Automatic feeding method and system for aquaculture

Abstract

The invention discloses an automatic feeding method and system for aquaculture, and relates to the technical field of aquaculture, and the method comprises the steps: collecting basic information and aquaculture information of each aquaculture unit; clustering the breeding information by adopting an algorithm, distributing a feeding scheme label for each feeding operation, and identifying a feeding scheme of the breeding unit according to a plurality of feeding scheme labels; performing automatic feeding according to the feeding scheme and the basic information; and constructing a meteorological network diagram based on the breeding meteorological information and the meteorological prediction information, analyzing the meteorological network diagram by using the trained neural network, judging the risk degree of the feeding and performing early warning. The feeding operation is clustered by adopting the DBSCAN algorithm, the feeding scheme label is identified, and the calculation efficiency and the calculation accuracy of the feeding scheme are improved. According to the method, the meteorological network diagram is constructed, and the trained neural network GNN is used for analyzing the meteorological network diagram, so that accurate judgment of the risk degree is realized, and the pond culture risk is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of aquaculture, and more particularly to an automatic feeding method and system for aquaculture. Background Art

[0002] China is a big fishing country. Aquaculture has become the main fishery for aquatic product production, especially fish farming, which is the main guarantee for animal protein supply. In the process of fish farming, the actual demand for fish feed is complex and changeable, which is determined by environmental factors such as the water temperature of the breeding tank and the actual growth stage of the fish. The feeding method of traditional fish farming cannot grasp the feeding status of aquatic organisms in real time. It is based on experience to feed at a fixed time and quantitatively. This feeding method cannot be accurately adjusted according to changes in the breeding environment and the growth habits of fish. It cannot accurately grasp the feeding status of different breeding organisms, different growth periods, and different environmental conditions, which will cause the problem of too much or too little feeding; too much feeding causes excess feed to sink to the bottom of the pond, and then the organic matter rots and deteriorates, exceeding the purification capacity of the pond itself, causing the growth of harmful bacteria and the enrichment of toxic gases, affecting the health of aquatic animals; excessive feeding also increases the cost of breeding; insufficient feeding causes aquatic animals to be unable to eat fully, affecting the growth of aquatic animals, resulting in reduced production and affecting the economic benefits of breeders; feeding cannot be adjusted according to meteorological changes, increasing the risk of pond breeding.

[0003] Therefore, how to propose an automatic feeding method and system for aquaculture, realize the automation of feeding, enhance the refinement of automatic feeding, realize the accurate judgment of the risk level of feeding scheme, and reduce the risk of pond farming, is a problem that technical personnel in this field urgently need to solve. Summary of the invention

[0004] In view of this, the present invention provides an automatic feeding method and system for aquaculture, which realizes the automated processing of feeding, enhances the refinement of automatic feeding, realizes the accurate judgment of the risk level of feeding scheme, and reduces the risk of pond farming. In order to achieve the above purpose, the present invention adopts the following technical scheme:

[0005] An automatic feeding method for aquaculture, comprising:

[0006] Collect basic information and breeding information of each breeding unit;

[0007] An algorithm is used to cluster the farming information, a feeding scheme label is assigned to each feeding operation, and the feeding scheme of the farming unit is identified based on multiple feeding scheme labels;

[0008] Automatically feed the breeding unit according to the feeding plan and basic information of the breeding unit;

[0009] A meteorological network diagram is constructed based on the current aquaculture meteorological information and meteorological forecast information. The trained neural network is used to analyze the meteorological network diagram to determine the risk level of this feeding and issue an early warning.

[0010] Optionally, the basic information of each breeding unit includes: the number, location, and activation status of the breeding unit.

[0011] Optionally, the breeding information includes: fish species information, input information, feeding water quality, breeding species, fish species specifications and quantity, feed information, breeding process information, wherein the feeding water quality includes: water temperature, dissolved oxygen, pH value, ammonia nitrogen and nitrite value, the feed information includes: nutritional indicators, specifications, varieties and brands, and the breeding process information includes: growth rate, death and disease information.

[0012] Optionally, the method further includes performing parameter optimization on the breeding information, performing multi-parameter continuous value regression analysis using the ReliefF algorithm, and obtaining parameter types with a greater correlation with automatic feeding through feature weight screening.

[0013] Optionally, the multi-parameter continuous value regression analysis using the ReliefF algorithm includes:

[0014] The sensor parameters of the time series data are screened for relevance using the ReliefF algorithm. The larger the characteristic weight of the parameter, the greater the influence factor of the parameter on automatic feeding, and the smaller the weight, the smaller the influence factor of its relevance.

[0015] Determine the total data sample set E and feature set G, and initialize them according to the sampling number n and the selected m nearest neighbor samples;

[0016] Randomly select a sample H from the training sample set E and find the m nearest neighbor samples R of H from the same sample set of H. i , i = 1, 2, 3, ..., m, and then find m nearest neighbor samples G from each sample set that is different from H i (D) and perform k repeated sampling;

[0017] Calculate the weight matrix O(T) of each parameter:

[0018]

[0019] Among them, p(D) represents the probability of category D appearing;

[0020] The weights of all sensor feature sets G are calculated in turn to obtain the parameter weight matrix O, and then reordered according to the weight size.

[0021] Optionally, clustering the farming information using an algorithm, assigning a feeding scheme label to each feeding operation, and identifying the feeding scheme of the farming unit according to multiple feeding scheme labels includes:

[0022] Select the DBSCAN algorithm, the algorithm parameters are the neighborhood radius of a point, the minimum number of points in a cluster, and the minimum number of feeding operations required to form an independent feeding plan label;

[0023] Get the data set obtained by re-sorting the weights, traverse each point in the data set for a single feeding operation, query all points in the field of the selected point feeding operation for clustering:

[0024] After obtaining the clustering results, the feeding quantity and feeding frequency of a cluster are compared with the corresponding set thresholds. If they are greater than the threshold, a label of increasing the amount or feeding frequency is added, otherwise the feeding plan is directly output.

[0025] Optionally, the querying of the selected point feeding operation and clustering of all points in the field include: if the number of points of all points in the field, i.e., the number of feeding operations, is greater than or equal to the minimum number of points, marking the selected point as a core point and creating a new cluster; adding the selected point and all points in the field to the cluster; if the number of all points in the field is less than the minimum number of points, any point is a boundary point; for each boundary point, finding the nearest core point and assigning it to the corresponding cluster; points in the data set that are not assigned to any cluster are marked as noise points.

[0026] Optionally, the automatic feeding of the breeding unit according to the identified feeding plan and basic information of the breeding unit includes: parsing the feeding plan, obtaining the feeding type, feeding quantity, and feeding frequency; if the feeding plan is accompanied by an increase in quantity or feeding frequency label, the feeding time is calculated according to the feeding frequency, and the feeding process is monitored, including the feeding quantity and the throwing distance.

[0027] Optionally, constructing a meteorological network diagram based on current aquaculture meteorological information and meteorological forecast information includes: obtaining meteorological information of the current aquaculture area according to sensors, obtaining meteorological forecast information of the current aquaculture area according to the network, obtaining historical meteorological information of the current aquaculture area, correcting the meteorological forecast information according to the historical meteorological information, and constructing a meteorological network diagram based on time nodes according to the corrected meteorological forecast information and the meteorological information of the current aquaculture area.

[0028] Optionally, an automatic feeding system for aquaculture, comprising:

[0029] Collection module: used to collect basic information and breeding information of each breeding unit;

[0030] Scheme identification module: used to cluster the farming information using an algorithm, assign a feeding scheme label to each feeding operation, and identify the feeding scheme of the farming unit based on multiple feeding scheme labels;

[0031] Automatic feeding module: used for the system to automatically feed the breeding unit according to the feeding plan and basic information of the breeding unit;

[0032] Early warning module: It is used to construct a meteorological network diagram based on the current aquaculture meteorological information and meteorological forecast information, and use the trained neural network to analyze the meteorological network diagram to determine the risk level of this feeding and issue an early warning.

[0033] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses an automatic feeding method and system for aquaculture, which has the following beneficial effects:

[0034] The present invention proposes an automatic feeding method for aquaculture, including: collecting basic information and breeding information of each breeding unit; clustering the breeding information using an algorithm, assigning a feeding scheme label to each feeding operation, and identifying the feeding scheme of the breeding unit according to multiple feeding scheme labels; automatically feeding the breeding unit according to the identified feeding scheme and basic information of the breeding unit; constructing a meteorological network diagram based on current breeding meteorological information and meteorological forecast information, using a trained neural network to analyze the meteorological network diagram, judging the risk level of this feeding and issuing an early warning. The present invention clusters the feeding operations using the DBSCAN algorithm, identifies the feeding scheme label, and assigns a feeding scheme label to each feeding operation, thereby improving the calculation efficiency and the accuracy of the feeding scheme calculation. Constructing a meteorological network diagram, using a trained neural network GNN to analyze the meteorological network diagram; judging the risk level of this feeding, achieving accurate judgment of the risk level and automated processing of the feeding operation, enhancing the automation level of automatic feeding, and reducing the risk of pond breeding. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0036] Figure 1 The present invention provides a schematic flow chart of an automatic feeding method for aquaculture.

[0037] Figure 2 This is a structural framework diagram of an automatic feeding system for aquaculture provided by the present invention. DETAILED DESCRIPTION

[0038] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0039] The embodiment of the present invention discloses an automatic feeding method for aquaculture, such as Figure 1 As shown, including:

[0040] Collect basic information and breeding information of each breeding unit;

[0041] An algorithm is used to cluster the farming information, a feeding scheme label is assigned to each feeding operation, and the feeding scheme of the farming unit is identified based on multiple feeding scheme labels;

[0042] Automatically feed the breeding unit according to the feeding plan and basic information of the breeding unit;

[0043] A meteorological network diagram is constructed based on the current aquaculture meteorological information and meteorological forecast information. The trained neural network is used to analyze the meteorological network diagram to determine the risk level of this feeding and issue an early warning.

[0044] Furthermore, the basic information of each breeding unit includes: the number, location, and activation status of the breeding unit.

[0045] Furthermore, the breeding information includes: fish species information, input information, feeding water quality, breeding species, fish species specifications and quantity, feed information, breeding process information, wherein the feeding water quality includes: water temperature, dissolved oxygen, pH value, ammonia nitrogen and nitrite value, the feed information includes: nutritional indicators, specifications, varieties and brands, and the breeding process information includes: growth rate, death and disease information.

[0046] Furthermore, it also includes parameter optimization of the breeding information, using the ReliefF algorithm to perform regression analysis on multi-parameter continuous values, and obtaining parameter types with a greater correlation with automatic feeding through feature weight screening.

[0047] Furthermore, the multi-parameter continuous value regression analysis using the ReliefF algorithm includes:

[0048] The sensor parameters of the time series data are screened for relevance using the ReliefF algorithm. The larger the characteristic weight of the parameter, the greater the influence factor of the parameter on automatic feeding, and the smaller the weight, the smaller the influence factor of its relevance.

[0049] Determine the total data sample set E and feature set G, and initialize them according to the sampling number n and the selected m nearest neighbor samples;

[0050] Randomly select a sample H from the training sample set E and find the m nearest neighbor samples R of H from the same sample set of H. i , i = 1, 2, 3, ..., m, and then find m nearest neighbor samples G from each sample set that is different from H i (D) and perform k repeated sampling;

[0051] Calculate the weight matrix O(T) of each parameter:

[0052]

[0053] Among them, p(D) represents the probability of category D appearing;

[0054] The weights of all sensor feature sets G are calculated in turn to obtain the parameter weight matrix O, and then reordered according to the weight size.

[0055] Furthermore, the algorithm is used to cluster the farming information, a feeding scheme label is assigned to each feeding operation, and the feeding scheme of the farming unit is identified according to multiple feeding scheme labels, including:

[0056] Select the DBSCAN algorithm, the algorithm parameters are the neighborhood radius of a point, the minimum number of points in a cluster, and the minimum number of feeding operations required to form an independent feeding plan label;

[0057] Get the data set obtained by re-sorting the weights. According to the filtered data set with large weights, traverse the single feeding operation of each point in the data set, query all the points in the field of the selected point feeding operation for clustering:

[0058] After obtaining the clustering results, the feeding quantity and feeding frequency of a cluster are compared with the corresponding set thresholds. If they are greater than the threshold, a label of increasing the amount or feeding frequency is added, otherwise the feeding plan is directly output.

[0059] Furthermore, the querying of the selected point feeding operation and clustering of all points in the field include: if the number of points of all points in the field, i.e., the number of feeding operations, is greater than or equal to the minimum number of points, marking the selected point as a core point and creating a new cluster; adding the selected point and all points in the field to the cluster; if the number of all points in the field is less than the minimum number of points, any point is a boundary point; for each boundary point, finding the nearest core point and assigning it to the corresponding cluster; points in the data set that are not assigned to any cluster are marked as noise points.

[0060] In a specific implementation, the selection of DBSCAN parameters includes: a neighborhood radius r of a point, a minimum number of points m of a cluster; wherein a point represents a single feeding operation, i.e., an increase in the number of times and / or a decrease in the number of times and / or an increase in the number of times and / or a decrease in the number of times and / or a feeding quantity and / or a feeding change frequency for a period of time; the neighborhood radius ε is a specific value, and the feeding quantity change and / or the increase in the number of times and / or the decrease in the number of times and / or the decrease in the number of times within the neighborhood radius ε are identified as the same feeding operation and are classified into the same category; at least m feeding operations are required on a specific road section to form an independent feeding plan label;

[0061] Traverse each point in the data set for a single feeding operation, and query all points in the r area where point a is fed: G r (a) = {b∈U|dist(a, b)≤r}; where U is a data set including quantity and / or feeding change frequency and / or increase and / or decrease and / or increase and / or decrease, dist is a distance function; a, b are two different points in the data set, i.e., two different feeding operations, G r (a) is the number of points in the neighborhood of point a r;

[0062] Among them, K a , K b are the feature vectors corresponding to the feeding operation at point a and the feeding operation at point b in the data set; 1 n is an n×n matrix with all 1s, where n is the number of data points; K is the K calculated by the radial basis function as the kernel function a , K b The kernel matrix between

[0063] If G r If the number of points in (a) is greater than or equal to m, mark a as the core point; create a new cluster; and combine a and G r All points in (a) are added to this cluster; if G r If the number of points in (a) is less than m, then a is a boundary point; for each boundary point, find the nearest core point and assign it to the corresponding cluster; points in the data set that are not assigned to any cluster are marked as noise points;

[0064] After the clustering results are obtained, the feeding quantity, increase times and / or decrease times and / or increase quantity and / or decrease quantity of a cluster are compared with the corresponding set thresholds. If it is greater than the threshold, an increase quantity or feeding frequency label is added, otherwise the feeding plan is directly output.

[0065] Furthermore, the automatic feeding of the breeding unit according to the identified feeding plan and basic information of the breeding unit includes: parsing the feeding plan, obtaining the feeding type, feeding quantity, and feeding frequency; if the feeding plan is accompanied by an increase in quantity or feeding frequency label, the feeding time is calculated according to the feeding frequency, and the feeding process is monitored, including the feeding quantity and the throwing distance.

[0066] In a specific implementation manner, the feeding scheme for automatic feeding is determined according to the information of the farmed fish species, the feeding water quality, the farmed species, the fish species specifications and quantity, and the farming process information, wherein the feeding water quality includes: water temperature, dissolved oxygen, pH value, ammonia nitrogen and nitrite value, and the farming process information includes: growth rate, death and disease information, and the input information and feed information in the scheme are obtained according to the analyzed feeding scheme. The feed information includes: nutritional indicators, specifications, species and brand, and the feeding is divided into times based on the comprehensive feeding quantity and feeding times, and the feeding process is monitored, including the feeding quantity, the remaining feeding quantity, and the throwing distance.

[0067] Furthermore, the construction of a meteorological network diagram based on the current aquaculture meteorological information and meteorological forecast information includes: obtaining meteorological information of the current aquaculture area according to sensors, obtaining meteorological forecast information of the current aquaculture area according to the network, obtaining historical meteorological information of the current aquaculture area, and compiling meteorological forecast information through meteorological historical information, and constructing a meteorological network diagram based on time nodes according to the revised meteorological forecast information and the meteorological information of the current aquaculture area.

[0068] In a specific implementation manner, the use of a trained neural network to analyze the meteorological network diagram to determine the risk level of this feeding includes:

[0069] Obtain the basic information of each breeding unit, including: the number, location, and activation status of the breeding unit; obtain the current meteorological data of the corresponding longitude and latitude of the breeding unit according to the meteorological network diagram of the time node; construct a reference series based on the historical meteorological information; and construct a comparison series based on the current meteorological data;

[0070] Normalize the meteorological factors, use the processed data to calculate the geometric similarity between each comparison series curve and the reference series curve, and regard the geometric similarity as the correlation coefficient between the comparison series and the reference series. Calculate the correlation of each meteorological factor, sort the correlation of each factor from large to small, and take the first four factors as the features of the neural network input vector;

[0071] Obtain the k nearest neighbors of the input data, then determine the relevant parameters in the DBSCAN clustering algorithm by analyzing the kth nearest neighbor distance data curve, and perform DBSCAN clustering operation on the input vector to remove the noise in the clustering result from the input data;

[0072] Construct a neural network, use the cluster center as the mean parameter of the Gaussian function corresponding to the network hidden layer node, train the network, and use the trained neural network to predict the risk level of the current time node.

[0073] In a specific implementation, the specific steps of the early warning are: D = d 基础 +β*d 预测 , D is the risk threshold of the current time node, d 基础 is the risk base value, β is an adjustment factor, d 预测 It is a risk value calculated by the neural network model. If the risk threshold of the current time node exceeds the preset threshold, the water level will be artificially lowered or the sunlight will be blocked. At the same time, the feeding time will be adjusted according to the changes in the weather to avoid losses in pond aquaculture.

[0074] In a specific implementation, the steps of the k-nearest neighbor method used to obtain the DBSCAN clustering hyperparameters are: using the k-nearest neighbor algorithm idea, obtain the L2 distance of the remaining input vectors to the current vector, and sort them from near to far to obtain the sequence dist1, dist2, dist3, dist4, ..., dist n-1 , where dist4 becomes the 4-neighbor distance, and the 4-neighbor distances of all input vectors are summarized to generate a curve graph with the vertical axis being the distance and the horizontal axis being the number of vectors corresponding to the distance. By analyzing the curve, it is found that the curve suddenly becomes flat after being steep to a certain neighbor distance, and it is easy to know that the points greater than the neighbor distance are noise, and the 4-neighbor distance at the turning point is used as the distance parameter value of DBSCAN. The present invention combines k-nearest neighbor, DBSCAN clustering algorithm, and neural network, and quickly determines network parameters through preliminary data analysis, thereby realizing rapid prediction of risk level.

[0075] In a specific embodiment, an automatic feeding system for aquaculture, such as Figure 2 As shown, including:

[0076] Collection module: used to collect basic information and breeding information of each breeding unit;

[0077] Scheme identification module: used to cluster the farming information using an algorithm, assign a feeding scheme label to each feeding operation, and identify the feeding scheme of the farming unit based on multiple feeding scheme labels;

[0078] Automatic feeding module: used for the system to automatically feed the breeding unit according to the feeding plan and basic information of the breeding unit;

[0079] Early warning module: It is used to construct a meteorological network diagram based on the current aquaculture meteorological information and meteorological forecast information, and use the trained neural network to analyze the meteorological network diagram to determine the risk level of this feeding and issue an early warning.

[0080] The present invention proposes an automatic feeding system for aquaculture, which automatically calculates the feeding quality, feeding time and frequency, and monitors the feeding process, including the amount of feed and the throwing distance, to achieve automated and digital management of the breeding process, and further upload the monitoring data to the AI ​​cloud server for data storage and recording. This eliminates waste, reduces costs for farmers, reduces feed sinking in ponds, and alleviates water pollution.

[0081] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0082] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An automatic feeding method for aquaculture, characterized in that: include: Collect basic information and breeding information of each breeding unit; An algorithm is used to cluster the farming information, a feeding scheme label is assigned to each feeding operation, and the feeding scheme of the farming unit is identified based on multiple feeding scheme labels; Automatically feed the breeding unit according to the feeding plan and basic information of the breeding unit; A meteorological network diagram is constructed based on the current aquaculture meteorological information and meteorological forecast information. The trained neural network is used to analyze the meteorological network diagram to determine the risk level of this feeding and issue an early warning.

2. The automatic feeding method for aquaculture according to claim 1, characterized in that: The basic information of each breeding unit includes: the number, location, and activation status of the breeding unit.

3. The automatic feeding method for aquaculture according to claim 1, characterized in that: The breeding information includes: fish species information, input information, feeding water quality, breeding species, fish species specifications and quantity, feed information, and breeding process information, among which, feeding water quality includes: water temperature, dissolved oxygen, pH value, ammonia nitrogen and nitrite value, feed information includes: nutritional indicators, specifications, varieties and brands, and breeding process information includes: growth rate, death and disease information.

4. The automatic feeding method for aquaculture according to claim 3, characterized in that: The method also includes optimizing parameters of the breeding information, performing regression analysis on multi-parameter continuous values ​​using a ReliefF algorithm, and obtaining parameter types with a greater correlation with automatic feeding through feature weight screening.

5. The automatic feeding method for aquaculture according to claim 4, characterized in that: The multi-parameter continuous value regression analysis using the ReliefF algorithm includes: The sensor parameters of the time series data are screened for relevance using the ReliefF algorithm. The larger the characteristic weight of the parameter, the greater the influence factor of the parameter on automatic feeding, and the smaller the weight, the smaller the influence factor of its relevance. Determine the total data sample set E and feature set G, and initialize them according to the sampling number n and the selected m nearest neighbor samples; Randomly select a sample H from the training sample set E and find the m nearest neighbor samples R of H from the same sample set of H. i , i = 1, 2, 3, ..., m, and then find m nearest neighbor samples G from each sample set that is different from H i (D) and perform k repeated sampling; Calculate the weight matrix O(T) of each parameter: Among them, p(D) represents the probability of category D appearing; The weights of all sensor feature sets G are calculated in turn to obtain the parameter weight matrix O, and then reordered according to the weight size.

6. The automatic feeding method for aquaculture according to claim 1, characterized in that: The algorithm is used to cluster the farming information, a feeding scheme label is assigned to each feeding operation, and the feeding scheme of the farming unit is identified according to multiple feeding scheme labels, including: Select the DBSCAN algorithm, the algorithm parameters are the neighborhood radius of a point, the minimum number of points in a cluster, and the minimum number of feeding operations required to form an independent feeding plan label; Get the data set obtained by re-sorting the weights, traverse each point in the data set for a single feeding operation, query all points in the field of the selected point feeding operation for clustering: After obtaining the clustering results, the feeding quantity and feeding frequency of a cluster are compared with the corresponding set thresholds. If they are greater than the threshold, a label of increasing the amount or feeding frequency is added, otherwise the feeding plan is directly output.

7. The automatic feeding method for aquaculture according to claim 6, characterized in that: The query selected point feeding operation clusters all points in the field, including: if the number of points of all points in the field, that is, the number of feeding operations, is greater than or equal to the minimum number of points, mark the selected point as a core point and create a new cluster; add all points in the selected point and all points in the field to the cluster; if the number of all points in the field is less than the minimum number of points, any point is a boundary point; for each boundary point, find the nearest core point and assign it to the corresponding cluster; points in the data set that are not assigned to any cluster are marked as noise points.

8. The automatic feeding method for aquaculture according to claim 1, characterized in that: The automatic feeding of the breeding unit according to the identified feeding plan and basic information of the breeding unit includes: parsing the feeding plan, obtaining the feeding type, feeding quantity, and feeding frequency; if the feeding plan is accompanied by an increase in quantity or feeding frequency tag, the feeding time is calculated according to the feeding frequency, and the feeding process is monitored, including the feeding quantity and the throwing distance.

9. The automatic feeding method for aquaculture according to claim 1, characterized in that: The method of constructing a meteorological network diagram based on the current aquaculture meteorological information and meteorological forecast information includes: obtaining meteorological information of the current aquaculture area according to sensors, obtaining meteorological forecast information of the current aquaculture area according to the network, obtaining meteorological historical information of the current aquaculture area, correcting the meteorological forecast information according to the meteorological historical information, and constructing a meteorological network diagram based on time nodes according to the corrected meteorological forecast information and the meteorological information of the current aquaculture area.

10. An automatic feeding system for aquaculture, characterized in that: include: Collection module: used to collect basic information and breeding information of each breeding unit; Scheme identification module: used to cluster the farming information using an algorithm, assign a feeding scheme label to each feeding operation, and identify the feeding scheme of the farming unit based on multiple feeding scheme labels; Automatic feeding module: used for the system to automatically feed the breeding unit according to the feeding plan and basic information of the breeding unit; Early warning module: It is used to construct a meteorological network diagram based on the current aquaculture meteorological information and meteorological forecast information, and use the trained neural network to analyze the meteorological network diagram to determine the risk level of this feeding and issue an early warning.

Images