A method and system for predicting feeding amount for breeding based on artificial intelligence

Abstract

The present invention belongs to the technical field of breeding management, and discloses a breeding feeding amount prediction method and system based on artificial intelligence. The method comprises the following steps: constructing an artificial intelligence model based on a cloud data center; collecting real-time breeding basic data of the current breeding species based on a data acquisition device, and uploading it to a cloud data center; adjusting information based on the real-time breeding basic data using a breeding feeding knowledge graph based on a cloud data center; using a breeding feeding amount prediction model to predict breeding feeding amount based on the real-time breeding basic data and the breeding feeding knowledge graph after information adjustment; using a breeding feeding adjustment model to adjust breeding feeding based on the real-time breeding basic data after the feeding amount prediction; sending the real-time breeding basic data after the feeding adjustment to the corresponding breeding feeding management device. The present invention solves the problems of lack of scientificity and accuracy, high breeding cost and low intelligence in the prior art.

Description

Technical Field

[0001] The present invention belongs to the technical field of aquaculture management, and in particular relates to an artificial intelligence-based aquaculture feeding amount prediction method and system. Background Art

[0002] Aquaculture management refers to the application of modern science and technology and methods to scientifically, standardizedly and meticulously manage the aquaculture industry in order to achieve the goals of high efficiency, ecological and sustainable development. The prediction of aquaculture feeding amount is an important part of aquaculture management. It is not only related to the economic benefits of farmers, but also to animal welfare, environmental protection and aquaculture survival rate. By accurately predicting the feeding amount, feed waste can be avoided, feed can be effectively used, and production costs can be reduced.

[0003] The existing aquaculture feeding amount prediction technology has the following defects:

[0004] 1) The determination of feeding amount in breeding mainly depends on the personal experience and intuition of the breeders. This method lacks scientificity and accuracy, and it is easy to lead to unreasonable feeding amount due to personal differences.

[0005] 2) Due to the lack of precise feeding guidance, farms often have overfeed or underfeed, resulting in low feed utilization and increased breeding costs.

[0006] 3) It is difficult to adapt to the needs of large-scale farming and the level of intelligence is low, which limits the further development and efficiency improvement of the farming industry. Summary of the invention

[0007] In order to solve the problems of lack of scientificity and accuracy, high breeding cost and low intelligence in the prior art, the present invention aims to provide a breeding feeding amount prediction method and system based on artificial intelligence.

[0008] The technical solution adopted by the present invention is:

[0009] A method for predicting feeding amount for breeding based on artificial intelligence comprises the following steps:

[0010] Based on the cloud data center, build a knowledge graph of aquaculture feeding for current aquaculture species, a prediction model for aquaculture feeding amount, and a model for aquaculture feeding adjustment;

[0011] Based on the data collection device, collect the real-time basic breeding data of the current breeding species, and upload the real-time basic breeding data to the cloud data center;

[0012] Based on the cloud data center, according to the real-time basic aquaculture data, the knowledge graph of aquaculture feeding is used to adjust the information and obtain the real-time basic aquaculture data after the information adjustment;

[0013] This includes extracting real-time breeding precautions from real-time breeding information data in real-time breeding basic data based on a cloud data center, and using a breeding and feeding knowledge graph to adjust the real-time breeding precautions to obtain real-time breeding precautions after information adjustment, and then replacing the real-time breeding precautions in the real-time breeding basic data according to the real-time breeding precautions after information adjustment to obtain real-time breeding basic data after information adjustment;

[0014] Based on the cloud data center, according to the real-time aquaculture basic data and aquaculture feeding knowledge graph after information adjustment, the aquaculture feeding amount prediction model is used to predict the aquaculture feeding amount, and the real-time aquaculture basic data after the feeding amount prediction is obtained;

[0015] The method includes using a breeding feeding amount prediction model based on a cloud data center to extract real-time graph features of a breeding feeding knowledge graph and M real-time key features of real-time breeding basic data after information adjustment, wherein M is the total number of key features, and performing feature splicing on the real-time graph features and the M real-time key features to obtain real-time splicing features, and performing breeding feeding amount prediction based on the real-time splicing features to obtain real-time breeding feeding amount prediction results, and then adding the real-time breeding feeding amount prediction results to the real-time breeding basic data after information adjustment to obtain real-time breeding basic data after feeding amount prediction;

[0016] Based on the cloud data center, the real-time basic aquaculture data after feeding amount prediction is used, and the aquaculture feeding adjustment model is used to adjust the aquaculture feeding to obtain the real-time basic aquaculture data after feeding adjustment;

[0017] Based on the cloud data center, the real-time basic breeding data after feeding adjustment is sent to the corresponding breeding feeding management device, and the next real-time basic breeding data is collected.

[0018] Furthermore, based on the cloud data center, a knowledge graph of aquaculture feeding, a prediction model of aquaculture feeding amount, and an aquaculture feeding adjustment model of the current aquaculture species are constructed, including the following steps:

[0019] Based on the cloud data center, some aquaculture knowledge of current aquaculture species is collected, and based on some aquaculture knowledge, a pre-trained natural language processing model is used to build an aquaculture feeding knowledge graph;

[0020] Collect some historical aquaculture basic data of the current aquaculture species, and use deep learning algorithms to build an aquaculture feeding amount prediction model based on some historical aquaculture basic data and aquaculture feeding knowledge graph, and generate some historical aquaculture feeding amount prediction results;

[0021] Based on several historical aquaculture feeding amount prediction results, a reinforcement learning algorithm is used to build an aquaculture feeding adjustment model, and several historical aquaculture feeding adjustment experiences are generated.

[0022] Further, the natural language processing model is constructed based on the BERT-LSTM-CRF algorithm, and the natural language processing model includes a first input layer, a BERT layer, a LSTM layer, a CRF layer and a first output layer connected in sequence;

[0023] The aquaculture feeding amount prediction model is constructed based on a multimodal fusion algorithm, and the aquaculture feeding amount prediction model includes a second input layer, a key feature extraction module constructed based on an RF algorithm, a graph feature extraction module constructed based on a GCN algorithm, an attention weight module constructed based on an Attention mechanism, an aquaculture feeding amount prediction module constructed based on an MLP algorithm, and a second output layer. The second input layer is connected to the key feature extraction module and the graph feature extraction module respectively, and the key feature extraction module and the graph feature extraction module are both connected to the attention weight module. The attention weight module, the aquaculture feeding amount prediction module, and the second output layer are connected in sequence;

[0024] The aquaculture feeding adjustment model is built based on the DQN algorithm.

[0025] Furthermore, the real-time breeding basic data includes real-time breeding information data of the current breeding species, real-time breeding monitoring data and real-time breeding feeding plan.

[0026] Furthermore, the real-time breeding information data includes real-time species information of the current breeding species, real-time breeding scale information, real-time breeding environment information and real-time breeding precautions;

[0027] Real-time aquaculture monitoring data includes real-time environmental monitoring data, real-time growth monitoring data and real-time feeding monitoring data;

[0028] The real-time aquaculture feeding plan includes the real-time aquaculture single feeding amount, the real-time aquaculture total feeding amount, the real-time aquaculture feeding method and the real-time aquaculture feeding cycle.

[0029] Furthermore, based on the cloud data center, according to the real-time basic aquaculture data, the aquaculture feeding knowledge graph is used to adjust the information to obtain the real-time basic aquaculture data after the information adjustment, including the following steps:

[0030] Furthermore, based on the cloud data center, according to the real-time basic aquaculture data after the feeding amount is predicted, the aquaculture feeding adjustment model is used to adjust the aquaculture feeding to obtain the real-time basic aquaculture data after the feeding adjustment, including the following steps:

[0031] Based on the cloud data center, according to the real-time basic data of aquaculture after the feeding amount is predicted, a search is performed in several historical aquaculture feeding adjustment experiences to obtain several matching historical aquaculture feeding adjustment experiences;

[0032] According to the real-time aquaculture basic data after the feeding amount prediction and several matching historical aquaculture feeding adjustment experiences, the model parameters of the aquaculture feeding adjustment model are updated to obtain an updated aquaculture feeding adjustment model;

[0033] Use the updated aquaculture feeding adjustment model to adjust aquaculture feeding and obtain real-time aquaculture feeding adjustment strategy;

[0034] According to the real-time aquaculture feeding adjustment strategy, the real-time aquaculture feeding plan in the real-time aquaculture basic data after the feeding amount prediction is adjusted to obtain an updated real-time aquaculture feeding plan;

[0035] According to the updated real-time aquaculture feeding plan, the corresponding real-time aquaculture feeding plan is replaced with information to obtain the real-time aquaculture basic data after feeding adjustment.

[0036] Furthermore, the real-time aquaculture feeding amount prediction result includes the real-time aquaculture single feeding amount prediction value, the real-time aquaculture total feeding amount prediction value and the real-time aquaculture feeding amount analysis prediction value;

[0037] The updated real-time aquaculture feeding plan includes the real-time aquaculture single feeding amount prediction value, the real-time aquaculture total feeding amount prediction value, the adjusted real-time aquaculture feeding method and the adjusted real-time aquaculture feeding cycle.

[0038] An artificial intelligence-based aquaculture feeding amount prediction system is used to implement an aquaculture feeding amount prediction method. The system includes a cloud data center, a plurality of data acquisition devices, and a plurality of aquaculture feeding management devices. The cloud data center is respectively connected to the plurality of data acquisition devices and the plurality of aquaculture feeding management devices in communication.

[0039] The cloud data center is provided with a model building unit, an information adjustment unit, a breeding feeding amount prediction unit, a breeding feeding adjustment unit and a data sending unit which are connected in sequence.

[0040] The beneficial effects of the present invention are:

[0041] The invention discloses a method and system for predicting aquaculture feeding amount based on artificial intelligence. The method and system combine artificial intelligence technology to predict aquaculture feeding amount, significantly reduce the dependence on manual experience, and improve the scientificity and accuracy of feeding decision-making. The method collects basic aquaculture data in real time, and combines aquaculture feeding knowledge graph, aquaculture feeding amount prediction model and aquaculture feeding adjustment model to dynamically monitor changes in aquaculture environment and adjust feeding plans in real time to adapt to changing environmental conditions, improve feed utilization, and reduce feed cost input. The method formulates personalized aquaculture feeding plans according to the biological characteristics, growth requirements and aquaculture advantages of different aquaculture species, thereby maximizing growth potential and reducing resource waste. Based on a cloud data center, the centralized management and analysis of aquaculture data is realized to adapt to the needs of large-scale aquaculture, and the automation and intelligence level of aquaculture management is improved through data-driven decision support.

[0042] Other beneficial effects of the present invention will be further described in the specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flowchart of the method for predicting feeding amount for breeding based on artificial intelligence in the present invention.

[0044] Figure 2 It is a structural block diagram of the breeding feeding amount prediction model in the present invention.

[0045] Figure 3 It is a structural block diagram of the artificial intelligence-based aquaculture feeding amount prediction system in the present invention. DETAILED DESCRIPTION

[0046] The present invention will be further explained below in conjunction with the accompanying drawings and specific embodiments.

[0047] Embodiment 1:

[0048] like Figure 1 As shown, this embodiment provides a method for predicting feeding amount for breeding based on artificial intelligence, comprising the following steps:

[0049] S1: Based on the cloud data center, construct the aquaculture feeding knowledge graph, aquaculture feeding amount prediction model and aquaculture feeding adjustment model of the current aquaculture species, including the following steps:

[0050] S1-1: Based on the cloud data center, some aquaculture knowledge of the current aquaculture species is collected, and based on some aquaculture knowledge, a pre-trained natural language processing model is used to construct an aquaculture feeding knowledge graph, including the following steps:

[0051] S1-1-1: Based on the cloud data center, some aquaculture knowledge of the current aquaculture species is collected, and some aquaculture knowledge is preprocessed to obtain some preprocessed aquaculture knowledge;

[0052] In this embodiment, the breeding species are aquaculture species, and the breeding knowledge includes breeding environment precautions, feeding precautions, etc. of aquaculture species;

[0053] S1-1-2: Use a pre-trained natural language processing model based on the Bidirectional Encoder Representations from Transformers (BERT)-Long Short-Term Memory (LSTM)-Conditional Random Fields (CRF) algorithm to extract several named entities of the pre-processed farming knowledge;

[0054] The natural language processing model includes a first input layer, a BERT layer, an LSTM layer, a CRF layer and a first output layer connected in sequence. The pre-processed aquaculture knowledge input into the natural language processing model is subjected to text cleaning (removing useless characters and unifying the format), word segmentation (segmenting the text into words or sub-word units) and annotation (annotating entity category labels for each word or sub-word unit) to obtain processed knowledge data. The first input layer is used to receive the processed knowledge data. The BERT layer is used to extract deep context features of the processed knowledge data, perform character embedding, and obtain a character-level vector sequence. The LSTM layer is used to learn long-distance dependencies in the character-level vector sequence and extract a global semantic feature sequence. The CRF layer is used to perform dependency processing on each named entity in the character-level vector sequence according to the global semantic feature sequence, and named entity labels are added to predict the optimal label sequence to obtain the corresponding named entity.

[0055] S1-2: Collect some historical aquaculture basic data of the current aquaculture species, and use the multimodal fusion algorithm to build an aquaculture feeding amount prediction model based on some historical aquaculture basic data and aquaculture feeding knowledge graph, and generate some historical aquaculture feeding amount prediction results; the architecture of the multimodal fusion algorithm is Random Forest (RF)-Graph Convolutional Network (GCN)-Attention-Multi-Layer Perceptron (MLP);

[0056] like Figure 2As shown, the aquaculture feeding amount prediction model includes a second input layer, a key feature extraction module based on the RF algorithm, a graph feature extraction module based on the GCN algorithm, an attention weight module based on the Attention mechanism, an aquaculture feeding amount prediction module based on the MLP algorithm, and a second output layer. The second input layer is connected to the key feature extraction module and the graph feature extraction module respectively, and the key feature extraction module and the graph feature extraction module are both connected to the attention weight module. The attention weight module, the aquaculture feeding amount prediction module, and the second output layer are connected in sequence;

[0057] The second input layer receives the basic aquaculture data. The key feature extraction module uses the internal Classification And Regression Tree (CART) to screen the key features of the input basic aquaculture data and extract several key features related to the data prediction label. The graph feature extraction module propagates features on the knowledge graph through operations similar to convolution, extracts the node features of named entities and the edge features between entities, and forms the graph features of the knowledge graph. The attention weight module uses the preset attention weights to splice several key features and graph features, integrate the scattered features, and enhance the influence of features on label prediction. The aquaculture feeding amount prediction module adopts the form of a fully connected network to learn the deep relationship between the spliced ​​features and the data prediction labels, and realizes efficient and accurate aquaculture feeding amount prediction.

[0058] S1-3: Based on several historical aquaculture feeding amount prediction results, the Deep Q Network (DQN) algorithm is used to build an aquaculture feeding adjustment model and generate several historical aquaculture feeding adjustment experiences, including the following steps:

[0059] S1-3-1: Generate the feeding adjustment problem as the simulation environment of the DQN algorithm, and build the intelligent agent and experience replay pool;

[0060] S1-3-2: According to each aquaculture feeding amount prediction state corresponding to the historical aquaculture feeding amount prediction results, define the state space of the DQN algorithm, and the parameters of the state space correspond to each aquaculture feeding amount prediction state; the state parameters in the state space include aquaculture single feeding amount prediction value, aquaculture total feeding amount prediction value and aquaculture feeding amount analysis prediction value, the aquaculture feeding amount prediction state corresponding to the aquaculture single feeding amount prediction value includes insufficient aquaculture single feeding amount, normal aquaculture single feeding amount and aquaculture single feeding amount exceeding the standard, the aquaculture feeding amount prediction state corresponding to the aquaculture total feeding amount prediction value includes insufficient total aquaculture feeding amount, normal total aquaculture feeding amount and aquaculture total feeding amount exceeding the standard, the aquaculture feeding amount prediction state corresponding to the aquaculture feeding amount analysis prediction value includes incorrect aquaculture feeding method, correct aquaculture feeding method, real-time aquaculture feeding cycle is too short, real-time aquaculture feeding cycle is normal and real-time aquaculture feeding cycle is too long, etc.;

[0061] S1-3-3: Define the action space of the DQN algorithm according to the action that needs to be output according to the aquaculture feeding adjustment; the action parameters of the action space include adjusting the aquaculture feeding method and adjusting the aquaculture feeding cycle. The possible actions for adjusting the aquaculture feeding method include adjusting the aquaculture feeding method to daytime feeding, nighttime feeding, centralized feeding, dispersed feeding, etc. The possible actions for adjusting the aquaculture feeding cycle include shortening the feeding cycle, delaying the feeding cycle, reducing the feeding frequency, increasing the feeding frequency, etc.;

[0062] S1-3-4: Based on the possible impact of each action in the action space, define the reward function of the DQN algorithm to evaluate the quality or impact of the action;

[0063] S1-3-5: Construct the input layer, several hidden layers, and output layer of the deep Q network, connect the input layer to the state space, and connect the output layer to the action space;

[0064] S1-3-6: Based on the state space, action space and reward function, according to several historical aquaculture feeding amount prediction results, the deep Q network and intelligent agent are optimized and trained to build an aquaculture feeding adjustment model, and several historical aquaculture feeding adjustment experiences are generated;

[0065] S1-3-7: Store some historical aquaculture feeding adjustment experiences into the experience playback pool of the aquaculture feeding adjustment model;

[0066] S2: Based on the data collection device, collect the real-time basic breeding data of the current breeding species, and upload the real-time basic breeding data to the cloud data center;

[0067] Real-time aquaculture basic data includes real-time aquaculture information data of the current aquaculture species, real-time aquaculture monitoring data and real-time aquaculture feeding plan;

[0068] Real-time aquaculture information data includes real-time species information of current aquaculture species, real-time aquaculture scale information, real-time aquaculture environment information and real-time aquaculture precautions;

[0069] In this embodiment, the current breeding species is an aquaculture species, and the real-time species information includes: silver carp, freshwater fish, diet: plankton, soy milk, rotifer, growth cycle: 34-36 weeks old; reproduction: 243 days;

[0070] Real-time breeding scale information: 60,000 fry, 50,000 adult fish, 1,000 mu of breeding ponds;

[0071] Real-time aquaculture environment information: water temperature XX°, oxygen content: YYY, water depth 1.8 meters;

[0072] Note: Inject some fertilizer water into the feed;

[0073] Real-time aquaculture monitoring data includes real-time environmental monitoring data, real-time growth monitoring data and real-time feeding monitoring data;

[0074] Real-time environmental monitoring data includes real-time environmental temperature data, real-time environmental humidity data, real-time environmental sunshine data, real-time environmental oxygen content data, etc.

[0075] Real-time growth monitoring data includes real-time fish density data, real-time adult fish length data, real-time weight gain data, etc.

[0076] Real-time feeding monitoring data includes real-time fed feed data, real-time feeding times, real-time feeding areas, etc.

[0077] The real-time aquaculture feeding plan includes the real-time aquaculture single feeding amount, the real-time aquaculture total feeding amount, the real-time aquaculture feeding method and the real-time aquaculture feeding cycle;

[0078] The single feeding amount of real-time breeding is: A kg, the total feeding amount of real-time breeding is B kg, the feeding method of real-time breeding is: centralized feeding, feeding during the day, and the feeding cycle of real-time breeding is: once every 8 hours;

[0079] S3: Based on the cloud data center, according to the real-time basic aquaculture data, the aquaculture feeding knowledge graph is used to adjust the information to obtain the real-time basic aquaculture data after the information adjustment, including the following steps:

[0080] S3-1: Based on the cloud data center, extract the real-time breeding information data from the real-time breeding basic data to pay attention to the real-time breeding;

[0081] In this embodiment, the precautions are as follows: some fertilizer water is injected into the feed;

[0082] S3-2: Use the aquaculture and feeding knowledge graph to adjust the information of real-time aquaculture precautions, and obtain the real-time aquaculture precautions after the information adjustment;

[0083] Real-time breeding precautions after information adjustment: inject some fertilizer water into the feed, stock 100-150 adult fish per mu, with the ratio of female fish to male fish being 1:1.25; the time for feeding is from 11pm to 12pm;

[0084] S3-3: According to the real-time breeding precautions after information adjustment, the real-time breeding precautions in the real-time breeding basic data are replaced to obtain the real-time breeding basic data after information adjustment;

[0085] S4: Based on the cloud data center, according to the real-time aquaculture basic data and aquaculture feeding knowledge graph after information adjustment, the aquaculture feeding amount prediction model is used to predict the aquaculture feeding amount, and the real-time aquaculture basic data after the feeding amount prediction is obtained, including the following steps:

[0086] S4-1: Based on the cloud data center, using the aquaculture feeding amount prediction model, extract the real-time graph features of the aquaculture feeding knowledge graph and M real-time key features of the real-time aquaculture basic data after information adjustment, where M is the total number of key features, including the following steps:

[0087] S4-1-1: Use the graph feature extraction module of the aquaculture feeding amount prediction model to extract the real-time graph features of the aquaculture feeding knowledge graph;

[0088] S4-1-2: Use the RF structure trained in the key feature extraction module to extract the feature contribution of M candidate features in the real-time aquaculture basic data after information adjustment;

[0089] The formula is:

[0090]

[0091] In the formula, For the Feature contribution of candidate features; For the The candidate features are in the random forest The feature contribution of each tree; is the CART tree indicator; is an alternative characteristic indicator; is the total number of CARTs;

[0092]

[0093] In the formula, CART tree node for random forest m ,node and nodes The Gini index of CART tree node m Medium Category The proportion of is the total number of categories; m , , is the node indicator; is the category indicator;

[0094] S4-1-3: normalizing the feature contributions of the M candidate features to obtain a number of corresponding normalized feature contributions;

[0095] The formula is:

[0096]

[0097] In the formula, It is the contribution of the feature after normalization; J is the total number of candidate features;

[0098] S4-1-4: Generate feature selection standard values ​​for several candidate features based on the normalized feature contribution;

[0099] The formula is:

[0100]

[0101] In the formula, For the Feature selection criterion values ​​for candidate features; For the The normalized feature contribution of the candidate features; is an alternative characteristic indicator;

[0102] S4-1-5: According to the feature selection standard value, the candidate features are sorted in descending order, and the first M candidate features are selected as real-time key features to obtain M real-time key features;

[0103] S4-2: According to the attention weight preset by the attention weight module, the real-time graph feature and the M real-time key features are spliced ​​to obtain the real-time spliced ​​feature, and the aquaculture feeding amount prediction module is used to predict the aquaculture feeding amount according to the real-time spliced ​​feature to obtain the real-time aquaculture feeding amount prediction result;

[0104] The real-time aquaculture feeding amount prediction results include the real-time aquaculture single feeding amount prediction value, the real-time aquaculture total feeding amount prediction value and the real-time aquaculture feeding amount analysis prediction value;

[0105] In this embodiment, the real-time breeding single feeding amount prediction value is: A+kg, the real-time breeding total feeding amount prediction value is B+kg, and the real-time breeding feeding amount prediction state corresponding to the real-time breeding feeding amount analysis prediction value includes: insufficient breeding single feeding amount, insufficient breeding total feeding amount, wrong breeding feeding method, and too long real-time breeding feeding cycle;

[0106] S4-3: adding the real-time breeding feeding amount prediction result to the real-time breeding basic data after information adjustment to obtain the real-time breeding basic data after feeding amount prediction;

[0107] S5: Based on the cloud data center, according to the real-time basic aquaculture data after the feeding amount is predicted, the aquaculture feeding adjustment model is used to adjust the aquaculture feeding to obtain the real-time basic aquaculture data after the feeding adjustment, including the following steps:

[0108] S5-1: Based on the cloud data center, according to the real-time basic data of aquaculture after the feeding amount is predicted, a search is performed in several historical aquaculture feeding adjustment experiences to obtain several matching historical aquaculture feeding adjustment experiences;

[0109] S5-2: According to the real-time aquaculture basic data after the feeding amount prediction and some matching historical aquaculture feeding adjustment experience, the model parameters of the aquaculture feeding adjustment model are updated to obtain an updated aquaculture feeding adjustment model, including the following steps:

[0110] S5-2-1: Update the state space of the aquaculture feeding adjustment model according to the real-time aquaculture single feeding amount prediction value and the real-time aquaculture total feeding amount prediction value in the real-time aquaculture basic data after the feeding amount prediction, and the real-time aquaculture feeding amount analysis prediction value, and obtain the updated state space ,in, For the updated Status value, is the state indicator, is the total number of state space dimensions;

[0111] S5-2-2: Update the action space of the aquaculture feeding adjustment model based on some matching historical aquaculture feeding adjustment experiences to obtain an updated action space ,in, For the updated Action value, is the action indicator, is the total number of action space dimensions;

[0112] S5-2-3: According to the updated state space and the updated action space, an updated aquaculture feeding adjustment model is obtained;

[0113] S5-3: Use the updated aquaculture feeding adjustment model to adjust aquaculture feeding and obtain a real-time aquaculture feeding adjustment strategy, including the following steps:

[0114] S5-3-1: Input the updated state space into the deep Q network of the aquaculture feeding adjustment model, and use the deep Q network to generate the Q value of each possible action in the updated action space based on the updated state space;

[0115] S5-3-2: Use the agent to obtain the reward value of each possible action in the updated action space according to the reward function, and update the Q value of the possible action according to the reward value to obtain the updated Q value of each possible action;

[0116] The formula is:

[0117]

[0118] In the formula, The updated status value and the updated action value The corresponding updated Q value; Status value and action value The corresponding Q value of the possible action; is the learning rate; is the Q value of the highest possible action; It is a comprehensive indicator; To update the parameters;

[0119] S5-3-3: Repeat the above Q-value updating steps until the number of iterations reaches the iteration number threshold, and use the greedy strategy to select the possible action with the highest updated Q-value in each iteration as the execution action;

[0120] S5-3-4: Use the intelligent agent to integrate several execution actions to obtain real-time aquaculture feeding adjustment strategy;

[0121] The real-time feeding adjustment strategy includes adjusting the feeding method to 11:00-12:00 pm and dispersed feeding, and adjusting the feeding cycle to once every 5 hours;

[0122] S5-4: According to the real-time aquaculture feeding adjustment strategy, the real-time aquaculture feeding plan in the real-time aquaculture basic data after the feeding amount prediction is adjusted to obtain an updated real-time aquaculture feeding plan;

[0123] The updated real-time aquaculture feeding plan includes the real-time aquaculture single feeding amount prediction value: A+kg, the real-time aquaculture total feeding amount prediction value: B+kg, the adjusted real-time aquaculture feeding method: 11pm-12pm and dispersed feeding, and the adjusted real-time aquaculture feeding cycle: once every 5 hours;

[0124] S5-5: According to the updated real-time aquaculture feeding plan, the corresponding real-time aquaculture feeding plan is replaced with information to obtain the real-time aquaculture basic data after feeding adjustment;

[0125] S6: Based on the cloud data center, the real-time basic breeding data after feeding adjustment is sent to the corresponding breeding feeding management device, and the next real-time basic breeding data is continuously collected.

[0126] Embodiment 2:

[0127] like Figure 3 As shown, this embodiment provides an artificial intelligence-based aquaculture feeding amount prediction system for implementing an aquaculture feeding amount prediction method. The system includes a cloud data center, a plurality of data acquisition devices, and a plurality of aquaculture feeding management devices. The cloud data center is respectively communicated with the plurality of data acquisition devices and the plurality of aquaculture feeding management devices;

[0128] A data collection device is used to collect real-time basic aquaculture data of the current aquaculture species and upload the real-time basic aquaculture data to a cloud data center;

[0129] Aquaculture feeding management device, used to receive and visualize real-time basic aquaculture data after feeding adjustment;

[0130] The cloud data center is provided with a model building unit, an information adjustment unit, a breeding feeding amount prediction unit, a breeding feeding adjustment unit and a data sending unit which are connected in sequence;

[0131] A model building unit, used to build a knowledge graph of aquaculture and feeding of the current aquaculture species, a prediction model for aquaculture and feeding amount, and a model for aquaculture and feeding adjustment;

[0132] An information adjustment unit is used to adjust information according to real-time basic aquaculture data using a knowledge graph of aquaculture and feeding, and obtain real-time basic aquaculture data after information adjustment;

[0133] The aquaculture feeding amount prediction unit is used to predict the aquaculture feeding amount according to the real-time aquaculture basic data and the aquaculture feeding knowledge graph after information adjustment, and use the aquaculture feeding amount prediction model to predict the aquaculture feeding amount, and obtain the real-time aquaculture basic data after the feeding amount prediction;

[0134] The aquaculture feeding adjustment unit is used to adjust the aquaculture feeding according to the real-time aquaculture basic data after the feeding amount is predicted, and use the aquaculture feeding adjustment model to obtain the real-time aquaculture basic data after the feeding adjustment;

[0135] The data sending unit is used to send the real-time basic breeding data after feeding adjustment to the corresponding breeding feeding management device.

[0136] The invention discloses a method and system for predicting aquaculture feeding amount based on artificial intelligence. The method and system combine artificial intelligence technology to predict aquaculture feeding amount, significantly reduce the dependence on manual experience, and improve the scientificity and accuracy of feeding decision-making. The method collects basic aquaculture data in real time, and combines aquaculture feeding knowledge graph, aquaculture feeding amount prediction model and aquaculture feeding adjustment model to dynamically monitor changes in aquaculture environment and adjust feeding plans in real time to adapt to changing environmental conditions, improve feed utilization, and reduce feed cost input. The method formulates personalized aquaculture feeding plans according to the biological characteristics, growth requirements and aquaculture advantages of different aquaculture species, thereby maximizing growth potential and reducing resource waste. Based on a cloud data center, the centralized management and analysis of aquaculture data is realized to adapt to the needs of large-scale aquaculture, and the automation and intelligence level of aquaculture management is improved through data-driven decision support.

[0137] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for predicting feeding amount for breeding based on artificial intelligence, characterized in that: The steps include: Based on the cloud data center, the knowledge graph of aquaculture feeding, the aquaculture feeding amount prediction model and the aquaculture feeding adjustment model of the current aquaculture species are constructed, including the following steps: Based on the cloud data center, some aquaculture knowledge of current aquaculture species is collected, and based on some aquaculture knowledge, a pre-trained natural language processing model is used to build an aquaculture feeding knowledge graph; Collect some historical aquaculture basic data of the current aquaculture species, and use deep learning algorithms to build an aquaculture feeding amount prediction model based on some historical aquaculture basic data and aquaculture feeding knowledge graph, and generate some historical aquaculture feeding amount prediction results; Based on several historical aquaculture feeding amount prediction results, a reinforcement learning algorithm is used to build an aquaculture feeding adjustment model and generate several historical aquaculture feeding adjustment experiences; Based on the data collection device, collect the real-time basic breeding data of the current breeding species, and upload the real-time basic breeding data to the cloud data center; The real-time breeding basic data includes real-time breeding information data of the current breeding species, real-time breeding monitoring data and real-time breeding feeding plan; The real-time breeding information data includes real-time species information of the current breeding species, real-time breeding scale information, real-time breeding environment information and real-time breeding precautions; The real-time breeding monitoring data includes real-time environmental monitoring data, real-time growth monitoring data and real-time feeding monitoring data; The real-time breeding and feeding scheme includes the real-time breeding single feeding amount, the real-time breeding total feeding amount, the real-time breeding feeding method and the real-time breeding feeding cycle; Based on the cloud data center, according to the real-time basic aquaculture data, the knowledge graph of aquaculture feeding is used to adjust the information and obtain the real-time basic aquaculture data after the information adjustment; Based on the cloud data center, according to the real-time aquaculture basic data and aquaculture feeding knowledge graph after information adjustment, the aquaculture feeding amount prediction model is used to predict the aquaculture feeding amount, and the real-time aquaculture basic data after the feeding amount prediction is obtained, including the following steps: Based on the cloud data center, the aquaculture feeding amount prediction model is used to extract the real-time graph features of the aquaculture feeding knowledge graph and M real-time key features of the real-time aquaculture basic data after information adjustment, where M is the total number of key features; Perform feature splicing on the real-time graph features and the M real-time key features to obtain real-time splicing features, and perform aquaculture feeding amount prediction based on the real-time splicing features to obtain a real-time aquaculture feeding amount prediction result; The real-time breeding feeding amount prediction result is added to the real-time breeding basic data after information adjustment to obtain the real-time breeding basic data after feeding amount prediction; Based on the cloud data center, the real-time basic aquaculture data after the feeding amount is predicted is used, and the aquaculture feeding adjustment model is used to adjust the aquaculture feeding to obtain the real-time basic aquaculture data after the feeding adjustment, including the following steps: Based on the cloud data center, according to the real-time basic data of aquaculture after the feeding amount is predicted, a search is performed in several historical aquaculture feeding adjustment experiences to obtain several matching historical aquaculture feeding adjustment experiences; According to the real-time aquaculture basic data after the feeding amount prediction and several matching historical aquaculture feeding adjustment experiences, the model parameters of the aquaculture feeding adjustment model are updated to obtain an updated aquaculture feeding adjustment model; Use the updated aquaculture feeding adjustment model to adjust aquaculture feeding and obtain real-time aquaculture feeding adjustment strategy; According to the real-time aquaculture feeding adjustment strategy, the real-time aquaculture feeding plan in the real-time aquaculture basic data after the feeding amount prediction is adjusted to obtain an updated real-time aquaculture feeding plan; According to the updated real-time aquaculture feeding plan, the corresponding real-time aquaculture feeding plan is replaced with information to obtain the real-time aquaculture basic data after feeding adjustment; Based on the cloud data center, the real-time basic breeding data after feeding adjustment is sent to the corresponding breeding feeding management device, and the next real-time basic breeding data is collected.

2. The method for predicting feeding amount for breeding based on artificial intelligence according to claim 1, characterized in that: The natural language processing model is constructed based on the BERT-LSTM-CRF algorithm, and the natural language processing model includes a first input layer, a BERT layer, a LSTM layer, a CRF layer and a first output layer connected in sequence; The aquaculture feeding amount prediction model is constructed based on a multimodal fusion algorithm, and the aquaculture feeding amount prediction model includes a second input layer, a key feature extraction module constructed based on an RF algorithm, a graph feature extraction module constructed based on a GCN algorithm, an attention weight module constructed based on an Attention mechanism, an aquaculture feeding amount prediction module constructed based on an MLP algorithm, and a second output layer, wherein the second input layer is respectively connected to the key feature extraction module and the graph feature extraction module, and the key feature extraction module and the graph feature extraction module are both connected to the attention weight module, and the attention weight module, the aquaculture feeding amount prediction module, and the second output layer are connected in sequence; The aquaculture feeding adjustment model is constructed based on the DQN algorithm.

3. The method for predicting feeding amount for breeding based on artificial intelligence according to claim 2, characterized in that: Based on the cloud data center, according to the real-time basic aquaculture data, the knowledge graph of aquaculture feeding is used to adjust the information, and the real-time basic aquaculture data after the information adjustment is obtained, including the following steps: Based on the cloud data center, extract the real-time breeding information data from the real-time breeding basic data to pay attention to the real-time breeding; Use the aquaculture and feeding knowledge graph to adjust the information of real-time aquaculture precautions, and obtain the real-time aquaculture precautions after the information adjustment; According to the real-time breeding precautions after information adjustment, the real-time breeding precautions in the real-time breeding basic data are replaced to obtain the real-time breeding basic data after information adjustment.

4. The method for predicting feeding amount for breeding based on artificial intelligence according to claim 3, characterized in that: The real-time aquaculture feeding amount prediction result includes the real-time aquaculture single feeding amount prediction value, the real-time aquaculture total feeding amount prediction value and the real-time aquaculture feeding amount analysis prediction value; The updated real-time aquaculture feeding plan includes a real-time aquaculture single feeding amount prediction value, a real-time aquaculture total feeding amount prediction value, an adjusted real-time aquaculture feeding method, and an adjusted real-time aquaculture feeding cycle.

5. An artificial intelligence-based aquaculture feeding amount prediction system, used to implement the aquaculture feeding amount prediction method according to any one of claims 1 to 4, characterized in that: The system includes a cloud data center, a plurality of data acquisition devices and a plurality of aquaculture and feeding management devices, wherein the cloud data center is respectively connected to the plurality of data acquisition devices and the plurality of aquaculture and feeding management devices for communication; The cloud data center is provided with a model building unit, an information adjustment unit, a breeding feeding amount prediction unit, a breeding feeding adjustment unit and a data sending unit which are connected in sequence.

Images