CN113792805A - Water quality evaluation method based on multi-source data fusion - Google Patents
Water quality evaluation method based on multi-source data fusion Download PDFInfo
- Publication number
- CN113792805A CN113792805A CN202111089373.7A CN202111089373A CN113792805A CN 113792805 A CN113792805 A CN 113792805A CN 202111089373 A CN202111089373 A CN 202111089373A CN 113792805 A CN113792805 A CN 113792805A
- Authority
- CN
- China
- Prior art keywords
- evidence
- water quality
- network model
- neural network
- probability distribution
- 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.)
- Granted
Links
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 title claims abstract description 171
- 238000013441 quality evaluation Methods 0.000 title claims abstract description 62
- 238000000034 method Methods 0.000 title claims abstract description 58
- 230000004927 fusion Effects 0.000 title claims abstract description 56
- 238000005315 distribution function Methods 0.000 claims abstract description 65
- 238000003062 neural network model Methods 0.000 claims abstract description 54
- 238000011156 evaluation Methods 0.000 claims abstract description 53
- 238000010606 normalization Methods 0.000 claims abstract description 12
- 238000012545 processing Methods 0.000 claims abstract description 11
- 238000012549 training Methods 0.000 claims description 23
- 230000006870 function Effects 0.000 claims description 20
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 claims description 11
- 229910052760 oxygen Inorganic materials 0.000 claims description 11
- 239000001301 oxygen Substances 0.000 claims description 11
- OAICVXFJPJFONN-UHFFFAOYSA-N Phosphorus Chemical compound [P] OAICVXFJPJFONN-UHFFFAOYSA-N 0.000 claims description 4
- XKMRRTOUMJRJIA-UHFFFAOYSA-N ammonia nh3 Chemical compound N.N XKMRRTOUMJRJIA-UHFFFAOYSA-N 0.000 claims description 4
- 229910052698 phosphorus Inorganic materials 0.000 claims description 4
- 239000011574 phosphorus Substances 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 4
- 230000002194 synthesizing effect Effects 0.000 claims description 3
- 238000001303 quality assessment method Methods 0.000 claims 3
- 238000013528 artificial neural network Methods 0.000 abstract description 26
- 238000012544 monitoring process Methods 0.000 description 8
- 210000002569 neuron Anatomy 0.000 description 7
- 239000002352 surface water Substances 0.000 description 7
- 230000008569 process Effects 0.000 description 6
- 230000015572 biosynthetic process Effects 0.000 description 4
- 238000011160 research Methods 0.000 description 4
- 238000003786 synthesis reaction Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 230000004913 activation Effects 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 3
- IJGRMHOSHXDMSA-UHFFFAOYSA-N Atomic nitrogen Chemical compound N#N IJGRMHOSHXDMSA-UHFFFAOYSA-N 0.000 description 2
- 241000282414 Homo sapiens Species 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 238000007499 fusion processing Methods 0.000 description 2
- 241000588722 Escherichia Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 239000003344 environmental pollutant Substances 0.000 description 1
- 238000013210 evaluation model Methods 0.000 description 1
- 230000002550 fecal effect Effects 0.000 description 1
- 238000012886 linear function Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 229910052757 nitrogen Inorganic materials 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 231100000719 pollutant Toxicity 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/18—Water
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Chemical & Material Sciences (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Food Science & Technology (AREA)
- Medicinal Chemistry (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention is suitable for the technical field of water quality evaluation, and provides a water quality evaluation method based on multi-source data fusion, which comprises the following steps: inputting the water quality evaluation index after pretreatment into a preset neural network model to obtain preliminary water quality judgment; carrying out normalization processing on the output value of the neural network to obtain a basic probability distribution function; and inputting the basic probability distribution function into a preset D-S evidence theoretical model to obtain a target fusion evaluation result. The method adopts three different neural networks, namely a BP neural network model, an RBF neural network model and an extreme learning machine network model, to objectively obtain the basic probability distribution of the evidence theory, further corrects the conflicting evidence by using the support degree among the evidence, iteratively corrects the fusion result, determines the target fusion evaluation result, and improves the accuracy of the evaluation result.
Description
Technical Field
The invention belongs to the technical field of water quality evaluation, and particularly relates to a water quality evaluation method based on multi-source data fusion.
Background
Water resources are the basis for human beings to live on, along with the continuous development of economy, the ecological environment is damaged while the water resource demand is increased day by day, large-area pollution to a plurality of water bodies is caused, and the contradiction between human beings and the water resources is highlighted day by day. How to protect water resources and the ecological environment on which people live is a difficult problem for all mankind. The water quality evaluation is used as qualitative or quantitative description of the water body condition, can accurately reflect the current water quality and pollution condition, plays a decisive role in water resource protection, and is an essential link for solving the water resource problem.
The research on the water quality evaluation method has achieved great results, but the development is not complete enough so far, and further intensive research is needed. Common water quality evaluation methods generally include a single factor evaluation method, a water quality index method, a fuzzy mathematical evaluation method, a neural network evaluation method, a data fusion evaluation method, and the like. The mechanism of the single-factor evaluation method is that the worst single-index classification of water quality is used for determining comprehensive water quality classification, the method is simple and clear, but the evaluation result is too conservative and cannot be comprehensively evaluated, and the accuracy of the evaluation result is poor. The water quality index model (WQI) is a comprehensive evaluation method which is widely applied, although the WQI model is easy to understand, most of the models are specific to a specific research water area, the universality is poor, and the uncertainty of the model is increased by the calculation process of WQI, particularly the judgment of parameter weight. However, according to the characteristics of multi-source, isomerism, uncertainty, nonlinearity and the like of water environment monitoring data, the data fusion technology is widely applied to water quality comprehensive evaluation. The data fusion technology makes full use of the observed data of a plurality of water quality sensors, synthesizes the redundant or complementary information of the sensors in space or time according to a certain criterion, and obtains the consistent explanation or description of the water quality.
However, the above method cannot solve the problem of poor accuracy of the water quality evaluation result.
Disclosure of Invention
In view of this, the embodiment of the invention provides a water quality evaluation method based on multi-source data fusion, so as to solve the problem of poor accuracy of an evaluation result in the prior art.
The embodiment of the invention provides a water quality evaluation method based on multi-source data fusion, which comprises the following steps:
inputting the water quality evaluation index after pretreatment into a preset neural network model to obtain the water quality category;
carrying out normalization processing on the water quality category to obtain a basic probability distribution function;
and inputting the basic probability distribution function into a preset D-S evidence theoretical model to obtain a target fusion evaluation result.
In one possible implementation manner, the preset neural network model includes a back propagation neural network model, a radial basis function neural network model and an extreme learning machine network model;
inputting the pretreated water quality evaluation index into a preset neural network model to obtain the water quality category, wherein the water quality category comprises the following steps:
and respectively inputting the pretreated water quality evaluation indexes into a back propagation neural network model, a radial basis neural network model and an extreme learning machine network model to obtain a first water quality category corresponding to the back propagation neural network model, a second water quality category corresponding to the radial basis neural network model and a third water quality category corresponding to the extreme learning machine network model.
In a possible implementation manner, the normalizing the water quality category to obtain the basic probability distribution function includes:
and respectively carrying out normalization processing on the first water quality type, the second water quality type and the third water quality type to obtain a first basic probability distribution function corresponding to the first water quality type, a second basic probability distribution function corresponding to the second water quality type and a third basic probability distribution function corresponding to the third water quality type.
In a possible implementation manner, inputting a basic probability distribution function to a preset D-S evidence theoretical model to obtain a target fusion evaluation result, including:
determining a first fusion evaluation result according to the first basic probability distribution function, the second basic probability distribution function and the third basic probability distribution function;
and performing n times of iterative computation on the first fusion evaluation result, the first basic probability distribution function, the second basic probability distribution function and the third basic probability distribution function by using a preset iterative method, and determining a target fusion evaluation result, wherein n is an integer greater than or equal to 1.
In a possible implementation manner, determining a first fusion evaluation result according to the first basic probability distribution function, the second basic probability distribution function, and the third basic probability distribution function includes:
taking the first basic probability distribution function as a first evidence, taking the second basic probability distribution function as a second evidence and taking the third basic probability distribution function as a third evidence, and calculating a first weight corresponding to the first evidence, a second weight corresponding to the second evidence and a third weight corresponding to the third evidence;
determining a fourth evidence, a fifth evidence and a sixth evidence according to the first weight and the preset weight, the second weight and the preset weight and the third weight and the preset weight respectively;
and synthesizing the fourth evidence, the fifth evidence and the sixth evidence according to a preset combination rule to obtain a first fusion evaluation result.
In one possible implementation manner, calculating a first weight corresponding to the first evidence, a second weight corresponding to the second evidence, and a third weight corresponding to the third evidence includes:
calculating a first probability distance and a first collision factor between the first evidence and the second evidence, a second probability distance and a second collision factor between the second evidence and the third evidence, and a third probability distance and a third collision factor between the first evidence and the third evidence;
determining a first degree of conflict between the first evidence and the second evidence, a second degree of conflict between the second evidence and the third evidence, and a third degree of conflict between the first evidence and the third evidence according to the first probability distance and the first conflict factor, the second probability distance and the second conflict factor, and the third probability distance and the third conflict factor, respectively;
determining a first similarity, a second similarity and a third similarity according to the first conflict degree, the second conflict degree and the third conflict degree respectively;
determining a first support degree corresponding to the first evidence, a second support degree corresponding to the second evidence and a third support degree corresponding to the third evidence according to the first similarity and the third similarity, the first similarity and the second similarity and the third similarity respectively;
and selecting the maximum support degree from the first support degree, the second support degree and the third support degree, and determining a first weight, a second weight and a third weight respectively based on the ratio of the first support degree to the maximum support degree, the ratio of the second support degree to the maximum support degree and the ratio of the third support degree to the maximum support degree.
In a possible implementation manner, before inputting the pretreated water quality evaluation index into a preset neural network model and obtaining a water quality category corresponding to the water quality evaluation index, the method further includes:
training the initial neural network model by using a training set to obtain a preset neural network model;
and (4) preprocessing the original water quality evaluation index to obtain a preprocessed water quality evaluation index.
In a possible implementation manner, training an initial neural network model by using a training set to obtain a preset neural network model includes:
selecting a plurality of sample data from original sample data as a training set;
and inputting the training set into an initial neural network model, and determining the preset neural network model when the iteration times reach preset times.
In one possible implementation, the pre-processing the original water quality evaluation index to obtain a pre-processed water quality evaluation index includes:
and carrying out normalization processing on the original water quality evaluation index by using a preset function to obtain the pretreated water quality evaluation index.
In one possible implementation, the raw water quality evaluation index includes: dissolved oxygen, permanganate index, five-day biochemical oxygen demand, ammonia nitrogen and total phosphorus.
Compared with the prior art, the embodiment of the invention has the following beneficial effects:
according to the water quality evaluation method based on multi-source data fusion, the water quality category is obtained by inputting the water quality evaluation index after pretreatment into a preset neural network model, then the water quality category is subjected to normalization treatment to obtain a basic probability distribution function, and finally the basic probability distribution function is input into a preset D-S evidence theoretical model to obtain a target fusion evaluation result. The method and the device sequentially adopt the preset neural network model and the preset D-S evidence theoretical model to process the water quality indexes, determine the target fusion evaluation result, and improve the accuracy of the evaluation result.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the embodiments or the prior art descriptions will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without inventive exercise.
Fig. 1 is a schematic flow chart of an implementation of a water quality evaluation method based on multi-source data fusion according to an embodiment of the present invention;
FIG. 2 is a schematic structural diagram of a preset neural network model according to an embodiment of the present invention;
fig. 3 is a schematic flow chart of an implementation of a water quality evaluation method based on multi-source data fusion according to another embodiment of the present invention;
FIG. 4 is a flow chart of a fusion evaluation of the improved D-S evidence theory provided by the embodiment of the present invention.
Detailed Description
In the following description, for purposes of explanation and not limitation, specific details are set forth, such as particular system structures, techniques, etc. in order to provide a thorough understanding of the embodiments of the invention. It will be apparent, however, to one skilled in the art that the present invention may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present invention with unnecessary detail.
In order to solve the problems, the application provides a water quality evaluation method based on multi-source data fusion.
In order to explain the technical means of the present invention, the following description will be given by way of specific examples.
Fig. 1 is a schematic flow chart of an implementation of a water quality evaluation method based on multi-source data fusion according to an embodiment of the present invention. As shown in fig. 1, a water quality evaluation method based on multi-source data fusion in this embodiment includes:
step S101: inputting the water quality evaluation index after pretreatment into a preset neural network model to obtain the water quality category;
step S102: carrying out normalization processing on the water quality category to obtain a basic probability distribution function;
step S103: and inputting the basic probability distribution function into a preset D-S evidence theoretical model to obtain a target fusion evaluation result.
Specifically, the surface water quality evaluation indexes are 21 indexes except for water temperature, total nitrogen and fecal coliform in the surface water environmental quality standard (GB-2002), and only a plurality of indexes with high influence degree are selected for actual water quality evaluation in consideration of the complexity of an evaluation model and the convenience of water quality index measurement. Selecting five water quality evaluation indexes of dissolved oxygen, permanganate index, five-day biochemical oxygen demand, ammonia nitrogen and total phosphorus according to the water quality characteristics and main pollutants of a research water area. The water quality evaluation standard is based on the environmental quality standard of surface water (GB-2002), and water quality evaluation is carried out according to six categories of I, II, III, IV, V and inferior V, as shown in Table 1.
TABLE 1 corresponding relationship table of water quality evaluation index and water quality category
In combination with fig. 2, based on the structure diagram of the preset neural network, the water quality evaluation index after the pretreatment is input to the preset neural network model to obtain the water quality category. The water quality category corresponding to each evaluation index is shown in table 1, for example, the value of category i corresponding to dissolved oxygen is 7.5, the value of category ii is 6, the value of category iii is 5, the value of category iv is 3, and the value of category v is 2, a preset condition for dissolved oxygen is set in the system, and the category satisfying the preset condition is taken as the category of dissolved oxygen, which is not described again.
In one embodiment, step S101 includes:
and respectively inputting the pretreated water quality evaluation indexes into a BP (Back Propagation) neural network model, an RBF (radial basis function) neural network model and an extreme learning machine network model to obtain a first water quality type corresponding to the BP neural network model, a second water quality type corresponding to the RBF neural network model and a third water quality type corresponding to the extreme learning machine network model.
The neuron number of the input layer of the three neural networks is determined by the selected water quality evaluation index, the neuron number of the output layer is determined by the water quality category number, and the water quality category value adopts binary coding, such as: 100000 for class i water and 000001 for poor class v water. The neural network training test data is generated by uniform random interpolation according to the quality standard of surface water environment (GB 3838-.
Specifically, the water quality evaluation is essentially a pattern recognition problem based on the surface water environment quality standard, the number of neurons in the input layer of the neural network is determined by the number of selected water quality evaluation factors (five water quality evaluation indexes are dissolved oxygen, permanganate index, five-day biochemical oxygen demand, ammonia nitrogen and total phosphorus), the number of neurons in the output layer is the water quality category number (I, II, III, IV, V and poor V), and the model structure is shown in FIG. 2. Specifically, the water quality classification values are represented in table 2 using binary codes.
TABLE 2 Water quality class code
BP neural network structural design: and selecting a three-layer network structure of a single hidden layer, wherein the number of neurons of the hidden layer has no definite selection rule, and the number is determined to be 11 by adopting a common empirical formula. The method comprises the steps of utilizing Matlab R2018a toolbox functions to realize neural network coding, adopting a feedback function to construct a BP neural network, selecting an S-type tangent function tansig as an implicit layer activation function in the network, selecting a linear function purelin as an output layer activation function, training by using a train LM (Levenberg-Marquardt) algorithm, setting a learning rate to be 0.05, setting an expected error to be 0.001 and setting the maximum training times of the network to be 1000 times.
Designing an RBF neural network structure: and (3) adopting a newrb function to create a radial basis network, wherein the expected error is 0.001, the radial basis function expansion speed spread value is 0.35, and other parameters are default values. Creating the RBF neural network is an attempted process that increases the number of hidden layer radial basis neurons until the desired error is met.
Designing a network structure of the extreme learning machine: and (3) compiling a function to create an extreme learning machine network, wherein the expected error is 0.001, the number of neurons in a hidden layer is 180, and the sigmoid function is selected by an activation function.
After the three network structures are obtained, the original neural network needs to be trained by adopting a training set to obtain a preset neural network, and then the output result of the preset neural network is determined by a test set, wherein the output result is a first sub-water quality class value, a second sub-water quality class value and a third sub-water quality class value respectively. The neural network training test data are generated by uniform random interpolation by using a rand function according to the quality standard of surface water environment (GB 3838-2002). 300 groups of data are generated for each type of water quality, 1800 groups of data are generated, and a training set and a testing set are divided according to a ratio of 7: 3. The data set was normalized using the mapminmax function.
In one embodiment, step S102 includes:
and respectively carrying out normalization processing on the first water quality type, the second water quality type and the third water quality type to obtain a first basic probability distribution function corresponding to the first water quality type, a second basic probability distribution function corresponding to the second water quality type and a third basic probability distribution function corresponding to the third water quality type.
Specifically, with reference to fig. 3, the preprocessed water quality evaluation index is input to a preset neural network model, and normalization processing is performed on the output result to obtain a basic probability distribution function. Because the preset neural network model comprises a BP neural network model, an RBF neural network model and an extreme learning machine network model, the numerical value category value corresponding to each neural network, namely the first water quality category, the second water quality category and the third sub-water quality category, is output through the three neural networks. And then, carrying out normalization processing on the first water quality type, the second water quality type and the third water quality type to obtain a basic probability distribution function corresponding to each water quality type, namely a first basic probability distribution function m1, a second basic probability distribution function m2 and a third basic probability distribution function m 3.
In one embodiment, step S103 includes:
step S201: and determining a first fusion evaluation result according to the first basic probability distribution function, the second basic probability distribution function and the third basic probability distribution function.
Specifically, a first basic probability distribution function is used as a first evidence, a second basic probability distribution function is used as a second evidence, and a third basic probability distribution function is used as a third evidence, and a first weight corresponding to the first evidence, a second weight corresponding to the second evidence and a third weight corresponding to the third evidence are calculated; determining a fourth evidence, a fifth evidence and a sixth evidence according to the first weight and the preset weight, the second weight and the preset weight and the third weight and the preset weight respectively; and synthesizing the fourth evidence, the fifth evidence and the sixth evidence according to a preset combination rule to obtain a first fusion evaluation result.
In an embodiment, calculating a first weight corresponding to the first evidence, a second weight corresponding to the second evidence, and a third weight corresponding to the third evidence includes:
calculating a first probability distance and a first collision factor between the first evidence and the second evidence, a second probability distance and a second collision factor between the second evidence and the third evidence, and a third probability distance and a third collision factor between the first evidence and the third evidence;
determining a first degree of conflict between the first evidence and the second evidence, a second degree of conflict between the second evidence and the third evidence, and a third degree of conflict between the first evidence and the third evidence according to the first probability distance and the first conflict factor, the second probability distance and the second conflict factor, and the third probability distance and the third conflict factor, respectively;
determining a first similarity, a second similarity and a third similarity according to the first conflict degree, the second conflict degree and the third conflict degree respectively;
determining a first support degree corresponding to the first evidence, a second support degree corresponding to the second evidence and a third support degree corresponding to the third evidence according to the first similarity and the third similarity, the first similarity and the second similarity and the third similarity respectively;
and selecting the maximum support degree from the first support degree, the second support degree and the third support degree, and determining a first weight, a second weight and a third weight respectively based on the ratio of the first support degree to the maximum support degree, the ratio of the second support degree to the maximum support degree and the ratio of the third support degree to the maximum support degree.
Specifically, the D-S evidence theory is a theory established on a non-empty set Θ, and represents a recognition framework composed of n mutually incompatible basic propositions with Θ, that is, { θ ═ θ(1),θ(2),...,θ(n)}。2ΘCalled the power set of theta, all sub-propositions in the problem domain belong to power set 2Θ。
The basic probability distribution function m is defined over Θ: 2Θ∈[0,1](ii) a m satisfies m (phi) is 0;
m (A) represents the degree of support for the occurrence of proposition A. Let m1And m2To identify the two basic probability distribution functions on the framework Θ, the combination rule of D-S evidence theory is expressed as follows:
wherein the content of the first and second substances,
for the collision factor, a larger value of k indicates a larger inter-evidence collision. The combination of the multiple evidences is actually the combination of every two evidences, and the combination rule accords with the combination law and the exchange law.
The specific steps of improving the D-S evidence theory comprise: calculating the conflict degree among the evidences to obtain the discount factors, namely the weights, of the evidences; correcting the conflict evidence; and fusing each evidence by using a D-S synthesis rule, and iteratively correcting a fusion result. The specific flow chart is shown in fig. 4.
1. Conflict measure and discount factor acquisition
m is the basic probability distribution on the recognition frame theta, and the Pignistic function BetP corresponding to mm：Θ→[0,1]Comprises the following steps:
BetP is preparedmExtension to power set 2 of ΘΘUpper, then BetPmComprises the following steps:
where | A | is the potential of set A.
According to BetPmFunction definition two basic probability assignments m1And m2The inter-Pignistic probability distance difBetP is:
measure of the degree of conflict between evidences:
similarity between evidences: sij＝1-confij (6)
Inter-evidence support:
and (3) sequencing the support degrees among the evidences, wherein the evidences with the maximum support degree are key evidences, and the discount factor (weight) is the ratio of the support degree of the evidence to the support degree of the key evidences:
2. correcting conflicting evidence
The correction of the conflicting evidence is the core of the improved method of the invention, and in order to fully utilize the effective information of the original evidence, the evidence without conflict is not corrected, and only the evidence with conflict is corrected. So-called conflicting evidence is defined as evidence with a weight less than the average evidence weight, i.e. when
Then (c) is performed.
The formula for correcting the evidence of conflict is as follows:
3. the first fusion process was as follows:
(1) determining the first evidence synthesis weight according to equation (8)
(2) Will weight
Formula (9) is substituted to correct the evidence of conflict;
(3) the corrected evidence is substituted into the formula (1) for synthesis to obtain a first fusion result (initial fusion evaluation result) R0。
Step S202: and performing n times of iterative computation on the first fusion evaluation result, the first basic probability distribution function, the second basic probability distribution function and the third basic probability distribution function by using a preset iterative method, and determining a target fusion evaluation result, wherein n is an integer greater than or equal to 1.
Specifically, the conflict degree, the similarity and the support degree between each piece of evidence and the first fusion evaluation result are calculated to obtain a new weight, the evidence is corrected by using the new weight, the second fusion evaluation result is synthesized, and when the distance between the first fusion evaluation result and the second fusion evaluation result is smaller than a specified threshold value, iteration is completed. It should be noted that, when the distance between the first fusion evaluation result and the second fusion evaluation result is not less than the designated threshold, the next fusion evaluation result needs to be calculated continuously, and the iteration is not finished until the distance between the next fusion evaluation result and the previous fusion evaluation result is less than the designated threshold.
With reference to FIG. 4, the results R of the k-1 synthesis of the original evidence are calculatedk-1Determining new evidence weight according to the conflict degree, the similarity and the support degree, then revising conflict evidence according to the weight, and calculating the fusion result R of the kth timek(ii) a Finally, judging the iterative convergence condition when
Time iteration is completed and R is outputkOtherwise, the iteration is continued.
In an embodiment, before step S102, the method further includes: training the initial neural network model by using a training set to obtain a preset neural network model; and (4) preprocessing the original water quality evaluation index to obtain a preprocessed water quality evaluation index. Wherein, utilize the training set to train initial neural network model, obtain predetermined neural network model, include: selecting a plurality of sample data from original sample data as a training set; and inputting the training set into an initial neural network model, and determining the preset neural network model when the iteration times reach preset times.
Based on the method, the simulation experiment is evaluated, and the method comprises the following specific steps:
the method disclosed by the invention is verified by adopting simulation software Matlab R2018a, and five monitoring sections of YueCheng reservoir outlet, Quzhou, Dongwu outlet, Houkui Wuqiao and Aixinzhu are selected for experimental verification by utilizing surface water quality data of Jinan region (Handan City, Pichentai City) in 2021 year 4 published by a Chinese environment monitoring central office. Table 3 shows the actual monitoring data of the five monitoring section sensors.
TABLE 3 monitoring section data in the wing area
The water quality conditions of the five monitoring sections were evaluated by different methods, and the results are shown in table 4.
TABLE 4 Water quality evaluation results of different methods
The water quality evaluation results in table 4 show that the single-factor evaluation method is too conservative to fully utilize the information provided by each water quality monitoring index. Different neural networks have different judgment results due to different parameter selection, learning algorithms and the like. All five water quality indexes at the outlet of the Yueyuan reservoir are in the range of I-type water, and various methods can accurately judge the water quality grade; two indexes of the curve period belong to I-class water, two indexes belong to III-class water, one index belongs to IV-class water, and the comprehensive evaluation result shows that the III-class water quality is more reasonable; the Dongwu export contains three I-class water quality indexes, and the other two II-class indexes are closer to the I-class standard value, so that the water is judged to be I-class water; the three indexes of the Wuqiao in the later west are I-class water, the two indexes are III-class water, and the result is that II-class water is more reasonable; similarly, the water quality of Esinozhuang is judged to be class III. In conclusion, the single-factor evaluation method is most conservative, the results of the three neural network methods are relatively reasonable, the evaluation method is high in comprehensiveness, and the results are most reasonable.
The water quality evaluation results of the three types of neural networks are not completely consistent, even if the same result is obtained, the output values of the neural networks corresponding to the water quality categories are not consistent and do not reach an ideal output value 1, so that the evaluation results are uncertain, and the accuracy needs to be improved. The method provided by the invention utilizes a D-S evidence theory to fuse three neural network outputs, and finally obtains a determined evaluation result.
Taking Exinzhuang as an example, the detailed fusion process is described in detail. Firstly, an evidence body is constructed, and the identification framework of water quality evaluation is six types of water quality, namely type I (A1), type II (A2), type III (A3), type IV (A4), type V (A5) and type inferior V (A6). The output results of the neural network corresponding to the Esinozhuang are normalized to obtain basic probability distribution functions m1, m2 and m3, which respectively represent the evaluation results of the BP neural network, the RBF neural network and the extreme learning machine, as shown below.
Basic probability distribution function of Escherchia:
m1：0.1468 0.3500 0.4635 0.0336 0.0057 0.0004
m2：0.0233 0.1298 0.6012 0.0169 0.0302 0.1986
m3：0.1153 0.0422 0.0055 0.5868 0.2108 0.0394
evidence 3 of the Eschinzhuang clearly conflicts with evidence 1 and evidence 2, the evidence 3 supports A4, and both the evidence 1 and the evidence 2 support A3 but the support degree is low, so that the decision judgment is not facilitated. The results are shown in Table 5 after the theoretical fusion of D-S evidence.
TABLE 5 Esinzhuang Water quality fusion results
By comparing and analyzing the method, the Murphy method and the traditional D-S evidence theory, the fusion result of the traditional evidence theory in the table 5 is inconsistent with the other two methods, and the main reason is that the three fused evidences conflict and a correct result cannot be obtained. On the contrary, the other two methods can effectively process the conflict between the evidences so as to make a correct decision, and the method of the invention is easy to find out that the identification precision is higher.
It should be understood that, the sequence numbers of the steps in the foregoing embodiments do not imply an execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.
In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and reference may be made to the related descriptions of other embodiments for parts that are not described or illustrated in a certain embodiment.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
The above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present invention, and are intended to be included within the scope of the present invention.
Claims (10)
1. A water quality evaluation method based on multi-source data fusion is characterized by comprising the following steps:
inputting the water quality evaluation index after pretreatment into a preset neural network model to obtain the water quality category;
carrying out normalization processing on the water quality categories to obtain a basic probability distribution function;
and inputting the basic probability distribution function into a preset D-S evidence theoretical model to obtain a target fusion evaluation result.
2. The method of claim 1, wherein the pre-set neural network model comprises a back propagation neural network model, a radial basis neural network model, and an extreme learning machine network model;
inputting the pretreated water quality evaluation index into a preset neural network model to obtain the water quality category, wherein the water quality category comprises the following steps:
and respectively inputting the pretreated water quality evaluation indexes into the back propagation neural network model, the radial basis function neural network model and the extreme learning machine network model to obtain a first water quality type corresponding to the back propagation neural network model, a second water quality type corresponding to the radial basis function neural network model and a third water quality type corresponding to the extreme learning machine network model.
3. The method of claim 2, wherein the normalizing the water quality categories to obtain a basic probability distribution function comprises:
and respectively carrying out normalization processing on the first water quality type, the second water quality type and the third water quality type to obtain a first basic probability distribution function corresponding to the first water quality type, a second basic probability distribution function corresponding to the second water quality type and a third basic probability distribution function corresponding to the third water quality type.
4. The method according to claim 3, wherein the inputting the basic probability distribution function into a preset D-S evidence theoretical model to obtain a target fusion evaluation result comprises:
determining a first fusion evaluation result according to the first basic probability distribution function, the second basic probability distribution function and the third basic probability distribution function;
and performing n times of iterative computation on the first fusion evaluation result, the first basic probability distribution function, the second basic probability distribution function and the third basic probability distribution function by using a preset iterative method, and determining a target fusion evaluation result, wherein n is an integer greater than or equal to 1.
5. The method of claim 4, wherein determining a first fused assessment result based on the first basic probability distribution function, the second basic probability distribution function, and the third basic probability distribution function comprises:
taking the first basic probability distribution function as a first evidence, the second basic probability distribution function as a second evidence and the third basic probability distribution function as a third evidence, and calculating a first weight corresponding to the first evidence, a second weight corresponding to the second evidence and a third weight corresponding to the third evidence;
determining a fourth evidence, a fifth evidence and a sixth evidence according to the first weight and the preset weight, the second weight and the preset weight and the third weight and the preset weight respectively;
and synthesizing the fourth evidence, the fifth evidence and the sixth evidence according to a preset combination rule to obtain the first fusion evaluation result.
6. The method of claim 5, wherein the calculating a first weight for the first evidence, a second weight for the second evidence, and a third weight for the third evidence comprises:
calculating a first probability distance and a first collision factor between the first evidence and the second evidence, a second probability distance and a second collision factor between the second evidence and the third evidence, and a third probability distance and a third collision factor between the first evidence and the third evidence;
determining a first degree of conflict between the first evidence and the second evidence, a second degree of conflict between the second evidence and the third evidence, and a third degree of conflict between the first evidence and the third evidence, based on the first probability distance and the first conflict factor, the second probability distance and the second conflict factor, and the third probability distance and the third conflict factor, respectively;
determining a first similarity, a second similarity and a third similarity according to the first conflict degree, the second conflict degree and the third conflict degree respectively;
determining a first support degree corresponding to the first evidence, a second support degree corresponding to the second evidence and a third support degree corresponding to the third evidence according to the first similarity and the third similarity, the first similarity and the second similarity and the third similarity respectively;
selecting a maximum support degree from the first support degree, the second support degree and the third support degree, and determining a first weight, a second weight and a third weight based on a ratio of the first support degree to the maximum support degree, a ratio of the second support degree to the maximum support degree and a ratio of the third support degree to the maximum support degree, respectively.
7. The method according to any one of claims 1 to 6, wherein before inputting the pre-processed water quality evaluation index into a preset neural network model and obtaining the water quality type corresponding to the water quality evaluation index, the method further comprises:
training an initial neural network model by using a training set to obtain the preset neural network model;
and preprocessing the original water quality evaluation index to obtain the preprocessed water quality evaluation index.
8. The method of claim 7, wherein the training an initial neural network model using a training set to obtain the predetermined neural network model comprises:
selecting a plurality of sample data from original sample data as the training set;
and inputting the training set into the initial neural network model, and determining the preset neural network model when the iteration times reach preset times.
9. The method of claim 7, wherein the pre-treating the raw water quality assessment indicator to obtain the pre-treated water quality assessment indicator comprises:
and carrying out normalization processing on the original water quality evaluation index by using a preset function to obtain the pretreated water quality evaluation index.
10. The method of claim 9, wherein the raw water quality assessment indicators comprise: dissolved oxygen, permanganate index, five-day biochemical oxygen demand, ammonia nitrogen and total phosphorus.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111089373.7A CN113792805B (en) | 2021-09-16 | 2021-09-16 | Water quality evaluation method based on multi-source data fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111089373.7A CN113792805B (en) | 2021-09-16 | 2021-09-16 | Water quality evaluation method based on multi-source data fusion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113792805A true CN113792805A (en) | 2021-12-14 |
| CN113792805B CN113792805B (en) | 2024-05-10 |
Family
ID=78878672
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111089373.7A Active CN113792805B (en) | 2021-09-16 | 2021-09-16 | Water quality evaluation method based on multi-source data fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113792805B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106934221A (en) * | 2017-02-27 | 2017-07-07 | 华南理工大学 | A kind of water quality assessment sorting technique based on neutral net |
| CN107729920A (en) * | 2017-09-18 | 2018-02-23 | 江苏海事职业技术学院 | A kind of method for estimating state combined based on BP neural network with D S evidence theories |
| CN108764520A (en) * | 2018-04-11 | 2018-11-06 | 杭州电子科技大学 | A kind of water quality parameter prediction technique based on multilayer circulation neural network and D-S evidence theory |
| WO2020147265A1 (en) * | 2019-01-14 | 2020-07-23 | 南京信息工程大学 | Mobile electronic commerce recommendation method and system employing multisource information fusion |
-
2021
- 2021-09-16 CN CN202111089373.7A patent/CN113792805B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106934221A (en) * | 2017-02-27 | 2017-07-07 | 华南理工大学 | A kind of water quality assessment sorting technique based on neutral net |
| CN107729920A (en) * | 2017-09-18 | 2018-02-23 | 江苏海事职业技术学院 | A kind of method for estimating state combined based on BP neural network with D S evidence theories |
| CN108764520A (en) * | 2018-04-11 | 2018-11-06 | 杭州电子科技大学 | A kind of water quality parameter prediction technique based on multilayer circulation neural network and D-S evidence theory |
| WO2020147265A1 (en) * | 2019-01-14 | 2020-07-23 | 南京信息工程大学 | Mobile electronic commerce recommendation method and system employing multisource information fusion |
Non-Patent Citations (2)
| Title |
|---|
| 崔东文;: "BP神经网络在盘龙河水质综合评价中的应用", 水科学与工程技术, no. 02 * |
| 梁楠;邹志红;: "结合新型模糊支持向量机和证据理论的多传感器水质数据融合", 电讯技术, no. 03 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113792805B (en) | 2024-05-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | Long short-term memory network based on neighborhood gates for processing complex causality in wind speed prediction | |
| CN112766078B (en) | GRU-NN power load level prediction method based on EMD-SVR-MLR and attention mechanism | |
| CN111860982A (en) | Wind power plant short-term wind power prediction method based on VMD-FCM-GRU | |
| CN112149879B (en) | New energy medium-and-long-term electric quantity prediction method considering macroscopic volatility classification | |
| CN111814956B (en) | Multi-task learning air quality prediction method based on multi-dimensional secondary feature extraction | |
| CN108090629B (en) | Load prediction method and system based on nonlinear autoregressive neural network | |
| CN111832839B (en) | Energy consumption prediction method based on sufficient incremental learning | |
| CN114912666A (en) | Short-time passenger flow volume prediction method based on CEEMDAN algorithm and attention mechanism | |
| CN111141879B (en) | Deep learning air quality monitoring method, device and equipment | |
| CN115964932A (en) | Gas prediction method based on EMD-BilSTM-Attention mechanism transformer digital twin model | |
| CN115496257A (en) | Short-term vehicle speed prediction based on space-time fusion | |
| CN115409258A (en) | Hybrid deep learning short-term irradiance prediction method | |
| CN109376937B (en) | Self-adaptive scheduling end-of-term water level prediction method based on ensemble empirical mode decomposition | |
| CN111667189A (en) | Construction engineering project risk prediction method based on one-dimensional convolutional neural network | |
| CN114707692A (en) | Wetland effluent ammonia nitrogen concentration prediction method and system based on hybrid neural network | |
| CN113792805A (en) | Water quality evaluation method based on multi-source data fusion | |
| CN112232570A (en) | Forward active total electric quantity prediction method and device and readable storage medium | |
| CN111310974A (en) | Short-term water demand prediction method based on GA-ELM | |
| CN116316600A (en) | Wind power day-ahead prediction self-adaptive filtering method based on seasonal typing | |
| CN116070768A (en) | Short-term wind power prediction method based on data reconstruction and TCN-BiLSTM | |
| CN114139783A (en) | Wind power short-term power prediction method and device based on nonlinear weighted combination | |
| CN114492964A (en) | Photovoltaic power ultra-short term probability prediction method based on wavelet decomposition and optimized deep confidence network | |
| Li et al. | Smoothed deep neural networks for marine sensor data prediction | |
| Xuefang et al. | Study on the risk assessment of real estate project based on BP neural network | |
| Duan et al. | Research on the stability of NDGM model with the fractional order accumulation and its optimization |
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 |