CN115879750B - Aquatic seedling environment monitoring management system and method - Google Patents
Aquatic seedling environment monitoring management system and method Download PDFInfo
- Publication number
- CN115879750B CN115879750B CN202310187538.7A CN202310187538A CN115879750B CN 115879750 B CN115879750 B CN 115879750B CN 202310187538 A CN202310187538 A CN 202310187538A CN 115879750 B CN115879750 B CN 115879750B
- Authority
- CN
- China
- Prior art keywords
- index
- reference index
- prediction
- category
- historical
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A40/00—Adaptation technologies in agriculture, forestry, livestock or agroalimentary production
- Y02A40/80—Adaptation technologies in agriculture, forestry, livestock or agroalimentary production in fisheries management
- Y02A40/81—Aquaculture, e.g. of fish
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the technical field of data processing, in particular to an aquatic seedling environment monitoring management system and method, wherein all historical reference indexes are classified according to an environment influence factor set to obtain matching index categories; classifying the historical reference indexes of the matching index categories according to the frequency values and the linear offset degree to obtain two characteristic categories, and acquiring adjustment parameters by combining the frequency values of each historical reference index in the matching index categories; and constructing a predictive linear regression equation according to the reference index data variable quantity and the adjustment parameters, carrying out nonlinear transformation on the predictive linear regression equation to obtain a predictive probability value of each predictive reference index, and carrying out weighting treatment on all the predictive probability values to obtain a final predictive reference index so as to accurately monitor aquatic seedling conditions on future dates. According to the invention, by optimizing the problem of nonlinear water quality parameter prediction accuracy, the practicability of nonlinear data is higher, and the accuracy of predicting the water quality change of the aquatic seedling environment is improved.
Description
Technical Field
The invention relates to the technical field of data processing, in particular to an aquatic seedling environment monitoring and management system and method.
Background
In aquatic breeding and cultivation, the quality of water quality management not only affects the production performance of fish and shrimp, but also affects the composition and richness of bait organisms. With popularization of aquaculture and aggravation of water pollution, many aquaculture farms have limitations due to lack of water quality management conditions and experience and management efficiency, so that the survival rate of aquaculture seedlings is low, and the yield of aquaculture is reduced. The existing environment data influencing the water quality change through various internet of things sensors are replaced by artificial monitoring, and the environment data is analyzed, processed and fed back by utilizing artificial intelligence, so that the high-efficiency management of the water quality of aquatic seedlings and farms is realized, but the timeliness of water quality management and adjustment cannot be met by real-time adjustment, the water quality change is predicted and prevented, and the aquatic product loss and influence caused by the water quality change can be reduced to the minimum. The existing water quality change prediction mode is to construct a prediction model according to influence relation by means of autocorrelation analysis between influence factors and water quality parameters to obtain a predicted value, but the method is only suitable for water quality parameters with linear water temperature, transmittance and the like, and part of water quality parameters are nonlinear changes, so that the prediction result of the part of water quality parameters is poor in practicability.
In the prior art, all historical data are analyzed to obtain the fish experience data to be predicted, and no matching data matched with the fish experience data to be predicted in the historical data is specifically analyzed, so that larger errors can be generated, and the accuracy of the final prediction result is poor. The existing method obtains the association degree between each water quality environment data variable and the healthy growth index of the fish in a mean value mode, and selects the water quality environment data variable with larger association degree with the healthy growth index of the fish as a key water quality environment data variable affecting the growth of the fish; and regulating and controlling the water quality in the aquaculture system according to the trained predictive regulation and control model. When the association degree is obtained, an averaging method is used, the prediction of the nonlinear data is rough, the final prediction result is error, and the prediction result is inaccurate.
Disclosure of Invention
In order to solve the technical problem that the reference index of water quality is not accurately predicted in the prior art, the invention aims to provide an aquatic seedling environment monitoring and management system and method, and the adopted technical scheme is as follows:
the invention provides an aquatic seedling environment monitoring and managing method, which comprises the following steps:
Obtaining a historical reference index and a corresponding environmental impact factor set of each historical date of aquaculture; the historical reference index of each historical date corresponds to the environmental impact factor set one by one; obtaining an environmental impact factor set of a future date;
classifying all historical reference indexes according to the environmental impact factor sets of each historical date and future date to obtain at least two reference index categories, and taking the reference index category to which the environmental impact factor set of the future date belongs as a matching index category;
classifying all the historical reference indexes according to the frequency value of each historical reference index in the matching index category to obtain two characteristic categories; acquiring adjusting parameters of the corresponding historical reference indexes according to the characteristic category to which each historical reference index belongs in the matching index category and the frequency value;
obtaining the numerical class of all the historical reference indexes in the matching index class; sequentially inserting the prediction reference indexes into the matching index categories by taking the numerical value corresponding to each numerical value category as a prediction reference index, and obtaining the reference index data variation of the matching index categories before and after insertion; constructing a predictive linear regression equation for any one of the predictive reference indexes according to the adjustment parameters of the predictive reference indexes and the reference index data variation;
Nonlinear transformation is carried out on the predictive linear regression equation, and a predictive probability value of each predictive reference index is obtained; weighting each prediction reference index according to the prediction probability value to obtain a final prediction reference index; and monitoring and managing the aquatic seedling raising environment on the future date according to the final prediction reference index.
Further, the set of environmental impact factors includes:
environmental indicators include air temperature, water vapor pressure, cloud cover, air pressure, wind speed and sunshine hours; all environmental indicators of each date are taken as an environmental impact factor set of the corresponding date.
Further, the method for acquiring the feature class comprises the following steps:
the feature categories include a first feature category and a second feature category;
sequencing the frequency values of all the historical reference indexes in the matching index category according to the sequence from the large frequency value to the small frequency value to obtain a frequency sequence; sequentially accumulating from the first frequency value in the frequency sequence until the accumulated value is larger than a preset ideal threshold value;
dividing all the historical reference indexes corresponding to the frequency values participating in the accumulation operation into classes to be classified, removing the historical reference index corresponding to the last frequency value participating in the accumulation operation from the classes to be classified, and taking the updated classes to be classified as the first characteristic classes; and classifying all historical reference indexes belonging to non-first characteristic categories in the matching index category into the second characteristic category.
Further, the method for acquiring the adjustment parameters comprises the following steps:
for any one of the history reference indexes in the matching index category, if the history reference index is the first characteristic category, the adjusting parameter corresponding to the history reference index is a preset stable parameter;
and if the historical reference index is in the second characteristic category, taking the ratio of the frequency value of the corresponding historical reference index to the ideal threshold value as a negative adjustment parameter, and subtracting the negative adjustment parameter from the constant one to obtain the adjustment parameter of the corresponding historical reference index.
Further, the method for acquiring the reference index data variation comprises the following steps:
taking the standard deviation of all the historical reference indexes in the matching index category as a first discrete degree of the matching index category before insertion;
taking the numerical value corresponding to each numerical value category as a prediction reference index, sequentially inserting the prediction reference index into the matching index category, forming a prediction category by all the historical reference indexes and the prediction reference index in the matching index category, and taking the standard deviation of the prediction category as the prediction discrete degree of the matching index category after insertion;
obtaining the absolute value of the difference between the predicted discrete degree of each matched index category after insertion and the first discrete degree of the matched index category before insertion, and taking the absolute value of the difference as the reference index data variation of the matched index categories before and after insertion; each prediction reference index corresponds to one reference index data variation.
Further, the construction method of the predictive linear regression equation comprises the following steps:
obtaining the change quantity of the reference index data corresponding to the prediction reference index of the matched index class after each insertion, and subtracting the frequency value of the matched index class of the predicted reference index after the insertion from a constant 1 to serve as an independent variable of the prediction linear regression equation; and taking the adjusting parameter as an error term and taking the reference index data variation as a linear regression parameter.
Further, the method for obtaining the prediction probability value comprises the following steps:
nonlinear conversion is carried out on the prediction linear regression equation by using a Sigmoid function, so that a corresponding first prediction equation is obtained;
taking the value corresponding to each value category as a prediction reference index, inputting each prediction reference index into the first prediction equation, and outputting the sub-prediction probability of the prediction reference index; and obtaining the prediction probability value of the prediction reference index according to the absolute value of the difference between the constant one and the sub-prediction probability.
Further, the method for obtaining the final prediction reference index comprises the following steps:
accumulating the prediction probability values of all the prediction reference indexes to obtain a comprehensive prediction probability value;
Obtaining the ratio of the predicted probability value of each predicted reference index to the comprehensive predicted probability value, and taking the ratio as the sub-weight of the predicted reference index; taking the product of the sub-weight and the prediction reference index value as a sub-prediction index of the prediction reference index; and accumulating all the sub-prediction indexes of the prediction reference indexes to obtain a final prediction reference index.
The invention also provides an aquatic breeding environment monitoring and managing system, which comprises a memory and a processor, wherein the processor executes a computer program stored in the memory to realize the aquatic breeding environment monitoring and managing method.
The invention has the following beneficial effects:
according to the method, all historical reference indexes are classified according to the environmental impact factor set, and the reference index category to which the environmental impact factor set of future date belongs is used as a matching index category; according to the invention, similar multidimensional environmental influence factors are introduced, so that the relevance between the prediction reference index and the historical reference index is improved, and the prediction result is more accurate. And classifying all the historical reference indexes according to the frequency value to obtain two characteristic categories, wherein the frequency value reflects the influence of the historical reference indexes on the prediction reference indexes, so that the influence weights between the two characteristic categories are different through classification, the prediction reference indexes are analyzed by combining the weights of the historical reference indexes on the influence of the historical reference indexes, and the obtained prediction effect is more accurate. According to the characteristic category and the frequency value of each historical reference index in the matching index category, the adjusting parameters are obtained, the matching degree of the reference index and the historical reference index of each historical date can be predicted by the adjusting parameters, and a basis is provided for a prediction linear regression equation obtained subsequently. The variation of the reference index data reflects the discrete degree difference between the post-insertion matching index category and the pre-insertion matching index category, and can more intuitively represent the structural variation of the pre-insertion and post-insertion matching index category. And constructing a predictive linear regression equation according to the regulating parameters and the reference index data variable quantity, carrying out nonlinear transformation on the predictive linear regression equation to obtain a predictive probability value, and carrying out weighting treatment on each predictive reference index according to the predictive probability value to obtain a final predictive reference index so as to accurately monitor aquatic seedling conditions on future dates. According to the invention, by optimizing the problem of nonlinear water quality parameter prediction accuracy, the practicability of nonlinear data is higher, the prediction result is more accurate, and the accuracy of predicting the water quality change of the aquatic seedling environment is improved.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions and advantages of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of an aquatic seedling environment monitoring and managing method according to an embodiment of the present invention.
Detailed Description
In order to further describe the technical means and effects adopted by the invention to achieve the preset aim, the following detailed description refers to the specific implementation, structure, characteristics and effects of the aquatic seedling environment monitoring and management system and method according to the invention in combination with the accompanying drawings and the preferred embodiment. In the following description, different "one embodiment" or "another embodiment" means that the embodiments are not necessarily the same. Furthermore, the particular features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The invention provides a monitoring and managing system and a method for aquatic seedling environment, which are specifically described below with reference to the accompanying drawings.
Referring to fig. 1, a flowchart of a method for monitoring and managing an aquatic seedling environment according to an embodiment of the present invention is shown, where the method includes:
step S1: obtaining a historical reference index and a corresponding environmental impact factor set of each historical date of aquaculture; the historical reference index of each historical date corresponds to the environmental impact factor set one by one; a set of environmental impact factors for a future date is obtained.
In the embodiment of the invention, the specific implementation scene is the adjustment and control of the water quality parameters in the aquaculture process. The water quality parameter indexes of the aquaculture comprise temperature, pH value, salinity, ammonia nitrogen, hydrogen sulfide, nitrite, available phosphorus, transparency, dissolved oxygen and the like, and the water quality is regulated according to the water quality parameter indexes.
In order to enhance timeliness and accuracy of water quality parameter prediction, characteristics of basic water quality parameter indexes in each period need to be analyzed, the change of the water quality parameter indexes is influenced by external factors and factors inside a seedling raising pool, but a plurality of water quality parameter indexes exist in the factors inside the seedling raising pool to interfere with each other, and if the water quality parameter indexes are used as influencing factors for subsequent prediction, the reliability of the water quality parameter indexes is greatly reduced. Therefore, in the embodiment of the invention, analysis is performed on a certain water quality parameter index, the concentration of dissolved oxygen is selected as the historical reference index of the analysis, the historical reference index of each historical date of aquaculture and a corresponding environmental impact factor set are obtained, and an implementer can determine the historical reference index of the analysis according to a specific implementation scene. Further, the environmental indexes of each historical period comprise air temperature, water vapor pressure, cloud cover, air pressure, air speed and sunshine hours, all the environmental indexes of each date are used as an environmental influence factor set of the corresponding date, and the historical reference indexes of each historical date are in one-to-one correspondence with the environmental influence factor set. In other embodiments, the practitioner may select other types of indices as reference indices, and the other types of indices are the same as the analysis method of the dissolved oxygen concentration according to the embodiment of the present invention, and are not described.
Because the invention aims to improve the timeliness and accuracy of water quality parameter prediction, the future date prediction reference index can be predicted only by acquiring the future date environmental influence factor set, and the future date environmental influence factor set can be directly obtained from a weather website.
Step S2: classifying all the historical reference indexes according to the environmental impact factor sets of each historical date and future date to obtain at least two reference index categories, and taking the reference index category to which the environmental impact factor set of the future date belongs as a matching index category.
A set of historical reference indicators and environmental impact factors for all historical dates, and a set of environmental impact factors for future dates are obtained from step S1. For the reference index, the reference index is influenced by environmental indexes such as air temperature, water vapor pressure, cloud cover, wind speed and the like, and the environmental influence factor set of future date is obtained through a weather website, so that when the prediction reference index of the seedling pool on the future date is predicted, the causal relation between the prediction reference index of the future date and the historical reference index of the historical date is constructed in a time sequence analysis mode.
The prediction reference index of the future date is predicted, and the prediction reference index of the future date can be accurately predicted by analyzing the history reference index corresponding to the history date under the condition that the multidimensional environment influence factors are the same by classifying the history reference index corresponding to the environment influence factor set of the history date which is the same or similar to the future date. Thus, all historical reference indicators are classified according to the set of environmental impact factors for each historical date and future date, resulting in at least two reference indicator categories. In the embodiment of the invention, the method for classifying the historical reference indexes of all historical dates is a Gaussian mixture model, and the Gaussian mixture model is generally used for clustering multidimensional data. The gaussian mixture model is a technical means well known to those skilled in the art, and a specific process is not repeated herein, but only a clustering process of the gaussian mixture model provided in one embodiment of the present invention is briefly described herein:
The first step: and inputting the historical reference index of each historical date and the corresponding environmental impact factor set thereof, adding the environmental impact factor set of the future date into the environmental impact factor set corresponding to the historical reference index of the historical date, and classifying the historical reference indexes of all the historical dates according to all the environmental impact factor sets. The method comprises the steps of taking an environment influence factor set as an independent variable x, taking a historical reference index as a dependent variable y, establishing a plane rectangular coordinate system according to the independent variable and the dependent variable, obtaining data points corresponding to each historical date by the independent variable x and the dependent variable y, and classifying all the data points according to a single Gao Simi-degree distribution condition which is met by each data point in a Gaussian mixture model, namely classifying all the historical reference indexes.
And a second step of: setting the number of components of a preset initialized Gaussian mixture model; the gaussian distribution parameters, i.e., the mean and variance, for each reference index category are randomly initialized. In the embodiment of the present invention, the number of components of the preset initialized gaussian mixture model is 6, and the specific numerical value can be specifically set according to the specific implementation manner.
And a third step of: the probability of each gaussian model to which the history reference index of each history date belongs, that is, the posterior probability, is calculated. The closer the history reference index of each history date is to the center of the gaussian distribution, the greater the posterior probability, i.e., the higher the likelihood that the history reference index of the corresponding history date belongs to the reference index category.
Fourth step: the prior distribution probability, the mean value and the covariance parameters are calculated, so that the probability of the historical reference index of each historical date is maximized, the probability weighting of the historical reference index of the historical date is used for calculating the new parameters, and the weight is the probability that the historical reference index of the corresponding historical date belongs to the reference index category.
Fifth step: the second and third steps are iterated until convergence.
Sixth step: and outputting a plurality of reference index category results, wherein the environmental impact factor sets corresponding to the historical reference indexes of all the historical dates in each reference index category are similar. The predicted reference indicators for the future date are also categorized into reference indicator categories that match the set of environmental impact factors for the future date.
The environmental impact factor sets of the historical reference indexes in each reference index category obtained through Gaussian mixture model classification are similar, but the corresponding historical reference indexes are not the same, namely the historical reference indexes in the reference index category are nonlinear data, although the prediction reference indexes are unknown, no matter the values of the reference indexes on future dates, the corresponding prediction reference indexes are also classified into the matching index category according to the multi-dimensional environmental impact factor set clustering mode. Therefore, under the condition that the prediction reference index of the future date is unknown, the prediction reference index of the future date is supposed to be inserted into the matching index category, but the structural change of the matching index category needs to be as small as possible when the prediction reference index is inserted, so that the maximum posterior probability of the history reference index of the history date in the matching index category to the prediction reference index can be ensured, and further the more accurate prediction reference index of the future date can be obtained. Since the prediction reference index of the future date is unknown, the prediction reference index of the future date needs to be predicted according to the history reference index of the history date in the same reference index category.
Therefore, the reference index category to which the environmental impact factor set of the future date belongs is used as the matching index category, and the prediction reference index of the future date is predicted according to the history reference index of the history date in the matching index category, so that the accuracy of the prediction reference index of the future date can be improved.
Step S3: classifying all the historical reference indexes according to the frequency value of each historical reference index in the matched index category to obtain two characteristic categories; and acquiring the adjusting parameters of the corresponding historical reference indexes according to the characteristic category and the frequency value of each historical reference index in the matching index category.
Step S2 shows that prediction reference indexes of future dates need to be predicted according to historical reference indexes of historical dates in the matching index category. It should be noted that, the logistic regression model is a technical means well known to those skilled in the art, and is not repeated herein, but only the basic principle of the logistic regression model is briefly described: the logistic regression model essentially is that a linear regression model is subjected to nonlinear conversion through a Sigmoid function to obtain a probability value between 0 and 1. The essence of the logistic regression model is the prediction probability, and the linear regression equation obtained according to the original algorithm is the causal relation between the historical reference index of the historical date and a certain environmental index, but the environmental influence factor set is a multi-dimensional parameter, even if the multi-dimensional environmental influence factor set is changed into one-dimensional data in a data dimension reduction mode, the finally output probability value is inaccurate. Therefore, in the embodiment of the invention, the prediction reference index is iterated according to all the historical reference indexes in the matching index category, the prediction reference index is inserted into the matching index category, the data change of the reference index in the matching index category after the insertion of the prediction reference index is obtained, and the causal relation of the linear regression equation is reconstructed.
In the embodiment of the invention, the independent variable value constructed in the linear regression equation is the frequency value of a certain type of history reference index in the matching index category, and when the prediction reference index takes any type of history reference index in the matching index category, compared with the original matching index category, the structure of the matching index category containing the prediction reference index changes, the influence of the prediction reference index with different values on the structure of the matching index category containing the prediction reference index is also different, and compared with the matching index category not containing the prediction reference index, the linear regression line corresponding to the matching index category containing the prediction reference index can deviate.
Therefore, the change of the matching index category structure containing the prediction reference index can feed back the fit degree of the prediction reference index and the history reference index of each history date. Because the traditional logistic regression carries out non-linear transformation on the linear equation by utilizing the Sigmoid function, likelihood estimation is carried out on the output result to obtain a binary probability result, and although the invention does not need to carry out binary classification, the data in the matching index category can be classified into two categories with higher influence weight on the prediction reference index and lower influence weight. Therefore, in order to make the prediction effect on the prediction reference index more accurate, determine the accurate range of the value of the prediction reference index, classify all the history reference indexes according to the frequency value of each history reference index in the matching index class, and obtain two feature classes, in the embodiment of the invention, the method specifically includes:
Firstly, the frequency value of each history reference index in the original matching index category needs to be acquired, and the larger the frequency value is, the more the frequency of the corresponding history reference index in the original matching index category is. The historical reference index with larger frequency value in the original matching index category has higher weight on the influence of the prediction reference index; the historical reference indexes with smaller frequency values in the original matching index category have lower weight on the influence of the prediction reference indexes. When the prediction reference index is a historical reference index with a larger frequency value in the original matching index category, the linear regression offset of the matching index category after the insertion of the prediction reference index is changed less; when the prediction reference index is a historical reference index with a smaller frequency value in the original matching index category, the linear regression of the matching index category after the insertion of the prediction reference index may have a larger offset, and the historical reference index with a smaller corresponding frequency value is changed from small to large due to the insertion of the similar prediction reference index, so that the class with a lower weight on the prediction reference index is converted into the class with a higher weight on the prediction reference index.
Secondly, according to the relation between the frequency values of each history reference index in the original matching index category, the screening conditions of different characteristic categories are determined, and the method specifically comprises the following steps: ordering the frequency values of all the historical reference indexes in the matching index category according to the sequence from the large frequency value to the small frequency value to obtain a frequency sequence; and sequentially accumulating from the first frequency value in the frequency sequence until the accumulated value is larger than a preset ideal threshold value. Dividing all the historical reference indexes corresponding to the frequency values participating in the accumulation operation into categories to be classified, removing the historical reference index corresponding to the last frequency value participating in the accumulation operation from the categories to be classified, and taking the updated categories to be classified as first characteristic categories; all the historical reference indexes belonging to the non-first characteristic category in the matching index category are divided into second characteristic categories, wherein the characteristic categories comprise the first characteristic category and the second characteristic category. The first characteristic category is a category with higher weight which affects the prediction reference index in the original matching index category; the second characteristic category is a category with lower weight which affects the prediction reference index in the original matching index category. In the embodiment of the invention, the preset ideal threshold value is 50%, and the value of the specific ideal threshold value can be specifically set according to specific implementation scenes. In order to more intuitively represent all the reference index classifications in the matching index category, the conditional formulas of the reference index classifications specifically include:
In the method, in the process of the invention,
representing the number of historical reference indexes contained in the accumulated values which are accumulated in sequence from the first frequency value and smaller than the ideal threshold value in the frequency sequence;
representing historical reference indicators
The frequency values in the original matching index category,
representing the ideal threshold value of the value,
representing the function of taking the maximum value. For example, the ideal threshold value is preset to be 50%, the accumulated values of the first three historical reference indicators are 48% and the accumulated values of the first four historical reference indicators are 51% in the frequency sequence, which are accumulated sequentially from the first frequency value. Therefore, the maximum accumulated value smaller than the ideal threshold at this time contains 3 historical reference indexes
I.e.
Then the first three historical reference indicators have higher weight on the influence of the prediction reference indicators, namely the first characteristic category; the historical reference indicators remaining in the frequency sequence have a lower weight on the prediction reference indicators, i.e. the second feature class.
Since the prediction reference index is the first feature class or the second feature class, the structural influence on the matching index class containing the prediction reference index is also different. The matching index type structure change containing the prediction reference index can feed back the fit degree of the prediction reference index and the history reference index of each history date, so that the adjusting factors of the prediction reference index can be constructed according to the characteristic type of each history reference index, and the adjusting factors are combined to obtain the prediction linear regression equation in the subsequent process. Therefore, according to the characteristic category and the frequency value to which each history reference index belongs in the matching index category, the adjusting parameters of the corresponding history reference index are obtained, and in the embodiment of the invention, the method specifically comprises the following steps:
For any one of the history reference indexes in the matched index category, if the history reference index is the first characteristic category, the adjusting parameter corresponding to the history reference index is a preset stable parameter; when the prediction reference index is the first characteristic class, the linear equation of the matching index class containing the prediction reference index is hardly shifted or has smaller shift, the weight of the influence of the history reference index of the first characteristic class on the prediction reference index is large, the influence weight of the similar prediction reference index after insertion is only larger, and the method accords with design expectations. Therefore, in the embodiment of the present invention, the preset stability parameter is set to be a value of 0, and the value of the specific preset stability parameter can be specifically set according to the specific implementation scenario.
If the historical reference index is in the second characteristic category, taking the ratio of the frequency value corresponding to the historical reference index to the ideal threshold value as a negative adjustment parameter, and subtracting the negative adjustment parameter from the constant one to obtain the adjustment parameter corresponding to the historical reference index. When the prediction reference index is the second characteristic category, the linear equation of the matching index category containing the prediction reference index is greatly deviated, and the adjusting coefficient is reversely corrected, namely the influence weight of the second characteristic category on the final prediction reference index is increased. The weight of the second characteristic category influencing the prediction reference index is small, and due to the insertion of the similar prediction reference indexes, the frequency value of the historical reference index in the second characteristic category is changed from small to large, and the category with lower weight influencing the prediction reference index is converted into the category with higher weight influencing the prediction reference index. Therefore, when the prediction reference index is the second feature class, the linear offset generated by the inserted matching index class directly affects the accuracy of the subsequent final prediction reference index, and deviates from the expectation. The acquisition formula of the adjustment parameters of the historical reference index is specifically expressed as follows:
In the method, in the process of the invention,
represents the adjustment coefficient of the historical reference index j,
representing historical reference indicators
The frequency values in the original matching index category,
representing an ideal threshold. The adjustment coefficient of each history reference index represents the possibility of shifting the corresponding linear equation after the corresponding history reference index is used as the prediction reference index to be inserted into the matching index class, and the weight of the influence of the corresponding history reference index on the prediction reference index is adjusted. When the prediction reference index is the first characteristic category, the linear regression offset of the matching index category inserted with the prediction reference index is less in change, and adjustment is not needed; when the prediction reference index is the second feature class, the offset of the linear regression of the matching index class after the insertion of the prediction reference index may be large, and the weight affecting the prediction reference index needs to be adjusted.
Step S4: obtaining the numerical class of all the historical reference indexes in the matched index class; taking the numerical value corresponding to each numerical value category as a prediction reference index, sequentially inserting the prediction reference index into the matching index categories, and obtaining the reference index data variation of the matching index categories before and after insertion; and constructing a predictive linear regression equation for any one predictive reference index according to the adjusting parameters of the predictive reference index and the data variation of the reference index.
As can be seen from step S3, each historical reference index in the matching index category has a corresponding adjustment coefficient. The adjustment coefficient represents the possibility of offset of the corresponding linear equation after the corresponding historical reference index is used as the prediction reference index and inserted into the matching index class, and then the weight of the corresponding historical reference index affecting the prediction reference index is adjusted. Because the values of the historical reference indexes in the matching index categories without the inserted predictive reference indexes are repeated, the adjustment coefficients corresponding to the historical reference indexes corresponding to the repeated values are the same, namely the adjustment degrees of the weights of the historical reference indexes corresponding to the repeated values on the influence of the predictive reference indexes are the same. Therefore, it is necessary to obtain the numerical categories of all the historical reference indicators within the matching indicator category, and then analyze the numerical values of the prediction reference indicators according to each numerical category.
Further, taking the numerical value corresponding to each numerical value category as a prediction reference index, sequentially inserting the prediction reference index into the matching index categories to obtain the reference index data variation of the matching index categories before and after insertion, and specifically comprising:
And taking the standard deviation of all the historical reference indexes in the matching index category as the first discrete degree of the matching index category before insertion. The first degree of dispersion reflects the degree of dispersion between all the historical reference indicators in the pre-insertion matching indicator category, the greater the first degree of dispersion, the more dispersed the numerical distribution of the historical reference indicators in the corresponding pre-insertion matching indicator category.
And taking the numerical value corresponding to each numerical value category as a prediction reference index, sequentially inserting the prediction reference index into the matching index category, forming a prediction category by all the historical reference indexes and the prediction reference indexes in the matching index category, and taking the standard deviation of the prediction category as the prediction discrete degree of the matching index category after insertion. The prediction discrete degree reflects the discrete degree among all the reference indexes in the matching index category after the prediction reference indexes of different numerical value categories are inserted in the matching index category, and the larger the prediction discrete degree is, the more the distribution of the reference index numerical values in the matching index category inserted with the prediction reference indexes is dispersed.
Obtaining the absolute value of the difference between the predicted discrete degree of each matched index category after insertion and the first discrete degree of the matched index category before insertion, and taking the absolute value of the difference as the reference index data variation of the matched index categories before and after insertion; each prediction reference index corresponds to a reference index data variation. The variation of the reference index data reflects the discrete degree difference between the matched index category after insertion and the matched index category before insertion, and the larger the variation of the reference index data is, the larger the discrete degree variation in the matched index category after the corresponding prediction reference index is inserted is.
The linear regression model is designed by sequentially inserting the prediction reference index into the matching index category by taking the numerical value of each numerical value category in the matching index category as the prediction reference index, knowing the frequency value of the prediction reference index, the first discrete degree of the matching index category before insertion and the prediction discrete degree of the matching index category after insertion. When the value of the prediction reference index is any value category, the smaller the variation of the reference index data after the prediction reference index is inserted into the matching index category, the higher the fitting degree of the final prediction reference index to the corresponding prediction reference index is, so that the maximum posterior probability of the history reference index to the final prediction reference index can be ensured, and simply, the reliability of the prediction reference index obtained by the history reference index can be maximized.
According to the embodiment of the invention, under the condition that similar multidimensional environment influence factors are introduced, the prediction reference index is used as a dynamic value to influence the structural variation of the matching index category, the relevance of the prediction reference index and the historical reference index is improved, so that a prediction result is more accurate, and the linear regression equation deviation generated by the matching index category after the insertion of the prediction reference index is required to be corrected according to the adjustment parameters of each historical reference index in the matching index category. The prediction reference index of each numerical class has a corresponding historical reference index in the matching index class, and each index class corresponds to the reference index data change quantity of the matching index class before and after the prediction reference index of each numerical class is inserted into the matching index class because each historical reference index in the matching index class has a corresponding adjusting parameter. Therefore, each numerical class of prediction reference index has a corresponding adjustment parameter and a corresponding reference index data variation. Further, for any one prediction reference index, a prediction linear regression equation is constructed according to the adjustment parameters of the prediction reference index and the reference index data variation, and in the embodiment of the invention, the method specifically includes:
Obtaining the reference index data variation corresponding to the prediction reference index of the matched index class after each insertion, and taking the frequency value of the matched index class after the insertion of the prediction reference index subtracted from the constant 1 as the independent variable of the prediction linear regression equation; and taking the adjustment parameter as an error term and the reference index data variation as a linear regression parameter. It should be noted that, the method for constructing the linear regression equation is a technical means well known to those skilled in the art, and is not repeated herein, but only the predictive linear regression equation provided in one embodiment of the present invention is briefly described as follows:
in the method, in the process of the invention,
the dependent variables representing the predictive linear regression equation,
representing the number of all numeric categories in the matching index category,
represent the first
The prediction reference index of the class value class,
represent the first
Adjusting parameters corresponding to the prediction reference indexes of the class numerical value class,
indicating insertion of the first
The degree of predictive discretization of the matching index class of the predictive reference index of the class numerical class,
a first degree of discretization is indicated,
represent the first
The reference index data variation corresponding to the prediction reference index of the class numerical class,
represent the first
The prediction reference index of the class numerical value class matches the frequency value in the index class after insertion; 
Representing the absolute value function.
In the predictive linear regression equation, the reference index data variation reflects the discrete degree difference between the post-insertion matching index category and the pre-insertion matching index category, and the larger the reference index data variation, the lower the degree of fit between the final predictive reference index and the corresponding predictive reference index is. The function of the adjustment parameters is to correct the linear regression equation offset generated by the matching index category after the prediction reference index is inserted.
Indicating that the frequency values of the prediction reference index in the matched index category after insertion are subjected to logic relation correction. When the inserted prediction reference index is a numerical value class in the matching index class, taking the prediction discrete degree of the matching index class after insertion as the prediction discrete degree
Can be related to the influence factor of (2)
And dependent variable
Is used as a linear regression equation of (c),
to insert into
The linear relation change amount of the matching index class after predicting the reference index of the class numerical value class, when the change amount of the reference index data is smaller and
the larger the dependent variable
The smaller the variation, i.e. the first
The higher the degree of fit of the prediction reference index of the class numerical class with the final prediction reference index.
Step S5: nonlinear transformation is carried out on the predictive linear regression equation, and a predictive probability value of each predictive reference index is obtained; weighting each prediction reference index according to the prediction probability value to obtain a final prediction reference index; and monitoring and managing the aquatic seedling raising environment on the future date according to the final prediction reference index.
And (4) obtaining a predictive linear regression equation through the analysis of the step (S4), wherein the predictive linear regression equation is subjected to nonlinear conversion through a Sigmoid function due to the nature of the logistic regression model. Therefore, the nonlinear transformation is performed on the predictive linear regression equation to obtain the predictive probability value of each predictive reference index, which in the embodiment of the invention specifically includes:
and performing nonlinear transformation on the predictive linear regression equation by using the Sigmoid function to obtain a corresponding first predictive equation. Taking the numerical value corresponding to each numerical value category as a prediction reference index, inputting each prediction reference index into a first prediction equation, and outputting the sub-prediction probability of the prediction reference index; and obtaining the prediction probability value of the prediction reference index according to the absolute value of the difference between the constant one and the sub-prediction probability. It should be noted that, the method of using Sigmoid function to perform nonlinear transformation on the linear regression equation is a technical means well known to those skilled in the art, and will not be described herein. The formula for obtaining the prediction probability value specifically comprises the following steps:
in the method, in the process of the invention,
represent the first
A predicted probability value for a predicted reference indicator of the class value class,
dependent variables representing predictive linear regression equations or predictive probability equations, 
Represents a natural constant of the natural product,
represent the first
Sub-prediction probabilities of the prediction references of the class value class,
representing the absolute value function.
When the predicted probability value is smaller, i.e. when
After the prediction reference index of the class numerical value class is inserted into the matching index class, the first
The smaller the linear relation change amount of the prediction reference index of the class numerical value class in the matching index class is, namely
The larger the size of the container,
the smaller, then follow the above will be
Logic where the structure of the inserted matching index class should be less affected by the prediction reference index of the class value class, i.e.
The smaller. Thus utilizing
Correcting logic relationship when
The smaller the time that is taken for the device to be,
the larger.
If the prediction probability value of the prediction reference index is larger, the probability that the corresponding prediction reference index is the final prediction reference index is higher; if the prediction probability value of the prediction reference index is smaller, the probability that the corresponding prediction reference index is the final prediction reference index is lower.
In order to obtain more accurate values of prediction reference indexes and realize accurate monitoring of aquatic seedling growing conditions on future dates, weighting each prediction reference index according to a prediction probability value to obtain a final prediction reference index, and the method specifically comprises the following steps of:
Accumulating the predicted probability values of all the predicted reference indexes to obtain a comprehensive predicted probability value; obtaining the ratio of the predicted probability value of each predicted reference index to the comprehensive predicted probability value, and taking the ratio as the sub-weight of the predicted reference index; taking the product of the sub-weight and the prediction reference index value as a sub-prediction index of the prediction reference index; and accumulating the sub-prediction indexes of all the prediction reference indexes to obtain a final prediction reference index. The obtaining formula of the final prediction reference index specifically comprises:
in the method, in the process of the invention,
representing the final prediction reference index(s),
representing the number of all numeric categories in the matching index category,
represent the first
The prediction reference index of the class value class,
represent the first
The value corresponding to the prediction reference index of the class value class,
represent the first
A predicted probability value for a predicted reference indicator of the class value class,
representing the composite predicted probability value.
In the formulation of the final prediction reference index,
represent the first
Sub-weights of prediction reference indicators of class numerical categories, use
Will be
And normalizing, and multiplying the numerical values of the corresponding prediction reference indexes to obtain sub-prediction indexes of the corresponding prediction reference indexes.
For weighted calculation, a final prediction reference index is obtained, the final prediction reference index represents a reference index of a future date obtained through prediction, and the monitoring management of the aquatic seedling raising environment of the future date can be realized through the final prediction reference index.
Because the concentration of the dissolved oxygen is selected as a historical reference index for analyzing the water body in the embodiment of the invention, the dissolved oxygen of the known water body is kept between 5mg/L and 8mg/L and at least at more than 4mg/L and not lower than 3mg/L; the later stage of high-density fine culture is not lower than 4mg/L. Therefore, in the embodiment of the invention, the preset reference index threshold value is 4mg/L.
Further, monitoring management is performed on the aquatic seedling raising environment on the future date according to the final prediction reference index, and the specific monitoring management process in the embodiment of the invention specifically comprises the following steps: if the final predicted reference index is smaller than the preset reference index threshold, the reference index of the water body is improved by improving the manual pumping power, prolonging the pumping time or increasing the oxygen in micropores at the bottom of the seedling raising pool so as to prevent the problems of low dissolved oxygen concentration of the seedling raising pool and bubble diseases of fish fries caused by environmental indexes such as air temperature, cloud quantity, wind speed and the like in future dates.
In summary, according to the invention, all the historical reference indexes are classified according to the environmental impact factor set, and the reference index category to which the environmental impact factor set of the future date belongs is used as the matching index category; according to the invention, similar multidimensional environmental influence factors are introduced, so that the relevance between the prediction reference index and the historical reference index is improved, and the prediction result is more accurate. And classifying all the historical reference indexes according to the frequency value to obtain two characteristic categories, wherein the frequency value reflects the influence of the historical reference indexes on the prediction reference indexes, so that the influence weights between the two characteristic categories are different through classification, the prediction reference indexes are analyzed by combining the weights of the historical reference indexes on the influence of the historical reference indexes, and the obtained prediction effect is more accurate. According to the characteristic category and the frequency value of each historical reference index in the matching index category, the adjusting parameters are obtained, the matching degree of the reference index and the historical reference index of each historical date can be predicted by the adjusting parameters, and a basis is provided for a prediction linear regression equation obtained subsequently. The variation of the reference index data reflects the discrete degree difference between the post-insertion matching index category and the pre-insertion matching index category, and can more intuitively represent the structural variation of the pre-insertion and post-insertion matching index category. And constructing a predictive linear regression equation according to the regulating parameters and the reference index data variable quantity, carrying out nonlinear transformation on the predictive linear regression equation to obtain a predictive probability value, and carrying out weighting treatment on each predictive reference index according to the predictive probability value to obtain a final predictive reference index so as to accurately monitor aquatic seedling conditions on future dates. According to the invention, by optimizing the problem of nonlinear water quality parameter prediction accuracy, the practicability of nonlinear data is higher, the prediction result is more accurate, and the accuracy of predicting the water quality change of the aquatic seedling environment is improved.
The invention also provides an aquatic seedling environment monitoring and managing system which comprises a memory and a processor, wherein the processor executes a computer program stored in the memory to realize the aquatic seedling environment monitoring and managing method.
It should be noted that: the sequence of the embodiments of the present invention is only for description, and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments.
Claims (6)
1. The aquatic seedling environment monitoring and managing method is characterized by comprising the following steps:
obtaining a historical reference index and a corresponding environmental impact factor set of each historical date of aquaculture; the historical reference index of each historical date corresponds to the environmental impact factor set one by one; obtaining an environmental impact factor set of a future date;
Classifying all historical reference indexes according to the environmental impact factor sets of each historical date and future date to obtain at least two reference index categories, and taking the reference index category to which the environmental impact factor set of the future date belongs as a matching index category;
classifying all the historical reference indexes according to the frequency value of each historical reference index in the matching index category to obtain two characteristic categories; acquiring adjusting parameters of the corresponding historical reference indexes according to the characteristic category to which each historical reference index belongs in the matching index category and the frequency value;
obtaining the numerical class of all the historical reference indexes in the matching index class; sequentially inserting the prediction reference indexes into the matching index categories by taking the numerical value corresponding to each numerical value category as a prediction reference index, and obtaining the reference index data variation of the matching index categories before and after insertion; constructing a predictive linear regression equation for any one of the predictive reference indexes according to the adjustment parameters of the predictive reference indexes and the reference index data variation;
nonlinear transformation is carried out on the predictive linear regression equation, and a predictive probability value of each predictive reference index is obtained; weighting each prediction reference index according to the prediction probability value to obtain a final prediction reference index; monitoring and managing aquatic seedling raising environment in future date according to the final prediction reference index;
The method for acquiring the feature class comprises the following steps:
the feature categories include a first feature category and a second feature category;
sequencing the frequency values of all the historical reference indexes in the matching index category according to the sequence from the large frequency value to the small frequency value to obtain a frequency sequence; sequentially accumulating from the first frequency value in the frequency sequence until the accumulated value is larger than a preset ideal threshold value;
dividing all the historical reference indexes corresponding to the frequency values participating in the accumulation operation into classes to be classified, removing the historical reference index corresponding to the last frequency value participating in the accumulation operation from the classes to be classified, and taking the updated classes to be classified as the first characteristic classes; dividing all historical reference indexes belonging to non-first characteristic categories in the matching index category into the second characteristic category;
the method for acquiring the adjustment parameters comprises the following steps:
for any one of the history reference indexes in the matching index category, if the history reference index is the first characteristic category, the adjusting parameter corresponding to the history reference index is a preset stable parameter;
if the historical reference index is in the second characteristic category, taking the ratio of the frequency value of the corresponding historical reference index to the ideal threshold value as a negative adjustment parameter, and subtracting the negative adjustment parameter from a constant one to obtain an adjustment parameter of the corresponding historical reference index;
The method for acquiring the reference index data variation comprises the following steps:
taking the standard deviation of all the historical reference indexes in the matching index category as a first discrete degree of the matching index category before insertion;
taking the numerical value corresponding to each numerical value category as a prediction reference index, sequentially inserting the prediction reference index into the matching index category, forming a prediction category by all the historical reference indexes and the prediction reference index in the matching index category, and taking the standard deviation of the prediction category as the prediction discrete degree of the matching index category after insertion;
obtaining the absolute value of the difference between the predicted discrete degree of each matched index category after insertion and the first discrete degree of the matched index category before insertion, and taking the absolute value of the difference as the reference index data variation of the matched index categories before and after insertion; each prediction reference index corresponds to one reference index data variation.
2. The method of claim 1, wherein the set of environmental impact factors comprises:
environmental indicators include air temperature, water vapor pressure, cloud cover, air pressure, wind speed and sunshine hours; all environmental indicators of each date are taken as an environmental impact factor set of the corresponding date.
3. The method for monitoring and managing the aquatic seedling environment according to claim 1, wherein the method for constructing the predictive linear regression equation comprises the following steps:
obtaining the change quantity of the reference index data corresponding to the prediction reference index of the matched index class after each insertion, and subtracting the frequency value of the matched index class of the predicted reference index after the insertion from a constant 1 to serve as an independent variable of the prediction linear regression equation; and taking the adjusting parameter as an error term and taking the reference index data variation as a linear regression parameter.
4. The method for monitoring and managing an aquatic seedling environment according to claim 3, wherein the method for obtaining the predicted probability value comprises the following steps:
nonlinear conversion is carried out on the prediction linear regression equation by using a Sigmoid function, so that a corresponding first prediction equation is obtained;
taking the value corresponding to each value category as a prediction reference index, inputting each prediction reference index into the first prediction equation, and outputting the sub-prediction probability of the prediction reference index; and obtaining the prediction probability value of the prediction reference index according to the absolute value of the difference value between the constant 1 and the sub-prediction probability.
5. The method for monitoring and managing an aquatic seedling environment according to claim 4, wherein the method for obtaining the final prediction reference index comprises:
accumulating the prediction probability values of all the prediction reference indexes to obtain a comprehensive prediction probability value;
obtaining the ratio of the predicted probability value of each predicted reference index to the comprehensive predicted probability value, and taking the ratio as the sub-weight of the predicted reference index; taking the product of the sub-weight and the prediction reference index value as a sub-prediction index of the prediction reference index; and accumulating all the sub-prediction indexes of the prediction reference indexes to obtain a final prediction reference index.
6. An aquatic breeding environment monitoring and management system comprising a memory and a processor, wherein the processor executes a computer program stored in the memory to implement an aquatic breeding environment monitoring and management method according to any one of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310187538.7A CN115879750B (en) | 2023-03-02 | 2023-03-02 | Aquatic seedling environment monitoring management system and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310187538.7A CN115879750B (en) | 2023-03-02 | 2023-03-02 | Aquatic seedling environment monitoring management system and method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115879750A CN115879750A (en) | 2023-03-31 |
| CN115879750B true CN115879750B (en) | 2023-05-05 |
Family
ID=85761742
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310187538.7A Active CN115879750B (en) | 2023-03-02 | 2023-03-02 | Aquatic seedling environment monitoring management system and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115879750B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116204566B (en) * | 2023-04-28 | 2023-07-14 | 深圳市欣冠精密技术有限公司 | Digital factory monitoring big data processing system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104392391A (en) * | 2014-11-14 | 2015-03-04 | 国家电网公司 | Power grid running safety risk quantification method |
| CN113344972A (en) * | 2021-05-25 | 2021-09-03 | 浙江工业大学 | Intensive culture-based fish track prediction method |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9955552B2 (en) * | 2011-04-14 | 2018-04-24 | Suntracker Technologies Ltd. | Predictive daylight harvesting system |
| CA2942048C (en) * | 2016-04-27 | 2023-08-29 | Sigma Space Corp. | Methods and apparatus for adaptive multisensor analysis and aggregation |
| CN114997775A (en) * | 2022-05-19 | 2022-09-02 | 深圳市展之信供应链管理有限公司 | Intelligent regulation and control method and system for improving transportation survival rate of aquatic animals |
-
2023
- 2023-03-02 CN CN202310187538.7A patent/CN115879750B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104392391A (en) * | 2014-11-14 | 2015-03-04 | 国家电网公司 | Power grid running safety risk quantification method |
| CN113344972A (en) * | 2021-05-25 | 2021-09-03 | 浙江工业大学 | Intensive culture-based fish track prediction method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115879750A (en) | 2023-03-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109816221B (en) | Project risk decision method, apparatus, computer device and storage medium | |
| CN110782658B (en) | Traffic prediction method based on LightGBM algorithm | |
| CN108710974B (en) | Water ammonia nitrogen prediction method and device based on deep belief network | |
| CN112906298B (en) | Blueberry yield prediction method based on machine learning | |
| CN115879750B (en) | Aquatic seedling environment monitoring management system and method | |
| CN113435707B (en) | Soil testing formula fertilization method based on deep learning and weighting multi-factor evaluation | |
| CN115529987B (en) | Air port regulating and controlling method, device, equipment and storage medium for crop facility | |
| CN113537639A (en) | Method and system for establishing fish growth prediction model as culture environment parameter | |
| CN114662742A (en) | Dynamic decision method and device for field irrigation and fertilization system | |
| CN114818888B (en) | Soil composition data fusion method and system based on multi-channel Kalman filtering | |
| CN115907195A (en) | Photovoltaic power generation power prediction method, system, electronic device and medium | |
| CN112184412A (en) | Modeling method, device, medium and electronic equipment of credit rating card model | |
| CN116882562A (en) | Agricultural carbon emission influence factor analysis method and system | |
| CN116595333B (en) | Soil-climate intelligent rice target yield and nitrogen fertilizer consumption determination method | |
| CN109472320B (en) | Method for automatically correcting variety parameters of crop growth period model under uncertain conditions | |
| Mulema et al. | Quality and productivity in aquaculture: Prediction of oreochromis mossambicus growth using a transfer function arima model | |
| CN114971005A (en) | Bay water temperature combination prediction method based on LSTM and differential regression model dynamic weighting | |
| CN117196881B (en) | Intelligent cultivation information management system based on big data | |
| CN117764762B (en) | Method for improving nitrogen fertilizer utilization rate in corn planting | |
| CN115796558B (en) | Tobacco planting area intelligent management system based on deep neural network | |
| CN116484193A (en) | Crop yield prediction method, system, equipment and medium | |
| CN116908405A (en) | Scene analysis method and system for soil monitoring data | |
| CN115564596A (en) | Reference index obtaining method and device based on artificial intelligence analysis, electronic equipment and computer readable storage medium | |
| Tang et al. | A Durian Yield Prediction Method Based on an Improved Multiple Regression Model | |
| CN116702841A (en) | GDP prediction method combining GAN and time sequence data periodic decomposition |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |