urban drainage system | |
| Nayak et al. | Fuzzy computing based rainfall–runoff model for real time flood forecasting | |
| Özger | Comparison of fuzzy inference systems for streamflow prediction | |
| CN113642230B (en) | Multi-target complex drainage system adjustable weir intelligent control method based on machine learning | |
| CN112084461B (en) | KNN real-time correction method based on historical flood learning | |
| CN117035201A (en) | Multi-target scheduling rule making method and system for plain river network water engineering cluster | |
| CN113050430A (en) | Drainage system control method based on robust reinforcement learning | |
| CN110728409A (en) | Flood process type similarity mining and rapid prediction method | |
| Olyaie et al. | Estimating discharge coefficient of PK-weir under subcritical conditions based on high-accuracy machine learning approaches | |
| CN115618720A (en) | Soil salinization prediction method and system based on altitude | |
| CN108922191B (en) | Travel time calculation method based on soft set | |
| CN116167513A (en) | River basin flood response time calculation method based on univariate optimization DMCA model | |
| CN102013050A (en) | MCMC method for identifying changing modes of sewage rate | |
| Dey et al. | Rainfall-runoff prediction using a Gustafson-Kessel clustering based Takagi-Sugeno Fuzzy model | |
| CN114528761A (en) | Method and system for optimizing storage stagnation relation of horizontal plain water net polder region-system outside polder | |
| CN110807475B (en) | Flood classification, identification and forecast method based on certainty coefficient | |
| CN114723336A (en) | MOP-DL-based intelligent decision-making method for underground drainage pipe network maintenance | |
| Tran et al. | Neural network based prediction models for structural deterioration of urban drainage pipes | |
| KR20170112521A (en) | Control methods of low impact development facilities using rainfall forecasting | |
| Qiao et al. | Simulation and optimal control for a long‐distance water diversion project under different rainfall types: A case study in the Middle Route of China's South‐to‐North Water Diversion Project | |
| CN117390842B (en) | Automatic calibration method for drain pipe network model based on Bayesian optimization framework | |
| CN118097969B (en) | Congestion early warning method based on short-time prediction of traffic parameters | |
| Vasiliev et al. | Sliding Window-based Particle Swarm optimization Algorithm for a Sewer Network | |
| Ravanbakhsh et al. | Implementing and evaluating various machine learning models for pipe burst prediction |
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 |