An online multi-parameter detection method for water quality in offshore culture area

Abstract

The application discloses a kind of offshore aquaculture area water quality on-line multi-parameter detection method, it is related to water quality monitoring technical field, comprising: the water quality parameter in real-time offshore aquaculture area is associated with the time-varying vector of the water quality requirement of each growth stage of fry species in offshore aquaculture area is evaluated, whether real-time water quality parameter satisfies the water quality requirement of fry species is judged;Water quality data of the same fry species in historical aquaculture area are acquired to carry out water quality state analysis, evaluate the time-varying vector of the water quality requirement of fry species in offshore aquaculture area with the historical water quality trend index in offshore aquaculture area, establish fry growth hidden danger evaluation model, the water quality parameter in real-time offshore aquaculture area is fitted with the growth hidden danger risk type of fry species in offshore aquaculture area, and the development direction of fry species in offshore aquaculture area is determined.The advantages of the present application are: improve the accuracy and timeliness of offshore aquaculture area water quality management.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of water quality monitoring, in particular to a kind of offshore aquaculture area water quality online multi-parameter detection method. BACKGROUND

[0002] The existing offshore aquaculture water quality detection scheme uses static threshold judgment and cannot adapt to the dynamic water quality demand of fry growth, can only carry out isolated analysis of single parameter and ignores the synergistic effect of multiple parameters, relies on empirical weight distribution, leads to strong subjectivity of risk assessment, and lacks differentiated analysis of natural factors and human influence, it is difficult to capture the time sequence variation characteristics of water quality parameters in time, cannot realize early warning of hidden danger and predict risk evolution trend, leading to response lag in aquaculture water quality management. SUMMARY

[0003] To solve the above technical problems, an offshore aquaculture area water quality online multi-parameter detection method is provided, which solves the above problems.

[0004] To achieve the above purposes, the technical scheme adopted by the present application is:

[0005] An offshore aquaculture area water quality online multi-parameter detection method, comprising:

[0006] Step one, based on the fry species in the historical offshore aquaculture area, analyze the water quality preference of the fry species in each growth stage in the offshore aquaculture area, and generate the water quality demand time-varying vector of the fry species in each growth stage in the offshore aquaculture area;

[0007] Step two, based on the edge water quality sensor, the real-time water quality parameters in the offshore aquaculture area and the water quality demand time-varying vector of the fry species in each growth stage in the offshore aquaculture area are associated and evaluated, to determine whether the real-time water quality parameters meet the water quality demand of the fry species, if yes, it is determined that the water quality is normal, if not, it is determined that the water quality is abnormal;

[0008] Step three, if it is determined that the water quality is normal, based on the fry species in the offshore aquaculture area to be detected, the water quality data of the historical aquaculture area with the same fry species are obtained for water quality state analysis, and the historical water quality trend index of the offshore aquaculture area to be detected is evaluated;

[0009] Step four, according to the historical water quality trend index of the offshore aquaculture area to be detected and the water quality demand time-varying vector of the fry species in the offshore aquaculture area, a fry growth hidden danger evaluation model is established, and the fry species growth hidden danger risk type in the offshore aquaculture area to be detected is evaluated;

[0010] Step five, fitting analysis is performed on the water quality parameters in the real-time offshore aquaculture area and the growth risk type of the fish fry species in the offshore aquaculture area to be detected, so as to determine the growth risk development direction of the fish fry species in the real-time offshore aquaculture area.

[0011] Preferably, a standardized aquaculture database of the fish fry species in the known offshore aquaculture area is determined.

[0012] Based on the standardized aquaculture database of the fish fry species in the known offshore aquaculture area, the growth stage of the fish fry species is taken as an observation window, and the standardized aquaculture parameters of the fish fry species are taken as observation objects, so as to establish an ideal water quality parameter time sequence matrix of the growth stage of the fish fry species in the known offshore aquaculture area.

[0013] The mean and standard deviation of the ideal water quality parameters in the ideal water quality parameter time sequence matrix of the growth stage of the fish fry species in the known offshore aquaculture area are calculated, and the upper and lower limit value intervals of the water quality parameters of the growth stage of the fish fry species in the known offshore aquaculture area are statistically calculated by using the percentile.

[0014] Based on fuzzy clustering, the growth speed of the growth stage of the fish fry species in the known offshore aquaculture area is taken as an initial cluster, and the upper and lower limit value intervals of the water quality parameters of the growth stage of the fish fry species in the known offshore aquaculture area are clustered and divided to obtain the upper limit value interval of the growth speed attention water quality parameter of the growth stage of the fish fry species in the known offshore aquaculture area.

[0015] Preferably, the upper limit value interval of the growth speed attention water quality parameter of the growth stage of the fish fry species in the known offshore aquaculture area is supplemented by using piecewise linear interpolation, and the upper and lower limit preference time sequence of the growth speed attention water quality parameter of the growth stage of the fish fry species in the known offshore aquaculture area is established.

[0016] The change distribution between the growth speed of the growth stage of the fish fry species per unit time and the attention water quality parameter in the upper and lower limit preference time sequence of the growth speed attention water quality parameter of the growth stage of the fish fry species in the known offshore aquaculture area is fitted by using Gaussian process regression, so as to determine the time-varying water quality preference vector of the growth stage of the fish fry species in the known offshore aquaculture area.

[0017] An AR time sequence autoregression is established to establish a fish fry species growth stage water quality preference model.

[0018] The standardized aquaculture data of the fish fry species in the known offshore aquaculture area are taken as feature data, and the time-varying water quality preference vector of the growth stage of the fish fry species in the known offshore aquaculture area is taken as label data, which are substituted into the fish fry species growth stage water quality preference model to generate a water quality demand time-varying vector of each growth stage of the fish fry species in the offshore aquaculture area.

[0019]

[0020] wherein V ij (t) is the kth water quality demand time-varying vector of the jth growth stage of the ith fish fry species in the offshore aquaculture area at the tth unit time, x ijk (t-i) is the kth water quality standardized aquaculture data of the jth growth stage of the ith fish fry species in the offshore aquaculture area at the t-i unit time, ∈ t is the error term at the tth unit time, a0 is the intercept, a i is the linear regression coefficient, and n is the total number of growth stages.

[0021] Preferably, the water quality parameters in the real-time offshore aquaculture area are vector-converted by using linear mapping to establish a water quality parameter vector time series in the real-time offshore aquaculture area;

[0022] According to the unit time stamp, the water quality parameter vector time series in the real-time offshore aquaculture area and the water quality demand time-varying vector of each growth stage of the fish fry species in the offshore aquaculture area are aligned;

[0023] The spatial distance between the water quality parameter vector time series in the real-time offshore aquaculture area and the water quality demand time-varying vector of each growth stage of the fish fry species in the offshore aquaculture area is calculated by using the Euclidean distance formula, and the water quality parameter vector deviation value in the real-time offshore aquaculture area is screened out;

[0024] The deviation degree of each water quality parameter vector deviation value in the real-time offshore aquaculture area and the water quality demand time-varying vector of each growth stage of the fish fry species in the offshore aquaculture area is calculated, and the water quality parameter vector deviation value in the real-time offshore aquaculture area is given a weight;

[0025] Based on the water quality parameter vector deviation value in the real-time offshore aquaculture area and the water quality parameter vector deviation value weight in the real-time offshore aquaculture area, the water quality overall deviation index in the real-time offshore aquaculture area at the unit time is calculated in the following manner:

[0026]

[0027] wherein D(t) is the water quality overall deviation index in the real-time offshore aquaculture area at the tth unit time, w k is the water quality parameter vector deviation value weight in the real-time offshore aquaculture area at the tth unit time, r k (t) is the actual value of the water quality parameter vector in the real-time offshore aquaculture area at the tth unit time, V ijk (t) is the kth water quality demand time-varying vector of the jth growth stage of the ith fish fry species in the offshore aquaculture area at the tth unit time, V an upper limit allowable value of a time-varying vector of water quality demand of a jth growth stage of an ith fish fry species in a coastal aquaculture area at a tth unit time, a lower limit allowable value of a time-varying vector of water quality demand of a jth growth stage of an ith fish fry species in a coastal aquaculture area at a tth unit time;

[0028] presetting a water quality demand threshold of each growth stage of the fish fry species in the coastal aquaculture area according to the percentile of the time-varying vector of water quality demand of each growth stage of the fish fry species in the coastal aquaculture area;

[0029] judging whether the overall deviation index of water quality in the real-time coastal aquaculture area at the unit time is less than the water quality demand threshold of each growth stage of the fish fry species in the coastal aquaculture area, if yes, determining that the water quality is normal, and if no, determining that the water quality is abnormal.

[0030] Preferably, based on the fish fry species in the to-be-detected coastal aquaculture area, the water quality data of the historical aquaculture area with the same fish fry species is obtained;

[0031] According to the natural factors and human factors affecting the water quality change of the aquaculture area, the water quality data of the historical aquaculture area with the same fish fry species is divided to form a water quality influence natural factor data-human factor multivariate time series matrix of the historical aquaculture area with the same fish fry species;

[0032] Based on the Random Forest, according to the water quality influence natural factor data-human factor data of the historical aquaculture area with the same fish fry species at the unit time, a series water quality influence natural-human factor decision tree is established, the water quality parameter is taken as a branch decision node of the series water quality influence natural-human factor decision tree, the water quality influence natural factor data-human factor multivariate time series matrix of the historical aquaculture area with the same fish fry species is taken as an output, the information gain of each influence natural and human factor to the branch decision node is calculated, and the contribution proportion of the natural factor and the human factor to the water quality change in the water quality data of the historical aquaculture area with the same fish fry species is generated.

[0033] Preferably, the contribution proportion of the natural factor and the human factor to the water quality change in the water quality data of the historical aquaculture area with the same fish fry species is subjected to distribution regression fitting, and the individual influence coefficient of the natural factor and the human factor to the water quality index is calculated;

[0034] According to the contribution proportion of the natural factor and the human factor to the water quality change in the water quality data of the historical aquaculture area with the same fish fry species, the natural factor and the human factor in the water quality data of the historical aquaculture area with the same fish fry species are given weights

[0035] The local slope of the natural-anthropological factor ratio, corresponding to the individual influence coefficients of natural and anthropological factors on water quality indicators, is fitted using the least squares method, along with the weights of natural and anthropological factors in historical aquaculture data for the same fish fry species. The historical water quality trend indicators for the nearshore aquaculture area to be tested are then calculated as follows:

[0036]

[0037] Among them, P Total (t) Historical water quality trend indicators of the nearshore aquaculture area to be tested at the t-th unit of time. This represents the influence coefficient of natural factors in water quality data for the same fish fry species in historical aquaculture areas. δ represents the human factor influence coefficient in the historical water quality data of the same fish fry species in the aquaculture area at the t-th unit time. Nat W represents the slope of the trend of the influence of natural factors on water quality. Nat (t) represents the weight of natural factors in the historical water quality data of the same fish fry species in the aquaculture area at the t-th unit of time, W Hum (t) represents the weight of human factors in the historical water quality data of the same fish fry species in the aquaculture area at the t-th unit time, δ Hum C represents the slope of the trend of human factors' influence on water quality. Nat C is the intercept for natural factors. Hum For human factors intercept, ∈ t η t All of these are random error terms.

[0038] Preferably, based on the known types of growth hazards of fish fry species at various growth stages in known nearshore aquaculture areas, the water quality inducing parameters corresponding to the types of growth hazards of fish fry species at various growth stages in known nearshore aquaculture areas are determined.

[0039] Using correlation coefficients, the degree of interaction between water quality inducing parameters corresponding to the types of growth hazards of fish fry species at various growth stages in known nearshore aquaculture areas is calculated, and a synergistic effect matrix of water quality inducing parameters corresponding to the types of growth hazards of fish fry species at various growth stages in known nearshore aquaculture areas is established.

[0040] Standardization was performed based on the synergistic effect matrix of water quality inducing parameters corresponding to the types of growth hazards of fish fry at various growth stages in known nearshore aquaculture areas.

[0041] Based on the SVM support vector machine, the binary classification kernel function corresponding to the growth hidden danger type of each growth stage of the fry species is trained, the water quality inducing parameter synergistic effect matrix corresponding to the growth hidden danger type of each growth stage of the fry species in the known near-sea breeding area is taken as the output, the growth hidden danger type decision hyperplane boundary of each growth stage of the fry species in the known near-sea breeding area is generated, and the growth hidden danger type decision feature vector of each growth stage of the fry species in the known near-sea breeding area is generated.

[0042] The growth hidden danger type decision vector of each growth stage of the fry species in the known near-sea breeding area is substituted into the sigmoid function, the fry species growth hidden danger risk type risk probability is generated, and the fry growth hidden danger evaluation model is obtained.

[0043] The historical water quality trend index in the to-be-detected near-sea breeding area and the fry species water quality demand time-varying vector in the near-sea breeding area are substituted into the SVM support vector machine, and the fry species growth hidden danger risk type risk probability in the to-be-detected near-sea breeding area is obtained.

[0044] Preferably, based on the water quality parameters in the real-time near-sea breeding area per unit time, the water quality parameter time sequence feature data in the real-time near-sea breeding area is counted.

[0045] Based on the cosine similarity, the similarity between the growth hidden danger type attention water quality parameter of each growth stage of the fry species in the known near-sea breeding area and the water quality parameter time sequence feature data in the real-time near-sea breeding area is calculated, divided and screened, and the fry species growth hidden danger risk type support water quality parameter data set in the real-time near-sea breeding area is obtained.

[0046] Based on the fry species growth hidden danger risk type support water quality parameter data set in the real-time near-sea breeding area, the SVM support vector machine is substituted, and the real-time near-sea breeding fry species growth hidden danger risk type set is determined.

[0047] By using Bayesian inference, the support evidence number of the real-time near-sea breeding fry species growth hidden danger risk type set for each hidden danger risk type according to the fry species growth hidden danger risk type support water quality parameter data set in the real-time near-sea breeding area is verified, the real-time near-sea breeding fry species growth hidden danger risk type development trend confidence is obtained, and the growth hidden danger risk development direction of the fry species in the real-time near-sea breeding area is determined.

[0048] Compared with the prior art, the beneficial effects of the present application are that:

[0049] The application provides an offshore aquaculture area water quality online multi-parameter detection method scheme, which realizes accurate adaptation through the construction of a water quality demand time-varying vector in the fry growth stage, combines real-time sensor data and historical trend analysis, adopts a machine learning algorithm to establish a risk prediction model, realizes real-time evaluation of water quality abnormalities, predicts the development trend of hidden dangers, and significantly improves the accuracy and timeliness of offshore aquaculture area water quality management, and provides intelligent protection for the whole process of fry healthy growth in offshore aquaculture areas. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 It is an offshore aquaculture area water quality online multi-parameter detection method flow chart; DETAILED DESCRIPTION

[0051] The following description is used to disclose the application so that those skilled in the art can implement the application. The preferred embodiments in the following description are only as examples, and other obvious modifications can be thought of by those skilled in the art.

[0052] REFERENCE Figure 1 As shown in the figure, an offshore aquaculture area water quality online multi-parameter detection method comprises:

[0053] Step one, based on the fish fry species in the historical offshore aquaculture area, analyze the water quality preference of the fish fry species in each growth stage in the offshore aquaculture area, and generate the water quality demand time-varying vector of the fish fry species in each growth stage in the offshore aquaculture area;

[0054] The step one comprises the following contents:

[0055] Determine the standardized aquaculture database of the fish fry species in the known offshore aquaculture area;

[0056] Based on the standardized aquaculture database of the fish fry species in the known offshore aquaculture area, taking the growth stage of the fish fry species as the observation window and the standardized aquaculture parameters of the fish fry species as the observation object, an ideal water quality parameter time sequence matrix of the fish fry species growth stage in the known offshore aquaculture area is established;

[0057] Calculate the ideal water quality parameter mean and standard deviation in the ideal water quality parameter time sequence matrix of the fish fry species growth stage in the known offshore aquaculture area, and use the percentile to statistically analyze the upper and lower limit value interval of the water quality parameter of the fish fry species growth stage in the known offshore aquaculture area;

[0058] Based on fuzzy clustering, taking the growth speed of the fish fry species growth stage in the known offshore aquaculture area as the initial cluster, the water quality parameter upper and lower limit value interval of the fish fry species growth stage in the known offshore aquaculture area is clustered and divided, and the growth speed attention water quality parameter upper limit value interval of the fish fry species growth stage in the known offshore aquaculture area is obtained;

[0059] The growth speed of the fish fry species in the known offshore aquaculture area is supplemented by using piecewise linear interpolation, and the upper limit value interval of the water quality parameter is adjacent to the stage of the growth speed of the fish fry species in the known offshore aquaculture area. The upper and lower limit preferred time sequence of the water quality parameter is established for the growth speed of the fish fry species in the known offshore aquaculture area;

[0060] The change distribution between the growth speed of the fish fry species in the known offshore aquaculture area and the water quality parameter per unit time in the upper and lower limit preferred time sequence of the water quality parameter is fitted by using Gaussian process regression, and the time-varying water quality preference vector of the fish fry species in the known offshore aquaculture area is determined.

[0061] According to the AR time series autoregression, the fish fry species growth stage water quality preference model is established.

[0062] The standardized aquaculture data of the fish fry species in the known offshore aquaculture area is taken as the feature data, and the time-varying water quality preference vector of the fish fry species growth stage in the known offshore aquaculture area is taken as the label data, which is substituted into the fish fry species growth stage water quality preference model to generate the water quality demand time-varying vector of each growth stage of the fish fry species in the offshore aquaculture area.

[0063]

[0064] Wherein, V ij (t) is the kth water quality demand time-varying vector of the jth growth stage of the ith fish fry species in the offshore aquaculture area per t unit time, x ijk (t-i) is the kth water quality standardized aquaculture data of the jth growth stage of the ith fish fry species in the offshore aquaculture area per t-i unit time, ∈ t is the error term per t unit time, α0 is the intercept, α i is the linear regression coefficient, and n is the total number of growth stages.

[0065] In use, the contents in the above steps are combined,

[0066] As a further content, the existing offshore aquaculture area water quality detection mainly uses fixed threshold detection, which cannot reflect the dynamic water quality demand of fish fry in different growth stages, resulting in poor aquaculture adaptability. Secondly, the water quality interval division depends on the experience threshold, and does not combine the dynamic clustering of growth speed, resulting in mismatch between the water quality limit value interval and the actual demand.

[0067] The scheme dynamically divides the upper and lower limit intervals of the water quality parameters of each growth stage of the fry by constructing an ideal water quality parameter time sequence matrix and a time-varying preference vector, combining fuzzy clustering and growth rate correlation analysis; the time-varying influence of water quality on the growth rate is predicted by using a Gaussian process regression and an AR autoregressive model, and the missing data is compensated by piecewise linear interpolation, and finally a time-varying vector of water quality demand is generated based on standardized aquaculture data, realizing the upgrade from static threshold detection to dynamic demand prediction, and significantly improving the accuracy and adaptability of offshore aquaculture water quality management.

[0068] Step two, based on the edge water quality sensor, the water quality parameters in the real-time offshore aquaculture area and the time-varying vector of water quality demand of each growth stage of the fry in the offshore aquaculture area are associated and evaluated to determine whether the real-time water quality parameters meet the water quality demand of the fry; if yes, the water quality is determined to be normal, and if no, the water quality is determined to be abnormal;

[0069] The step two includes the following contents:

[0070] The water quality parameters in the real-time offshore aquaculture area are converted into vectors by using linear mapping, and a time sequence of the water quality parameter vectors in the real-time offshore aquaculture area is established;

[0071] According to the unit timestamp, the time sequence of the water quality parameter vectors in the real-time offshore aquaculture area and the time-varying vector of water quality demand of each growth stage of the fry in the offshore aquaculture area are aligned;

[0072] The spatial distance between the time sequence of the water quality parameter vectors in the real-time offshore aquaculture area and the time-varying vector of water quality demand of each growth stage of the fry in the offshore aquaculture area is calculated by using the Euclidean distance formula, and the water quality parameter vector deviation value in the real-time offshore aquaculture area is selected;

[0073] The deviation degree of each water quality parameter vector deviation value in the real-time offshore aquaculture area and the time-varying vector of water quality demand of each growth stage of the fry in the offshore aquaculture area is calculated, and the water quality parameter vector deviation value in the real-time offshore aquaculture area is given a weight;

[0074] Based on the water quality parameter vector deviation value in the real-time offshore aquaculture area and the water quality parameter vector deviation value weight in the real-time offshore aquaculture area, the overall deviation index of the water quality in the real-time offshore aquaculture area under the unit time is calculated, and the method is as follows:

[0075]

[0076] Wherein, D(t) is the overall deviation index of the water quality in the real-time offshore aquaculture area under the tth unit time, w k is the water quality parameter vector deviation value weight in the real-time offshore aquaculture area under the tth unit time, r k(t) is the actual value of the water quality parameter vector in the real-time offshore aquaculture area in the tth unit time, V ijk (t) is the optimal value of the kth water quality demand time-varying vector of the jth growth stage of the ith fry species in the offshore aquaculture area in the tth unit time, is the upper limit allowed value of the kth water quality demand time-varying vector of the jth growth stage of the ith fry species in the offshore aquaculture area in the tth unit time, is the lower limit allowed value of the kth water quality demand time-varying vector of the jth growth stage of the ith fry species in the offshore aquaculture area in the tth unit time;

[0077] Based on the water quality demand time-varying vector of each growth stage of the fry species in the offshore aquaculture area, the water quality demand threshold of each growth stage of the fry species in the offshore aquaculture area is preset according to the percentile;

[0078] As a further content, the water quality demand threshold of each growth stage is based on the percentile, and the percentile is 95% or 90% of the water quality demand time-varying vector of each growth stage of the fry species, and depends on the range set by the implementer for the growth quality of the final fry species, which is not described in detail here.

[0079] It is judged whether the overall water quality deviation index in the real-time offshore aquaculture area in the unit time is less than the water quality demand threshold of each growth stage of the fry species in the offshore aquaculture area. If yes, it is determined that the water quality is normal, and if no, it is determined that the water quality is abnormal;

[0080] In use, the contents in the above steps are combined,

[0081] As a further content, the traditional water quality detection method has defects such as static threshold judgment, single-point data isolated analysis, subjective weight distribution, and lack of real-time performance. It only relies on fixed standard evaluation and cannot dynamically adapt to the growth demand of fry, uses local measurement and ignores time trend, sets weight by experience and leads to evaluation deviation, and offline analysis mode is difficult to respond to abnormality in time, which seriously affects the accuracy and timeliness of aquaculture water quality management.

[0082] The scheme realizes accurate matching of fry growth stage and real-time water quality through dynamic data alignment of water quality parameter vector time series and demand time-varying vector; objectively quantifies the degree of abnormality by using Euclidean distance and dynamic weight to calculate parameter deviation; effectively captures water quality time characteristics based on time stamp continuous monitoring to avoid single-point detection deviation; finally realizes rapid abnormality early warning through real-time deviation index calculation, which significantly improves the accuracy and timeliness of aquaculture water quality monitoring.

[0083] Step three, if the water quality is determined to be normal, based on the type of fry in the offshore aquaculture area to be detected, the water quality data of the historical aquaculture area with the same type of fry is obtained to analyze the water quality state, and the historical water quality trend index in the offshore aquaculture area to be detected is evaluated;

[0084] The step three includes the following contents:

[0085] Based on the type of fry in the offshore aquaculture area to be detected, the water quality data of the historical aquaculture area with the same type of fry is obtained;

[0086] According to the natural factors and human factors affecting the water quality change of the aquaculture area, the water quality data of the historical aquaculture area with the same type of fry is divided to form a water quality influence natural factor data-human factor multivariate time series matrix of the historical aquaculture area with the same type of fry;

[0087] Based on the Random Forest random forest, according to the water quality influence natural factor data-human factor data of the historical aquaculture area with the same type of fry in unit time, a series of water quality influence natural-human factor decision tree is established, the water quality parameter is taken as the branch decision node of the series of water quality influence natural-human factor decision tree, the water quality influence natural factor data-human factor multivariate time series matrix of the historical aquaculture area with the same type of fry is taken as the output, the information gain of each influence natural and human factor to the branch decision node is calculated, and the contribution proportion of the natural factor and the human factor to the water quality change in the water quality data of the historical aquaculture area with the same type of fry is generated.

[0088] Based on the contribution proportion of the natural factor and the human factor to the water quality change in the water quality data of the historical aquaculture area with the same type of fry, the distribution regression fitting is carried out, and the individual influence coefficient of the natural factor and the human factor on the water quality index is calculated.

[0089] According to the contribution proportion of the natural factor and the human factor to the water quality change in the water quality data of the historical aquaculture area with the same type of fry, the natural factor and the human factor in the water quality data of the historical aquaculture area with the same type of fry are given weights

[0090] The local slope of the natural factor-human factor corresponding to the individual influence coefficient of the natural factor and the human factor on the water quality index is fitted by using the least square method, and the historical water quality trend index in the offshore aquaculture area to be detected is calculated based on the weight of the natural factor and the human factor in the water quality data of the historical aquaculture area with the same type of fry, in the following manner:

[0091]

[0092] Wherein, P Total (t) the historical water quality trend index in the offshore aquaculture area to be detected in the tth unit time, A natural factor influence coefficient in water quality data of the historical breeding area and the same fish fry species, A human factor influence coefficient in water quality data of the historical breeding area and the same fish fry species at the t-th unit time, δ Nat A natural factor influence trend slope on water quality, W Nat A natural factor component weight in water quality data of the historical breeding area and the same fish fry species at the t-th unit time, W Hum A human factor component weight in water quality data of the historical breeding area and the same fish fry species at the t-th unit time, δ Hum A human factor influence trend slope on water quality, C Nat A natural factor intercept, C Hum A human factor intercept, ∈ t , η t are all random error terms.

[0093] In use, the contents in the above steps are combined,

[0094] As further contents, the traditional offshore aquaculture water quality detection only monitors water quality parameters without distinguishing the differentiated influences of natural factors (such as tides, temperature) and human factors (such as feeding, medication); a fixed proportion static weight distribution is adopted, which cannot adapt to the dynamic changes of factor contribution degrees in each stage of aquaculture; there is a lack of causal modeling between water quality parameters and influencing factors, making it difficult to quantify specific influence paths; and a simple linear regression analysis is relied on, which cannot effectively capture the nonlinear characteristics of natural-human factor interactions, resulting in insufficient trend prediction accuracy.

[0095] The present scheme realizes accurate attribution analysis of the influences of factors such as tides and feeding on water quality parameters by constructing a natural-human factor multivariate time series matrix and adopting a random forest algorithm to establish a decision tree model; a dynamic weight system is constructed based on information gain and least squares method to evaluate the adaptive optimization of indicators with the breeding stage; the water quality change causal chain is analyzed through the decision tree, and the nonlinear interaction between factors is fitted by using distribution regression, thereby improving the water quality trend prediction accuracy in complex scenarios.

[0096] Step four, according to the historical water quality trend index in the to-be-detected offshore breeding area and the time-varying vector of water quality demand of the fish fry species in the offshore breeding area, a fish fry growth hidden danger evaluation model is established to evaluate the growth hidden danger risk type of the fish fry species in the to-be-detected offshore breeding area;

[0097] The step four includes the following contents:

[0098] Based on the growth hidden danger types of the fish fry species at each growth stage in the known offshore breeding area, water quality inducement parameters corresponding to the growth hidden danger types of the fish fry species at each growth stage in the known offshore breeding area are determined.

[0099] Using the correlation coefficient, the interaction degree between the water quality inducing parameters corresponding to the growth hazard type of the fish fry species at each growth stage in the known offshore aquaculture area is calculated, and a water quality inducing parameter synergy effect matrix corresponding to the growth hazard type of the fish fry species at each growth stage in the known offshore aquaculture area is established.

[0100] Based on the water quality inducing parameter synergy effect matrix corresponding to the growth hazard type of the fish fry species at each growth stage in the known offshore aquaculture area, standardization processing is performed.

[0101] Based on the SVM support vector machine, a binary classification kernel function corresponding to the growth hazard type of the fish fry species at each growth stage is trained, the water quality inducing parameter synergy effect matrix corresponding to the growth hazard type of the fish fry species at each growth stage in the known offshore aquaculture area is taken as the output, a growth hazard type decision hyperplane boundary of the fish fry species at each growth stage in the known offshore aquaculture area is generated, and a growth hazard type decision feature vector of the fish fry species at each growth stage in the known offshore aquaculture area is generated.

[0102] The growth hazard type decision vector of the fish fry species at each growth stage in the known offshore aquaculture area is substituted into the sigmoid function to generate a fish fry growth hazard risk type risk probability, and a fish fry growth hazard evaluation model is obtained.

[0103] The historical water quality trend indicators in the to-be-detected offshore aquaculture area and the fish fry species water quality demand time-varying vector in the offshore aquaculture area are substituted into the SVM support vector machine to obtain a fish fry species growth hazard risk type risk probability in the to-be-detected offshore aquaculture area.

[0104] In use, the contents in the above steps are combined,

[0105] As further contents, the traditional water quality detection technology relies on the bank hazard identification to observe the abnormal fish fry and make lagging judgment, and cannot realize early warning; only focuses on a single parameter exceeding the standard and ignores the synergistic effect of multiple parameters (such as the interaction between pH and dissolved oxygen); adopts a three-level extensive risk classification of high / medium / low, lacks accurate probabilistic evaluation; and uses a fixed threshold as the determination standard, which is difficult to adapt to the dynamic risk characteristics of different breeding species and growth stages, resulting in insufficient accuracy and timeliness of risk assessment.

[0106] The scheme improves the timeliness of early warning of offshore aquaculture water quality hazards by constructing a water quality parameter synergistic effect matrix and analyzing the interaction of water quality parameters; a dynamic probability model is constructed by combining SVM support vector machine and sigmoid function to improve the risk assessment accuracy; through training of growth stage specific classifiers, the differentiated needs of different aquaculture species are automatically adapted; at the same time, based on the decision feature vector, the risk traceability is realized, and the composite hazards such as sudden change of water temperature + overfeeding are accurately positioned.

[0107] Step five, fitting analysis is performed on the water quality parameters in the real-time offshore aquaculture area and the fish fry species growth hazard risk type in the to-be-detected offshore aquaculture area, and the fish fry species growth hazard risk development direction in the real-time offshore aquaculture area is determined;

[0108] The step five includes the following contents:

[0109] Based on the water quality parameters in the real-time offshore aquaculture area per unit time, the water quality parameter time sequence characteristic data in the real-time offshore aquaculture area is counted;

[0110] Based on the cosine similarity, the similarity between the growth hazard type attention water quality parameters of each growth stage of the fish fry species in the known offshore aquaculture area and the water quality parameter time sequence characteristic data in the real-time offshore aquaculture area is calculated for division and screening, and the fish fry species growth hazard risk type support water quality parameter data set in the real-time offshore aquaculture area is obtained;

[0111] Based on the fish fry species growth hazard risk type support water quality parameter data set in the real-time offshore aquaculture area, the SVM support vector machine is substituted to determine the real-time offshore aquaculture fish fry species growth hazard risk type set;

[0112] Using Bayesian inference, the support evidence number of the fish fry species growth hazard risk type support water quality parameter data set for each hazard risk type in the real-time offshore aquaculture fish fry species growth hazard risk type set is verified to obtain the real-time offshore aquaculture fish fry species growth hazard risk type development trend confidence, and the fish fry species growth hazard risk development direction in the real-time offshore aquaculture area is determined.

[0113] In use, the contents in the above steps are combined,

[0114] As a further content, the traditional water quality detection can only statically evaluate the current risk state and cannot predict the hidden danger evolution trend (such as the development path of acute poisoning or chronic disease) in the risk prediction aspect; the dynamic similarity matching mechanism between the water quality parameters and the risk types is lacked, which leads to the inaccurate judgment of the risk development path; meanwhile, the risk judgment lacks the probabilistic support, and it is difficult to quantify the credibility of specific threats (such as the certainty of suffocation risk caused by the decrease of dissolved oxygen); and it is also unable to effectively handle the multi-risk concurrent scene (such as the simultaneous occurrence of algal explosion and ammonia nitrogen exceeding standard), which leads to the difficulty in distinguishing the primary and secondary risks, and seriously affects the accuracy and timeliness of the prevention and control decision.

[0115] The scheme dynamically matches the water quality parameters and the risk characteristics through the cosine similarity algorithm, constructs a risk prediction engine by combining the Bayesian inference, improves the prediction accuracy of hidden dangers, constructs a multi-dimensional evidence chain matrix based on SVM, synchronously tracks the correlation of multiple parameters such as dissolved oxygen and algae, quantifies the confidence of the risk development path by using the Bayesian posterior probability, and visually displays the complete risk transmission chain such as the overfeeding leading to the increase of ammonia nitrogen leading to the gill disease, so as to provide the intelligent decision of the whole process from prediction and early warning to tracing and regulation for the aquaculture risk prevention and control.

[0116] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above-mentioned embodiments, and the above-mentioned embodiments and descriptions in the specification are only the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection claimed by the present application is defined by the appended claims and their equivalents.

Claims

1. A method for online multi-parameter detection of water quality in nearshore aquaculture areas, characterized in that, include: Step 1: Based on the fish fry species in the historical nearshore aquaculture area, analyze the water quality preferences of the nearshore aquaculture area at each growth stage of the fish fry species, and generate a time-varying vector of water quality requirements for each growth stage of the fish fry species in the nearshore aquaculture area. Step 2: Based on the edge water quality sensor, obtain the real-time water quality parameters in the nearshore aquaculture area and the time-varying vector of water quality requirements of fish fry at various growth stages in the nearshore aquaculture area, and evaluate the correlation between them to determine whether the real-time water quality parameters meet the water quality requirements of the fish fry. If yes, determine that the water quality is normal; otherwise, determine that the water quality is abnormal. Step 3: If the water quality is determined to be normal, based on the fish fry species in the nearshore aquaculture area to be tested, obtain historical water quality data of the same fish fry species in the aquaculture area to analyze the water quality status and evaluate the historical water quality trend indicators in the nearshore aquaculture area to be tested. Step 4: Based on the historical water quality trend indicators of the nearshore aquaculture area to be tested and the time-varying vector of water quality requirements for fish fry species in the nearshore aquaculture area, establish a fish fry growth hazard assessment model to assess the risk types of fish fry growth hazards in the nearshore aquaculture area to be tested, including: Based on historically known types of growth hazards for fish fry species at various growth stages in nearshore aquaculture areas, water quality inducing parameters corresponding to these growth hazard types were determined. Using correlation coefficients, the degree of interaction between water quality inducing parameters corresponding to the types of growth hazards of fish fry species at various growth stages in known nearshore aquaculture areas is calculated, and a synergistic effect matrix of water quality inducing parameters corresponding to the types of growth hazards of fish fry species at various growth stages in known nearshore aquaculture areas is established. Standardization was performed based on the synergistic effect matrix of water quality inducing parameters corresponding to the types of growth hazards of fish fry at various growth stages in known nearshore aquaculture areas. Based on SVM support vector machine, a binary classification kernel function corresponding to the growth hazard type of fish fry at each growth stage is trained. The water quality inducing parameter synergy effect matrix corresponding to the growth hazard type of fish fry at each growth stage in the known nearshore aquaculture area is used as the output to generate the decision hyperplane boundary of the growth hazard type of fish fry at each growth stage in the known nearshore aquaculture area, and to generate the decision feature vector of the growth hazard type of fish fry at each growth stage in the known nearshore aquaculture area. Substitute the decision vector of the growth hazard type of fish fry species at each growth stage in the known nearshore aquaculture area into the sigmoid function to generate the risk probability of the growth hazard type of fish fry species, and obtain the fish fry growth hazard assessment model. By substituting the historical water quality trend indicators of the nearshore aquaculture area to be tested and the time-varying vector of water quality requirements of fish fry species in the nearshore aquaculture area into the SVM support vector machine, the risk probability of the risk type of the growth hazard of fish fry species in the nearshore aquaculture area to be tested can be obtained. Step 5: Use the real-time water quality parameters in the nearshore aquaculture area to perform a fitting analysis with the risk types of potential growth hazards of fish fry species in the nearshore aquaculture area to determine the development direction of potential growth hazards of fish fry species in the real-time nearshore aquaculture area.

2. The method for online multi-parameter detection of water quality in nearshore aquaculture areas according to claim 1, characterized in that, Step 1 includes: Establish a standardized aquaculture database for fish fry species in known nearshore aquaculture areas; Based on a standardized aquaculture database of fish fry species in known nearshore aquaculture areas, and using the growth stage of fish fry species as the observation window and the standardized aquaculture parameters of fish fry species as the observation objects, a time series matrix of ideal water quality parameters for the growth stage of fish fry species in known nearshore aquaculture areas is established. Calculate the mean and standard deviation of the ideal water quality parameters in the time series matrix of the ideal water quality parameters for the growth stages of fish fry in the known nearshore aquaculture area, and use percentiles to statistically determine the upper and lower limit ranges of the water quality parameters for the growth stages of fish fry in the known nearshore aquaculture area. Based on fuzzy clustering, the growth rate of fish fry at known growth stages in known nearshore aquaculture areas is used as the initial cluster. Clustering is then performed on the upper and lower limit ranges of water quality parameters at known growth stages in known nearshore aquaculture areas to obtain the upper limit range of water quality parameters related to the growth rate of fish fry at known growth stages in known nearshore aquaculture areas.

3. The method for online multi-parameter detection of water quality in nearshore aquaculture areas according to claim 2, characterized in that, Step 1 further includes: Using piecewise linear interpolation, the upper and lower limit preferences of water quality parameters for the growth rate of fish fry at different growth stages in known nearshore aquaculture areas are supplemented for adjacent stages, thus establishing a time series of preferences for the upper and lower limits of water quality parameters for the growth rate of fish fry at different growth stages in known nearshore aquaculture areas. Gaussian process regression was used to fit the distribution of the change between the growth rate of fish fry at different growth stages and the water quality parameters under the upper and lower limits of the preference time series of the water quality parameters under the known nearshore aquaculture area, and the time-varying water quality preference vector of fish fry at different growth stages under the known nearshore aquaculture area was determined. Based on AR time-series autoregression, a water quality preference model for fish fry growth stages was established. Using standardized aquaculture data of fish fry species in known nearshore aquaculture areas as feature data, and time-varying water quality preference vectors of fish fry species at different growth stages in known nearshore aquaculture areas as label data, we input them into the water quality preference model of fish fry species at different growth stages to generate time-varying vectors of water quality requirements for fish fry species at different growth stages in nearshore aquaculture areas. ，， in, Let be the time-varying vector of water quality requirements for the k-th growth stage of the i-th fish fry species in the nearshore aquaculture area at the t-th unit of time. For the first Standardized aquaculture data for the k-th water quality of the i-th fish fry species at the j-th growth stage in a nearshore aquaculture area within a unit of time. This is the error term at the t-th unit of time. The intercept is... These are the linear regression coefficients. This represents the total number of growth stages.

4. The method for online multi-parameter detection of water quality in nearshore aquaculture areas according to claim 3, characterized in that, Step 2 includes: Using linear mapping, water quality parameters in the real-time nearshore aquaculture area are transformed into vectors to establish a vector time series of water quality parameters in the real-time nearshore aquaculture area. Based on the unit timestamp, the data is aligned between the real-time water quality parameter vector time series in the nearshore aquaculture area and the time-varying vector of water quality requirements for various growth stages of fish fry in the nearshore aquaculture area. Using the Euclidean distance formula, the spatial distance between the real-time water quality parameter vector time series in the nearshore aquaculture area and the time-varying vector of water quality requirements of fish fry at various growth stages in the nearshore aquaculture area is calculated, and the deviation value of the water quality parameter vector in the real-time nearshore aquaculture area is screened out. Calculate the deviation of the water quality parameter vector in each real-time nearshore aquaculture area from the time-varying vector of water quality requirements for each growth stage of fish fry in the nearshore aquaculture area, and assign weights to the deviation values ​​of the water quality parameter vector in the real-time nearshore aquaculture area. Based on the real-time deviation values ​​of water quality parameter vectors within the nearshore aquaculture area and their weights, the overall water quality deviation index within the real-time nearshore aquaculture area per unit time is calculated as follows: ，， in, Let be the overall water quality deviation index in the nearshore aquaculture area at the t-th unit time. Weights of the deviation values ​​of water quality parameters in the nearshore aquaculture area at the t-th unit of time. This represents the actual value of the real-time water quality parameter vector within the nearshore aquaculture area at the t-th unit of time. Let be the optimal time-varying vector value of the water quality requirement for the k-th growth stage of the i-th fish fry species in the nearshore aquaculture area at the t-th unit time. Let $t$ be the upper limit of the time-varying vector of water quality requirements for the $k$-th growth stage of the $i$-th fish fry species in the nearshore aquaculture area at the $j$-th unit of time. Let be the lower limit of the time-varying vector of water quality demand for the kth growth stage of the i-th fish fry species in the nearshore aquaculture area at the t-th unit of time. Based on the time-varying vector of water quality requirements for different growth stages of fish fry species in nearshore aquaculture areas, water quality requirement thresholds for different growth stages of fish fry species in nearshore aquaculture areas are preset according to percentiles. Determine whether the overall deviation index of water quality in the nearshore aquaculture area under real-time conditions is less than the water quality requirement threshold for each growth stage of fish fry in the nearshore aquaculture area. If yes, the water quality is determined to be normal; otherwise, the water quality is determined to be abnormal.

5. The method for online multi-parameter detection of water quality in nearshore aquaculture areas according to claim 1, characterized in that, Step 3 includes: Based on the fish fry species in the nearshore aquaculture area to be tested, obtain historical water quality data of the same fish fry species in the aquaculture area. Based on the natural and anthropogenic factors affecting water quality changes in aquaculture areas, water quality data for historical aquaculture areas and fish fry species are divided to construct a multivariate time series matrix of natural and anthropogenic factors affecting water quality in historical aquaculture areas and fish fry species. Based on Random Forest, a cascaded decision tree for water quality impact on natural and anthropogenic factors is established using historical data on natural and anthropogenic factors affecting the same fish fry species in the same aquaculture area at a given time. Water quality parameters are used as the decision nodes of the cascaded decision tree, and a multivariate time series matrix of natural and anthropogenic factors affecting water quality in the same aquaculture area for the same fish fry species is used as the output. The information gain of each natural and anthropogenic factor affecting the decision nodes is calculated, generating the contribution ratio of natural and anthropogenic factors to water quality changes in the historical water quality data of the same aquaculture area for the same fish fry species.

6. The method for online multi-parameter detection of water quality in nearshore aquaculture areas according to claim 1, characterized in that, Step 3 also includes: Based on the historical water quality data of the same fish fry species in the aquaculture area, a distribution regression fitting was performed on the contribution ratio of natural factors and human factors to water quality changes, and the individual influence coefficients of natural factors and human factors on water quality indicators were calculated. Based on the contribution ratio of natural and anthropogenic factors to water quality changes in historical aquaculture areas and for the same fish fry species, weights are assigned to natural and anthropogenic factors in the water quality data for historical aquaculture areas and for the same fish fry species. The local slope of the natural-anthropological factor ratio, corresponding to the individual influence coefficients of natural and anthropological factors on water quality indicators, is fitted using the least squares method, along with the weights of natural and anthropological factors in historical aquaculture data for the same fish fry species. The historical water quality trend indicators for the nearshore aquaculture area to be tested are then calculated as follows: ，， in, The historical water quality trend indicators of the nearshore aquaculture area to be tested at the t-th unit of time. This represents the influence coefficient of natural factors in water quality data for the same fish fry species in historical aquaculture areas. Let be the coefficient of human influence in the historical water quality data of the same fish fry species in the aquaculture area at the t-th unit of time. The slope of the trend of the influence of natural factors on water quality. Let be the weight of the natural factor component in the historical water quality data of the same fish fry species in the t-th unit of time. Let be the weight of the human factor component in the historical water quality data of the same fish fry species in the t-th unit of time for the aquaculture area. The slope of the trend of human factors' impact on water quality. For natural factors intercept, For human factors intercept, , All are random error terms.

7. The method for online multi-parameter detection of water quality in nearshore aquaculture areas according to claim 6, characterized in that, Step 5 includes: Based on the real-time water quality parameters in the nearshore aquaculture area per unit time, statistical analysis was conducted on the time-series characteristics of the water quality parameters in the nearshore aquaculture area. Based on cosine similarity, the similarity between the water quality parameters of the fish fry species at each growth stage in the known nearshore aquaculture area and the time series feature data of water quality parameters in the real-time nearshore aquaculture area is calculated to classify and filter them, so as to obtain the data set of supporting water quality parameters for the risk types of growth hazards of fish fry species in the real-time nearshore aquaculture area. Based on the water quality parameter dataset supporting the risk types of fish fry growth hazards in real-time nearshore aquaculture areas, the set of real-time nearshore aquaculture fish fry growth hazard risk types is determined by substituting it into an SVM support vector machine. Using Bayesian inference, we verify the number of supporting evidences for each risk type in the set of risk types for fish fry growth in real-time nearshore aquaculture areas based on the risk types for fish fry growth in real-time nearshore aquaculture areas. We obtain the confidence level of the development trend of risk types for fish fry growth in real-time nearshore aquaculture areas and determine the development direction of risk for fish fry growth in real-time nearshore aquaculture areas.

Images