Water quality prediction system based on carbon-water circulation in karst area

Abstract

The invention discloses a karst area water quality prediction system based on carbon-water circulation, and relates to the technical field of environmental protection, a data collection module respectively collects carbon circulation parameters and water circulation parameters in a fixed time period, and the collected parameters are subjected to key parameter screening through a parameter screening module; the carbon cycle key parameters and the water cycle key parameters obtained through screening are sent to a carbon cycle water quality evaluation module and a water cycle water quality evaluation module through a main control module respectively, and a carbon cycle water quality evaluation value and a water cycle water quality evaluation value are obtained through calculation respectively; and the comprehensive evaluation module obtains a final water quality evaluation value according to the carbon cycle water quality evaluation value and the water cycle water quality evaluation value. According to the method, water quality predictive evaluation of the karst area water body is realized by respectively combining parameters related to water quality evaluation in carbon circulation and water circulation.

Description

Technical Field

[0001] The present invention relates to the technical field of environmental protection, and more specifically, to a water quality prediction system based on the carbon-water cycle in karst areas. Background Art

[0002] The unique hydrogeological characteristics of karst areas are a special terrain formed by soluble rocks (such as limestone, dolomite, etc.) under the action of water, which are widely distributed around the world. The water resource management in karst areas faces many challenges, which mainly stem from its unique hydrogeological characteristics and complex ecosystem. Among them, the interaction between the carbon cycle and the water cycle in karst areas has a significant impact on water quality changes.

[0003] In karst areas, carbonic acid formed by the combination of carbon dioxide in precipitation and rainwater promotes the dissolution process of rocks. This process not only changes the chemical composition of water bodies, but also significantly affects the hardness and pH value of groundwater, thus having a profound impact on the aquatic ecosystem. Correspondingly, the metabolic activities of plants and microorganisms in the ecosystem further affect the carbon cycle, resulting in dynamic changes in water quality; the water cycle in karst areas is highly uneven, and the precipitation, soil type and terrain changes make the water recharge and flow processes complex and variable. Rainwater affects the water quality through surface infiltration and groundwater recharge, and may carry surface pollutants (such as pesticides, heavy metals, etc.) during rainfall, thus leading to water quality deterioration. Therefore, understanding the characteristics of the water cycle and its impact on water quality will help better predict and manage water resources.

[0004] Therefore, how to predictively evaluate the water quality of water bodies in karst areas is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a water quality prediction system based on the carbon-water cycle in karst areas, which realizes the predictive evaluation of the water quality of water bodies in karst areas by combining the parameters related to water quality assessment in the carbon cycle and the water cycle respectively.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A water quality prediction system based on the carbon-water cycle in a karst area, comprising: a data collection module, a parameter screening module, a main control module, a carbon cycle water quality assessment module, a water cycle water quality assessment module and a comprehensive assessment module. The data collection module respectively collects carbon cycle parameters and water cycle parameters within a fixed period, and performs key parameter screening on the collected parameters through the parameter screening module. The obtained key carbon cycle parameters and key water cycle parameters are respectively sent to the carbon cycle water quality assessment module and the water cycle water quality assessment module through the main control module, and the carbon cycle water quality assessment value and the water cycle water quality assessment value are respectively calculated. The comprehensive assessment module obtains the final water quality assessment value according to the carbon cycle water quality assessment value and the water cycle water quality assessment value.

[0008] Preferably, the parameter screening module uses the ReliefF algorithm to perform regression analysis on multi-parameter continuous values, and obtains the parameter types with greater correlation with water quality through feature weight screening, specifically as follows:

[0009] Difference calculation module, calculating the parameter difference of the nearest neighbor sample features, D i 、D j is the acquisition value of the v-th parameter at times i and j, and S v is the standardized unit of this parameter. The parameter difference diff(D i,v , D j,v ) = (D i,v , D j,v ) / S v ;

[0010] Sampling module, determining the total data sample set D and the feature set M, and initializing according to the sampling times m and the selected k nearest neighbor samples; randomly select a sample R in the total data sample set D, and find k nearest neighbor samples H j , j = 1, 2, 3... k, and then find k nearest neighbor samples M j (C) from the different-class sample set of the sample R, and repeat sampling m times;

[0011] Parameter matrix calculation module, calculating the weight matrix of each parameter:

[0012]

[0013] Among them, p(C) represents the probability of the occurrence of category C;

[0014] Screening module, calculating the weight matrix of each parameter in the feature set M and sorting them, setting the feature weight threshold δ = avg(W), and removing the parameters with weights lower than the feature weight threshold as key parameters.

[0015] Preferably, the carbon cycle water quality assessment module includes data dimensionality reduction and feature extraction of the screened parameters based on the principal component analysis method, and calculates the carbon cycle water quality assessment value, specifically:

[0016] The centralization processing module, the parameter set corresponding to the carbon cycle after screening is F = (f1, f2, …, f n ) T ∈R t×n , f i is the time series data of the i-th parameter. First, perform data centralization processing on the parameter set F to make the mean return to 0 and the variance return to 1, and obtain the decentralized matrix F * ;

[0017] The projection calculation module calculates the eigenvalues λ i and eigenvectors ξ of the corresponding covariance matrix n×n, sorts the eigenvalues λ i from large to small, calculates the cumulative influence rate of different numbers of characteristic parameters Select the eigenvectors corresponding to the first P eigenvalues exceeding the influence threshold to form the projection matrix

[0018] The influence index calculation module multiplies the processed sample matrix by the projection matrix to obtain the dimensionality-reduced principal component data set X = (x1, x2, …, x p ) T , X is defined as the principal component elements of the parameter set F, and x1, x2, … are called the first principal component, the second principal component, … in turn; use deviation normalization to normalize the principal component parameters, and the calculation method is x max , x min are the maximum and minimum values of the corresponding parameter sequences, and the carbon cycle water quality influence index CQI is obtained through weight fusion. The calculation formula is ρ j is the weight value of the P parameters after dimensionality reduction;

[0019] The carbon cycle water quality assessment value calculation module predicts the water quality state by combining the pre-trained HMM model with the carbon cycle water quality influence index, obtains the water quality grade, calculates the likelihood value according to the water quality grade through the Viterbi algorithm, obtains the differences between multiple carbon cycle water quality influence indexes closest to the likelihood value, and predicts the differences between the indexes using the weighted average method to obtain the carbon cycle water quality assessment value.

[0020] Preferably, the carbon cycle water quality assessment value calculation module specifically includes:

[0021] Water quality status prediction module, constructing the HMM model, using the carbon cycle water quality impact index CQI as the training set for model training to evaluate the water quality status corresponding to the current index. The water quality status is divided into 5 grades: excellent, good, medium, poor, and inferior, corresponding to 5 hidden states of the HMM model. Taking CQI as the observation sequence O = {O1, O2, …, O t}, estimating the parameter values of the model HMM: λ = (A, B, π), making the probability of the observation sequence P(O|λ) the largest. Based on the forward-backward algorithm, defining the forward probability variable α t (j), the backward probability variable β t (j) and the probability variable γ t (j);

[0022] α t (j) = P(O1, O2, …, O t , i t = q j |λ), β t (j) = P(O t+1 , O t+2 , …, O T |i t = q j , λ);

[0023] γ t (j) = P(i t = q j |O, λ), t = 1, 2, …, T j = 1, 2, …, N;

[0024] Substitute the above parameters into the expectation maximization algorithm to calculate and solve P(O|λ), and obtain the probability relationship of the water body in different states, calculate the state expectation value, obtain the parameter re-estimation formula of the HMM, and continuously iterate and update the model until the optimal model is obtained κ is a pre-set convergence condition, and the water quality status is evaluated through the optimal model ;

[0025] Evaluation value calculation module, calculating the likelihood value through the Viterbi algorithm, obtaining the differences of multiple carbon cycle water quality impact indexes closest to this likelihood value, and predicting the differences between indexes using the weighted average method to obtain the carbon cycle water quality evaluation value:

[0026]

[0027] where CQI i is the i-th historical index similar to the index CQI t at time t, and W i represents the index difference CQI i+1 - CQI iThe proportion, N represents the number of weighting times, LL t represents the exponential likelihood value at time t, LL i represents the one t historically exponential likelihood value that is the i-th closest to LL.

[0028] Preferably, the water cycle water quality assessment module specifically includes:

[0029] A coding table acquisition unit, which obtains corresponding parameter classifications according to the key parameters of the water cycle to form parameter indicators, determines the evaluation quantities of each parameter in the water cycle according to the parameter indicators, constructs a multi-parameter evaluation coding table, and obtains the deterioration degree of each parameter;

[0030] A standard acquisition unit, which combines the weight values and deterioration degrees of the evaluation quantities of each parameter with the corresponding standard rules to obtain the deduction scores of single-parameter indicators, obtains a clear parameter evaluation standard, and formulates a multi-parameter indicator rule table;

[0031] An overall evaluation unit, which then formulates a water cycle water quality evaluation rule table based on the deduction situation of each parameter, conducts a status evaluation on the karst area, and obtains a water cycle water quality evaluation value.

[0032] Preferably, the overall evaluation unit specifically includes:

[0033]

[0034] Among them, E s is the water cycle water quality evaluation value, Y represents the total deduction score of each parameter of the key parameters of the water cycle, Y max represents the total deduction score of all parameters of the key parameters of the water cycle, N c represents the corresponding evaluation score of the fuzzy evaluation set, W s represents the weight coefficient of the overall water quality evaluation under different water cycle conditions.

[0035] Preferably, the comprehensive evaluation module calculates the final water quality evaluation value according to the water cycle water quality evaluation value, the carbon cycle water quality evaluation value and their corresponding weights by the weighted summation method, and conducts a grade evaluation according to the preset water quality scoring standard.

[0036] As can be seen from the above technical solutions, compared with the prior art, the present invention discloses a water quality prediction system based on the carbon-water cycle in karst areas, comprehensively considering the impacts of the carbon cycle and the water cycle on the water quality in karst areas, breaking through the limitations of traditional single-factor evaluation. In the carbon cycle of karst areas, the dissolution of rocks is affected by carbon dioxide, which changes the chemical composition of water bodies. In the water cycle, processes such as precipitation and infiltration affect water quality. The system comprehensively considers these factors to more accurately grasp the changes in water quality. The system collects a variety of parameters, which are used for evaluation after screening, improving the evaluation efficiency and accuracy, and providing data support for accurately evaluating water quality; the carbon cycle water quality evaluation module uses the principal component analysis method for dimensionality reduction and the HMM model for prediction, which can effectively process complex data and accurately evaluate the impact of the carbon cycle on water quality. The principal component analysis method extracts key features, and the HMM model analyzes dynamic changes. The combination of the two improves the accuracy and reliability of the evaluation. The water cycle water quality evaluation module constructs a coding table and formulates rules to comprehensively evaluate the impacts of various parameters, determines the evaluation quantity according to parameter classification, combines weights and standards to determine the deduction score, and overall evaluates the water quality, providing a basis for water resource management. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on the provided drawings.

[0038] Figure 1 It is a schematic structural diagram provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0039] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0040] The embodiments of the present invention disclose a water quality prediction system based on the carbon-water cycle in karst areas, as Figure 1As shown in the figure, it includes: a data collection module, a parameter screening module, a main control module, a carbon cycle water quality assessment module, a water cycle water quality assessment module, and a comprehensive assessment module. The data collection module collects carbon cycle parameters and water cycle parameters within a fixed time period respectively, and screens the collected parameters through the parameter screening module to obtain key parameters. The key carbon cycle parameters and key water cycle parameters obtained by screening are sent to the carbon cycle water quality assessment module and the water cycle water quality assessment module respectively through the main control module, and the carbon cycle water quality assessment value and the water cycle water quality assessment value are calculated respectively. The comprehensive assessment module obtains the final water quality assessment value based on the carbon cycle water quality assessment value and the water cycle water quality assessment value.

[0041] In a specific embodiment, the carbon cycle parameters include: soil organic carbon content and decomposition rate, plant carbon absorption and release amount, dissolved inorganic carbon (DIC) and dissolved organic carbon (DOC) concentrations in water bodies, karstification carbon flux, etc.; the water cycle parameters include: precipitation amount and precipitation frequency, surface runoff volume and velocity, groundwater level and groundwater recharge amount, migration and transformation parameters of pesticides and heavy metals in water bodies, pesticide and heavy metal contents in atmospheric precipitation, leaching amount of pesticides and heavy metals in soil, pesticide and heavy metal contents in surface runoff, etc.

[0042] In a specific embodiment, the parameter screening module uses the ReliefF algorithm to perform regression analysis on multi-parameter continuous values, and obtains the parameter types with greater correlation with water quality through feature weight screening, as follows:

[0043] Difference calculation module, calculating the parameter difference of the nearest neighbor sample features, D i 、D j is the acquisition value of the v-th parameter at times i and j, S v is the standardized unit of this parameter, and the parameter difference diff(D i,v ,D j,v )=(D i,v ,D j,v ) / S v ;

[0044] Sampling module, determining the total data sample set D and the feature set M, and initializing according to the sampling times m and the selected k nearest neighbor samples; randomly select a sample R in the total data sample set D, and find k nearest neighbor samples H j , j = 1, 2, 3... k, and then find k nearest neighbor samples M j (C) from the different-class sample set of the sample R, and repeat the sampling m times;

[0045] Parameter matrix calculation module, calculating the weight matrix of each parameter:

[0046]

[0047] Among them, p(C) represents the probability of the occurrence of category C;

[0048] A screening module calculates a weight matrix for each parameter in the feature set M and sorts them, sets the feature weight threshold δ = avg(W), and removes the parameters with weights lower than the feature weight threshold as key parameters.

[0049] In a specific embodiment, the carbon cycle water quality assessment module includes data dimensionality reduction and feature extraction of the screened parameters based on the principal component analysis method, and calculates the carbon cycle water quality assessment value, specifically:

[0050] A centralization processing module, the parameter set corresponding to the carbon cycle after screening is F = (f1, f2,..., f n ) T ∈R t×n , f i is the time series data of the i-th parameter. First, perform data centralization processing on the parameter set F to make the mean return to 0 and the variance return to 1, and obtain the decentralized matrix F * ;

[0051] A projection calculation module calculates the eigenvalues λ i and eigenvectors ξ of the corresponding covariance matrix n×n, sorts the eigenvalues λ i from largest to smallest, and calculates the cumulative influence rate of different numbers of feature parameters Select the eigenvectors corresponding to the top P eigenvalues exceeding the influence threshold to form a projection matrix

[0052] An influence index calculation module multiplies the processed sample matrix by the projection matrix to obtain the dimensionality-reduced principal component data set X = (x1, x2,..., x p ) T , X is defined as the principal component elements of the parameter set F, and x1, x2,... are called the first principal component, the second principal component,... in turn; use deviation normalization to normalize the principal component parameters, and the calculation method is x max , x min are the maximum and minimum values of the corresponding parameter sequences, and the carbon cycle water quality influence index CQI is obtained through weight fusion. The calculation formula is ρ j is the weight value of the P parameters after dimensionality reduction;

[0053] The carbon cycle water quality evaluation value calculation module predicts the water quality status by combining a pre-trained HMM model with the carbon cycle water quality impact index, obtains the water quality grade, calculates the likelihood value through the Viterbi algorithm based on the water quality grade, obtains the differences of multiple carbon cycle water quality impact indexes closest to the likelihood value, and predicts the differences between the indexes using the weighted average method to obtain the carbon cycle water quality evaluation value.

[0054] In a specific embodiment, the carbon cycle water quality evaluation value calculation module specifically includes:

[0055] The water quality status prediction module constructs an HMM model, uses the carbon cycle water quality impact index CQI as the training set for model training to evaluate the water quality status corresponding to the current index, divides the water quality status into 5 grades: excellent, good, medium, poor, and inferior, corresponding to 5 hidden states of the HMM model, takes CQI as the observation sequence O = {O1, O2, …, O t}, i t = q j |λ is used to describe the hidden state of the water quality at time t under the given model λ, estimates the parameter values of the model HMM: λ = (A, B, π), maximizes the probability of the observation sequence P(O|λ), and based on the forward-backward algorithm, defines the forward probability variable α t (j), the backward probability variable β t (j) and the probability variable γ t (j);

[0056] α t (j) = P(O1, O2, …, O t , i t = q j |λ), β t (j) = P(O t+1 , O t+2 , …, O T |i t = q j , λ);

[0057] γ t (j) = P(i t = q j |O, λ), t = 1, 2, …, T j = 1, 2, …, N;

[0058] Substitute the above parameters into the expectation maximization algorithm to calculate and solve P(O|λ), specifically:

[0059]

[0060] Under the entire observation sequence and the given model, define that the water quality is in the state q j at time t and in the state q at time t + 1k The probability is ξ(j, k), and there is

[0061]

[0062] According to the derivation of the probability formula, there is:

[0063] p(i t = q j , i t+1 = q j , 0|λ) = a jk b k (0 t+1 )α t (j)β t+1 (j);

[0064]

[0065] It is obtained that:

[0066]

[0067] Calculate the state expectation value: Assume that the observed value at time t is v1, 1 ≤ t ≤ T. Summing γ t (j) directly with respect to t can obtain the average value of the number of times the device is in state q j , that is, the conditional expectation of the variable γ t (j) and the state expectation of γ t (j)

[0068] According to the above calculation results, the parameter re - estimation formula of HMM is:

[0069]

[0070] Obtain the probability relationship of the water body in different states, calculate the state expectation value, obtain the parameter re - estimation formula of HMM, and continuously iterate and update the model until an optimal model is obtained κ is a pre - set convergence condition, and the water quality state is evaluated through the optimal model ;

[0071] An evaluation value calculation module calculates the likelihood value through the Viterbi algorithm, obtains the differences of multiple carbon cycle water quality impact indices closest to this likelihood value, and predicts the differences between the indices using the weighted average method to obtain the carbon cycle water quality evaluation value:

[0072]

[0073] Among them, CQI i is the index CQI at time tt The i-th similar historical index, W i Denote the index difference CQI i+1 -CQI i The proportion, N represents the weighted number of times, LL t Denote the index likelihood value at time t, LL i Denote the same as LL t The likelihood value of the i-th nearest historical index.

[0074] In a specific embodiment, the water cycle water quality assessment module specifically includes:

[0075] The coding table acquisition unit obtains the corresponding parameter classification according to the key parameters of the water cycle to form parameter indicators, determines the evaluation quantities of each parameter of the water cycle according to the parameter indicators, constructs a multi-parameter evaluation coding table, and obtains the deterioration degree of each parameter;

[0076] The standard acquisition unit obtains the deduction score of the single parameter index according to the weight value and deterioration degree of each parameter evaluation quantity in combination with the corresponding standard rules, obtains the clear parameter evaluation standard, and formulates a multi-parameter index rule table;

[0077] The overall evaluation unit then formulates a water cycle water quality evaluation rule table according to the deduction situation of each parameter, evaluates the state of the karst area, and obtains the water cycle water quality evaluation value.

[0078] In a specific embodiment, the multi-parameter evaluation coding table is shown in Table 1:

[0079] Table 1 Multi-parameter evaluation coding table

[0080]

[0081]

[0082] In a specific embodiment, the overall evaluation unit specifically includes:

[0083]

[0084] Among them, E s Is the water cycle water quality evaluation value, Y represents the total deduction score of each parameter of the key parameters of the water cycle, Y max Represents the total deduction score of all parameters of the key parameters of the water cycle, N c Represents the corresponding evaluation score of the fuzzy evaluation set; W s Represents the weight coefficient of the overall water quality evaluation under different water cycle conditions. Among them, the water cycle water quality evaluation value is the water quality evaluation for a future period of time.

[0085] In a specific embodiment, the comprehensive evaluation module calculates the final water quality evaluation value according to the water cycle water quality evaluation value, the carbon cycle water quality evaluation value and their corresponding weights by the weighted summation method, and conducts grade evaluation according to the preset water quality scoring standard, including:

[0086] The weight determination unit uses the Analytic Hierarchy Process (AHP) to construct a hierarchical structure model that includes the impacts of the carbon cycle and the water cycle on the water quality in the karst area. Invite experts in the fields of karst area hydrology, geology, ecology, etc. to compare the importance of each factor pairwise and construct a judgment matrix. By calculating the eigenvector of the judgment matrix and conducting a consistency test, determine the weight W 碳 of the carbon cycle water quality evaluation value and the weight W 水 of the water cycle water quality evaluation value to ensure that the weight distribution is scientific and reasonable. The weight determination takes into account the characteristics of different regions in the karst area. In areas where karst fissures are developed and the carbon cycle is active, appropriately increase the carbon cycle weight; in areas with abundant precipitation and a large impact of surface runoff on water quality, increase the water cycle weight.

[0087] The final water quality prediction value calculation unit uses the weighted summation formula to calculate the final water quality prediction value Q. The formula is Q = W 水 ·E s +W 碳 ·(CQI t+1 -CQI t ), where W 水 +W 碳 = 1. This formula comprehensively considers the impacts of the carbon-water cycle on water quality and intuitively reflects the comprehensive water quality status in the karst area. Based on the calculated final water quality prediction value, combined with the water quality management objectives and historical data in the karst area, divide the water quality prediction grades, such as excellent (Q≥90), good (80≤Q<90), medium (60≤Q<80), poor (40≤Q<60), and inferior (Q≤40). Through the clear grade division, it is convenient for managers to quickly understand the water quality status in the karst area and provide clear guidance for subsequent decision-making.

[0088] In this specification, each embodiment is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. For the same or similar parts among the embodiments, reference can be made to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and reference can be made to the description in the method part for related parts.

[0089] The foregoing description of the disclosed embodiments enables those skilled in the art to practice or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A water quality prediction system based on carbon-water cycle in karst areas, characterized in that: include: A data collection module, a parameter screening module, a main control module, a carbon cycle water quality assessment module, a water cycle water quality assessment module and a comprehensive assessment module. The data collection module collects carbon cycle parameters and water cycle parameters within a fixed time period respectively, performs key parameter screening on the collected parameters through the parameter screening module, and sends the screened carbon cycle key parameters and water cycle key parameters to the carbon cycle water quality assessment module and the water cycle water quality assessment module respectively through the main control module, and calculates and obtains the carbon cycle water quality assessment value and the water cycle water quality assessment value respectively. The comprehensive assessment module obtains the final water quality assessment value according to the carbon cycle water quality assessment value and the water cycle water quality assessment value.

2. A water quality prediction system based on carbon-water cycle in karst areas according to claim 1, characterized in that: The parameter screening module uses the ReliefF algorithm to perform regression analysis on multiple parameter continuous values, and obtains parameter types with greater correlation with water quality through feature weight screening, as follows: The difference calculation module calculates the parameter difference of the neighboring sample features, D i , D j is the collected value of the vth parameter at time i and j, S v is the standardized unit of the parameter, and the parameter difference diff(D i,v ,D j,v )=(D i,v ,D j,v ) / S v ; The sampling module determines the total data sample set D and feature set M, and initializes them according to the sampling number m and the selected k nearest neighbor samples; randomly selects sample R from the total data sample set D, and finds k nearest neighbor samples H from the same sample set of sample R j , j = 1, 2, 3...k, and then find k nearest neighbor samples M from different class sample sets of sample R j (C), repeated sampling m times; Parameter matrix calculation module, calculates the weight matrix of each parameter: Among them, p(C) represents the probability of category C appearing; The screening module calculates the weight matrix for each parameter in the feature set M and sorts them, sets the feature weight threshold δ=avg(W), and removes the parameters with weights lower than the feature weight threshold as key parameters.

3. The water quality prediction system based on carbon-water cycle in karst areas according to claim 1 is characterized in that: The carbon cycle water quality assessment module includes data dimension reduction and feature extraction based on the principal component analysis method for the screened parameters, and calculating the carbon cycle water quality assessment value, specifically: Centralized processing module, after screening, the parameter set corresponding to the carbon cycle is F = (f1, f2, ..., f n ) T ∈R t×n , f i is the time series data of the i-th parameter. First, the parameter set F is centrally processed to return the mean to 0 and the variance to 1, and the decentralized matrix F is obtained. * ; Projection calculation module, calculates the eigenvalue λ of the corresponding covariance matrix n×n i and eigenvector ξ, and eigenvalue λ i Sort from large to small and calculate the cumulative impact rate of different number of characteristic parameters Select the eigenvectors corresponding to the first P eigenvalues ​​exceeding the influence threshold to form a projection matrix Influence index calculation module, multiplying the processed sample matrix with the projection matrix Get the principal component data set X after dimensionality reduction = (x1, x2, ..., x p ) T , X is defined as the principal component element of the parameter set F, x1, x2, ... are called the first principal component, the second principal component, ...; the principal component parameters are normalized using deviation standardization, and the calculation method is x max 、x min is the maximum and minimum value of the corresponding parameter sequence, which is fused into the carbon cycle water quality impact index CQI through weight fusion. The calculation formula is: ρ j is the weight value of P parameters after dimensionality reduction; The carbon cycle water quality assessment value calculation module predicts the water quality status through the pre-trained HMM model combined with the carbon cycle water quality impact index, obtains the water quality grade, and calculates the likelihood value through the Viterbi algorithm according to the water quality grade, obtains the difference between multiple carbon cycle water quality impact indexes that are nearest to the likelihood value, and uses the weighted average method to predict the difference between the indexes to obtain the carbon cycle water quality assessment value.

4. A water quality prediction system based on carbon-water cycle in karst areas according to claim 3, characterized in that: The carbon cycle water quality assessment value calculation module specifically includes: The water quality state prediction module constructs the HMM model, takes the carbon cycle water quality impact index CQI as the training set for model training, so as to evaluate the water quality state corresponding to the current index, and divides the water quality state into five levels: excellent, good, medium, poor, and bad, corresponding to the five hidden states of the HMM model, and takes CQI as the observation sequence O = {O1, O2, ..., O t }, estimate the parameter values ​​of the model HMM: λ=(A,B,π) to maximize the probability of the observation sequence P(O|λ), and define the forward probability variable α based on the forward-backward algorithm t (j), backward probability variable β t (j) and probability variable γ t (j); α t (j)=P(O1,O2,…,O t ,i t =q j |λ),β t (j)=P(O t+1 ,O t+2 ,…,O T |i t =q j ,λ); γ t (j)=P(i t =q j |O,λ),t=1,2,…,T j=1,2,…,N; Substitute the above parameters into the expectation maximization algorithm to calculate P(O|λ), and obtain the probability relationship of the water body under different states, calculate the state expectation value, obtain the parameter re-estimation formula of HMM, and continuously iterate and update the model until Get the optimal model κ is a pre-set convergence condition, which is obtained by the optimal model Assess water quality status; The evaluation value calculation module calculates the likelihood value through the Viterbi algorithm, obtains the difference between multiple carbon cycle water quality impact indexes that are closest to the likelihood value, and uses the weighted average method to predict the difference between the indexes to obtain the carbon cycle water quality evaluation value: Among them, CQI i is the index CQI at time t t The i-th similar historical index, W i CQI i+1 -CQI i The proportion, N represents the weighted number of times, LL t represents the exponential likelihood value at time t, LL i Indicates that LL t The i-th similar historical exponential likelihood value.

5. The water quality prediction system based on carbon-water cycle in karst areas according to claim 1 is characterized in that: The water cycle water quality assessment module specifically includes: A coding table acquisition unit, which acquires corresponding parameter classifications according to the key water cycle parameters to form parameter indicators, determines the evaluation quantities of each water cycle parameter according to the parameter indicators, constructs a multi-parameter quantity evaluation coding table, and obtains the degree of degradation of each parameter; The standard acquisition unit obtains the deduction value of the single parameter index according to the weight value of each parameter evaluation quantity and the degree of degradation combined with the corresponding standard rules, obtains a clear parameter evaluation standard, and formulates a multi-parameter index rule table; The overall assessment unit then formulates a water cycle water quality assessment rule table based on the deduction of each parameter, conducts a status assessment on the karst area, and obtains the water cycle water quality assessment value.

6. A water quality prediction system based on carbon-water cycle in karst areas according to claim 5, characterized in that: The overall assessment unit specifically includes: Among them, E s is the water quality assessment value of the water cycle, Y represents the total deduction value of each key parameter of the water cycle, Y max Represents the total deduction value of all parameters of the key parameters of the water cycle, N c represents the evaluation score corresponding to the fuzzy evaluation set, W s Represents the weight coefficient of overall water quality assessment under different water cycle conditions.

7. The water quality prediction system based on carbon-water cycle in karst areas according to claim 1 is characterized in that: The comprehensive assessment module calculates the final water quality assessment value according to the water cycle water quality assessment value, the carbon cycle water quality assessment value and the corresponding weights through the weighted summation method, and performs a grade assessment according to a preset water quality scoring standard.

Images