Wave energy control and allocation method and system for offshore aquaculture

Abstract

Disclosed in the present invention are a wave energy control and allocation method and system for offshore aquaculture, comprising: acquiring aquaculture cycles of sub-aquaculture zones of an offshore aquaculture zone, sorting remaining aquaculture cycles of the sub-aquaculture zones in an ascending order, and obtaining a plurality of working cycles on the basis of a sorting result; obtaining predicted wave energy production within a following working cycle by means of a preset neural network model; obtaining an importance coefficient value sorting result of the aquaculture zones by means of a preset RFE model; and on the basis of device types of aquaculture devices, aquaculture zones where the aquaculture devices are located, the predicted wave energy production, and the importance coefficient value sorting result, sequentially adjusting operating cycles and operating powers of first-type aquaculture devices, second-type aquaculture devices, and third-type aquaculture devices. In the present invention, by establishing a prediction model, the wave energy production within the following working cycle can be predicted in advance, and then operating plans of the aquaculture devices are arranged on the basis of a prediction result, thereby avoiding the waste or shortage of resources.

Description

Wave energy control distribution method and system for offshore aquaculture TECHNICAL FIELD

[0001] The present application relates to the technical field of wave energy data processing, in particular to a wave energy control distribution method and system for offshore aquaculture. BACKGROUND

[0002] In the offshore aquaculture environment, wave energy as a clean and renewable energy is widely used in the power supply of aquaculture equipment. However, the yield of wave energy is affected by many factors, including the height, speed, direction of sea waves and weather conditions. These factors change over time, resulting in fluctuations in the yield of wave energy. Therefore, in actual operation, if there is no effective management and control strategy, the following problems may occur:

[0003] First, resource waste: if the high yield of wave energy is not used in time, it will cause waste of resources; second, resource shortage: when the yield of wave energy is low, if the aquaculture equipment still runs in a fixed mode, it may lead to insufficient power supply, affecting the normal operation of aquaculture activities; third, low efficiency of equipment operation: due to the fluctuation of wave energy yield, if the equipment operation parameters are fixed, it may lead to low efficiency of equipment operation in some periods, and may exceed the needs in other periods. SUMMARY

[0004] The present application provides a wave energy control distribution method and system for offshore aquaculture, which realizes the effective use of wave energy resources, ensures the full use of these resources when the yield of wave energy is high, and reduces unnecessary energy consumption when the yield is low.

[0005] The first aspect of the present application provides a wave energy control distribution method for offshore aquaculture, comprising:

[0006] Obtaining the aquaculture period of each sub-aquaculture area in the offshore aquaculture area, sorting the remaining aquaculture periods of each sub-aquaculture area from small to large, and obtaining a plurality of working periods according to the sorting result;

[0007] The input layer of the preset neural network model is set to three dimensions, an LSTM layer and a GRU layer are selected to constitute the hidden layer of the preset neural network model; a plurality of first neurons and a second neuron are set in the output layer, each first neuron corresponds to a group of sensors deployed on the sea; the output of the second neuron is the sum of the outputs of all first neurons; each group of sensors is responsible for information collection of a marine sub-area;

[0008] convert historical marine data into a three-dimensional tensor; a first dimension of the three-dimensional tensor is a batch sample number, the batch sample number is equal to a number of groups of marine sensors deployed on the sea; a second dimension of the three-dimensional tensor is a time step, the time step is equal to a time span of the historical marine data; and a third dimension of the three-dimensional tensor is a feature number;

[0009] input the three-dimensional tensor into a preset neural network model for training;

[0010] obtain a predicted wave energy production of a next working period through the preset neural network model;

[0011] obtain a sorting result of importance coefficient values of each aquaculture area through the preset RFE model;

[0012] adjust, according to a device type of each aquaculture device, an aquaculture area where each aquaculture device is located, the predicted wave energy production, and the sorting result of the importance coefficient values, a running period and a running power of the first type of aquaculture device, the second type of aquaculture device, and the third type of aquaculture device.

[0013] In a possible implementation manner of the first aspect, the sorting of the remaining aquaculture periods of each sub-aquaculture area from small to large, and obtaining a plurality of working periods according to a sorting result, specifically comprises:

[0014] statistically obtaining the remaining aquaculture periods of each sub-aquaculture area;

[0015] sorting a plurality of the remaining aquaculture periods from small to large, and taking a first of the remaining aquaculture periods as a next working period;

[0016] for the remaining aquaculture periods other than the first of the remaining aquaculture periods, taking an end date of a previous one of the remaining aquaculture periods as a start date of a corresponding working period, and taking an end date of a current one of the remaining aquaculture periods as an end date of the corresponding working period.

[0017] In a possible implementation manner of the first aspect, before the adjusting of the running period and the running power of the first type of aquaculture device, the second type of aquaculture device, and the third type of aquaculture device, the method further comprises:

[0018] collecting historical running data of each aquaculture device;

[0019] feature extraction is performed on historical operation data of the respective aquaculture devices to obtain five-dimensional device feature vectors corresponding to the respective aquaculture devices; a first dimension of the five-dimensional device feature vector is a device type value, a second dimension of the five-dimensional device feature vector is a device working environment value, a third dimension of the five-dimensional device feature vector is a device working time length value; a fourth dimension of the five-dimensional device feature vector is the latitude and longitude value of the device; and a fifth dimension of the five-dimensional device feature vector is a quarterly yield value of an aquaculture farm to which the device belongs.

[0020] K-means clustering is performed on all the five-dimensional device feature vectors to obtain three clusters; the five-dimensional device feature vectors in the first cluster correspond to the first type of aquaculture device, the five-dimensional device feature vectors in the second cluster correspond to the second type of aquaculture device, and the five-dimensional device feature vectors in the third cluster correspond to the third type of aquaculture device.

[0021] In a possible implementation manner of the first aspect, the number of features is five, and the elements of each array in the three-dimensional tensor are, in sequence, a wave height value, a wave period, a wave direction angle, a wind speed, and a wind direction angle.

[0022] In a possible implementation manner of the first aspect, the importance coefficient value ranking result of each aquaculture zone is obtained by using a preset RFE model, and specifically includes the following steps.

[0023] Data of different aquaculture zones are collected to form an aquaculture data set, and an aquaculture yield is taken as a target variable.

[0024] The aquaculture data set is divided into an aquaculture training set and an aquaculture test set.

[0025] The preset logistic regression model is trained using the aquaculture training set.

[0026] The aquaculture test set is predicted by using the logistic regression model to obtain importance coefficient ranking results of multiple aquaculture zones, and each importance coefficient decreases in size in sequence; the size of each importance coefficient reflects the size of the influence of the aquaculture zone on the aquaculture yield.

[0027] In a possible implementation manner of the first aspect, the data of different aquaculture zones are collected to form an aquaculture data set, and specifically includes the following steps.

[0028] Residual aquaculture periods, latitudes and longitudes, dissolved oxygen values, pH values, temperatures, feed types, and degrees of manual intervention of different aquaculture zones are collected.

[0029] Missing values and abnormal values in the sample are processed.

[0030] Numerical features are normalized to obtain an aquaculture data set formed by multiple samples.

[0031] In a possible implementation manner of the first aspect, the operation period and operation power of the first type of aquaculture equipment, the second type of aquaculture equipment and the third type of aquaculture equipment are adjusted in sequence according to the device type of each aquaculture equipment, the aquaculture area where each aquaculture equipment is located, the predicted wave energy production and the importance coefficient value ranking result, and specifically comprising:

[0032] The wave energy required by the first type of aquaculture equipment, the second type of aquaculture equipment and the third type of aquaculture equipment in the next working period is counted.

[0033] The importance coefficient value ranking of each first type of aquaculture equipment, the importance coefficient value ranking of each second type of aquaculture equipment and the importance coefficient value ranking of each third type of aquaculture equipment are confirmed according to the aquaculture area where each aquaculture equipment is located and the importance coefficient value ranking result.

[0034] If the wave energy required by all the first type of aquaculture equipment is greater than the predicted wave energy production, the operation period and operation power of each first type of aquaculture equipment are adjusted in sequence according to the importance coefficient value ranking of each first type of aquaculture equipment.

[0035] If the wave energy required by all the first type of aquaculture equipment is less than or equal to the predicted wave energy production, and the wave energy required by all the first type of aquaculture equipment and all the second type of aquaculture equipment is greater than the predicted wave energy production, after meeting the operation period requirement and operation power requirement of all the first type of aquaculture equipment, the operation period and operation power of each second type of aquaculture equipment are adjusted in sequence according to the importance coefficient value ranking of each second type of aquaculture equipment.

[0036] If the wave energy required by all the first type of aquaculture equipment and all the second type of aquaculture equipment is less than or equal to the predicted wave energy production, and the wave energy required by all the first type of aquaculture equipment, all the second type of aquaculture equipment and all the third type of aquaculture equipment is greater than the predicted wave energy production, after meeting the operation period requirement and operation power requirement of all the first type of aquaculture equipment and all the second type of aquaculture equipment, the operation period and operation power of each third type of aquaculture equipment are adjusted in sequence according to the importance coefficient value ranking of each third type of aquaculture equipment.

[0037] In a possible implementation manner of the first aspect, the first type of aquaculture equipment includes water quality monitoring equipment, feed feeding equipment and waste treatment equipment; the second type of aquaculture equipment includes underwater camera monitoring equipment, water pumps, filtration equipment and disease prevention equipment; and the automatic control equipment, greenhouse and incubation equipment and aquatic product processing equipment.

[0038] In a possible implementation manner of the first aspect, before the operation period and operation power of the first type of cultivation equipment, the second type of cultivation equipment and the third type of cultivation equipment are sequentially adjusted, the method further includes:

[0039] adjusting the operation period and operation power of all the safety equipment, the safety equipment including lifesaving equipment, fire extinguishing equipment and emergency power supply; the power supply power of the emergency power supply is greater than the total rated power of all the first type of cultivation equipment.

[0040] The second aspect of the embodiment of the application provides a wave energy control distribution system for offshore cultivation, including:

[0041] a working period module, configured to acquire cultivation periods of each sub-cultivation area of an offshore cultivation area, sort the remaining cultivation periods of each sub-cultivation area from small to large, and obtain a plurality of working periods according to the sorting result;

[0042] a wave energy prediction module, configured to obtain a predicted wave energy yield of a next working period through a preset neural network model;

[0043] an importance sorting module, configured to obtain an importance coefficient value sorting result of each cultivation area through a preset RFE model;

[0044] a device adjustment module, configured to sequentially adjust the operation period and operation power of a first type of cultivation equipment, a second type of cultivation equipment and a third type of cultivation equipment according to the device type of each cultivation equipment, the cultivation area where each cultivation equipment is located, the predicted wave energy yield and the importance coefficient value sorting result.

[0045] Compared with the prior art, the embodiment of the application provides a wave energy control distribution method and system for offshore cultivation. In the first aspect, a prediction model is established to predict the wave energy yield in the next working period in advance. In this way, the operation plan of the cultivation equipment can be arranged according to the prediction result, and the waste or shortage of resources can be avoided. On the other hand, the remaining cultivation periods of each sub-cultivation area are sorted, and then different working periods are divided according to the sorting result. In this way, the demand of the sub-cultivation area that is about to end the cultivation period can be preferentially met when the wave energy yield is high, so as to improve the cultivation efficiency. On the other hand, the operation period and operation power of different types of cultivation equipment are adjusted according to the predicted wave energy yield and the importance coefficient sorting result of the cultivation area. For example, the operation power of the equipment can be appropriately increased during the period when the wave energy yield is high, and the operation period and power of the equipment can be appropriately reduced during the period when the yield is low, so as to save energy. BRIEF DESCRIPTION OF DRAWINGS

[0046] FIG. 1 is a flowchart of a wave energy control distribution method for offshore cultivation provided by an embodiment of the application;

[0047] Fig. 2 is a structural schematic diagram of a wave energy control and distribution system for offshore aquaculture according to an embodiment of the present application. DETAILED DESCRIPTION

[0048] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0049] Referring to Fig. 1, the present application provides a wave energy control and distribution method for offshore aquaculture, comprising:

[0050] S10, obtaining the aquaculture periods of each sub-aquaculture area in the offshore aquaculture area, sorting the remaining aquaculture periods of each sub-aquaculture area from small to large, and obtaining a plurality of working periods according to the sorting result.

[0051] S11, obtaining the predicted wave energy yield of the next working period through a preset neural network model.

[0052] S12, obtaining the importance coefficient value sorting result of each aquaculture area through a preset RFE model.

[0053] S13, adjusting the operation period and operation power of the first type of aquaculture equipment, the second type of aquaculture equipment and the third type of aquaculture equipment in turn according to the equipment type of each aquaculture equipment, the aquaculture area where each aquaculture equipment is located, the predicted wave energy yield and the importance coefficient value sorting result.

[0054] In S10, the aquaculture periods are obtained and sorted, and the aquaculture information of each sub-aquaculture area is collected. The remaining aquaculture periods of each sub-aquaculture area are sorted from small to large. Because different sub-aquaculture areas cultivate different aquatic products, there are different time requirements for the corresponding aquatic products to reach full growth (harvesting period). The working periods are divided according to the time length close to the harvesting period (the smallest remaining aquaculture period). This way of dividing the working periods can prioritize the harvesting of aquatic products close to the harvesting period in each working period, effectively guaranteeing the output of aquatic products in the offshore aquaculture factory and ensuring that the wave energy distribution is more reasonable.

[0055] In defining the structure of the neural network model in S11, the input layer is set to three dimensions, such as (batch_size, time_steps, features).

[0056] The hidden layer is set to one LSTM layer and one GRU layer. By using a neural network model containing LSTM and GRU layers, the long-term dependencies in time series data can be better captured, thereby improving the accuracy of wave energy yield prediction.

[0057] The output layer is configured with several first neurons (each corresponding to a group of sensors) and a second neuron (which aggregates the outputs of all first neurons).

[0058] Suppose in S12, there are 3 groups of sensors deployed on the sea, and each group of sensors collects 5 features (such as wave height value, wave period, wave direction angle, wind speed and wind direction angle). Suppose the time span of the historical data is 10 time steps. Then the shape of the three-dimensional tensor is (3, 10, 5).

[0059] In S13, the three-dimensional tensor is input into the neural network model for training, and the obtained preset neural network model can predict the wave energy yield of each working period. Based on accurate wave energy yield prediction, resource allocation can be planned in advance to ensure that resources are fully utilized during periods of high wave energy yield and to avoid resource waste; by accurately predicting wave energy yield, the number of unnecessary device start-ups and shutdowns can be reduced, thereby reducing operating costs.

[0060] S14 predicts the wave energy yield of the next working period through the trained neural network model, which is to determine whether the wave energy yield can meet the demand of the sub-culture area culture equipment that is about to end the culture period. By accurately predicting the wave energy yield, the number of unnecessary device start-ups and shutdowns can be reduced, thereby reducing operating costs. S15 uses the Recursive Feature Elimination (RFE) model to determine which culture areas have a greater impact on culture yield, thereby identifying key culture areas. Subsequently, by giving priority to culture areas with higher importance, culture activities in these areas can be fully supported, thereby improving culture efficiency. S16 optimizes device operating parameters to ensure that culture activities are carried out under sufficient resources, which helps to improve culture yield.

[0061] Through S10-S16, according to different device types, culture areas, predicted wave energy yields, and importance coefficient value ranking results, the operation period and power of the device can be fine-tuned to ensure that resources are effectively utilized; by reasonably adjusting the operation period and power of the device, the device can be fully utilized during periods of high wave energy yield, avoiding device idling; by reducing device operation during periods of low wave energy yield, energy can be saved and operating costs can be reduced; by optimizing device operating parameters to ensure that culture activities are carried out under sufficient resources, culture yield can be improved.

[0062] By way of example, the remaining culture period of each sub-culture area is sorted from small to large, and a plurality of working periods are obtained according to the sorting results, specifically including:

[0063] The remaining culture period of each sub-culture area is counted.

[0064] Sort the multiple remaining cultivation periods from small to large, and take the first remaining cultivation period as the next working period.

[0065] For the remaining cultivation periods other than the first remaining cultivation period, take the end date of the last remaining cultivation period as the start date of the corresponding working period, and take the end date of the current remaining cultivation period as the end date of the corresponding working period.

[0066] By sorting the remaining cultivation periods of each sub-cultivation area from small to large and obtaining multiple working periods according to the sorting results, effective management and optimization of the cultivation period can be achieved. This approach helps to ensure that the needs of sub-cultivation areas with shorter cultivation periods are prioritized during periods of high wave energy production, thereby improving cultivation efficiency and resource utilization efficiency.

[0067] Suppose there are four sub-cultivation areas, and their remaining cultivation periods are as follows: sub-cultivation area A: 30 days; sub-cultivation area B: 45 days; sub-cultivation area C: 60 days; sub-cultivation area D: 15 days.

[0068] The specific division is as follows:

[0069] First working period: start date: 1st day; end date: 15th day; related sub-cultivation area: D.

[0070] Second working period: start date: 16th day; end date: 30th day; related sub-cultivation area: A;

[0071] Third working period: start date: 31st day; end date: 45th day; sub-cultivation area: B;

[0072] Fourth working period: start date: 46th day; end date: 60th day; related sub-cultivation area: C.

[0073] Exemplarily, before obtaining the predicted wave energy production of the next working period through the preset neural network model, the following steps are further included:

[0074] The input layer of the preset neural network model is set to three dimensions, and an LSTM layer and a GRU layer are selected to form the hidden layer of the preset neural network model.

[0075] In the output layer, a plurality of first neurons and a second neuron are set, each first neuron corresponds to a group of sensors deployed on the sea; the output of the second neuron is the sum of the outputs of all first neurons; each group of sensors is responsible for information collection of a sub-ocean area.

[0076] The historical marine data is converted into a three-dimensional tensor; a first dimension of the three-dimensional tensor is a batch sample number, the batch sample number is equal to a number of groups of marine sensors deployed; a second dimension of the three-dimensional tensor is a time step, the time step is equal to a time span of the historical marine data; and a third dimension of the three-dimensional tensor is a number of features. The three-dimensional tensor containing time series data can better capture the time dependence in the data, thereby improving the accuracy of the prediction model.

[0077] The three-dimensional tensor is input into a preset neural network model for training.

[0078] Exemplarily, the number of features is five, and elements of each array in the three-dimensional tensor are wave height value, wave period, wave direction angle, wind speed, and wind direction angle in turn.

[0079] The wave height value, the wave period, the wave direction angle, the wind speed, and the wind direction angle are selected as the features because these features are closely related to wave energy production. By using features directly related to wave energy production, the accuracy of the prediction model can be improved. Then the historical marine data is converted into a three-dimensional tensor, where each array represents data of a group of sensors, and elements in each array are wave height value, wave period, wave direction angle, wind speed, and wind direction angle in turn.

[0080] Suppose there are 3 groups of sensors deployed on the sea, and each group of sensors collects the above-mentioned 5 features.

[0081] Suppose the time span of the historical data is 10 time steps.

[0082] Then the shape of the three-dimensional tensor is (3, 10, 5), where: the first dimension (3) represents 3 groups of sensors. The second dimension (10) represents the time step, i.e. the time span of the historical data. The third dimension (5) represents the number of features, i.e. wave height value, wave period, wave direction angle, wind speed, and wind direction angle.

[0083] Exemplarily, before the operation period and operation power of the first type of aquaculture equipment, the second type of aquaculture equipment, and the third type of aquaculture equipment are adjusted in turn, the method comprises:

[0084] Collecting historical operation data of each aquaculture equipment.

[0085] Extracting features from the historical operation data of each aquaculture equipment to obtain a five-dimensional equipment feature vector corresponding to each aquaculture equipment; a first dimension of the five-dimensional equipment feature vector is an equipment type value, a second dimension of the five-dimensional equipment feature vector is an equipment working environment value, a third dimension of the five-dimensional equipment feature vector is an equipment working time length value; a fourth dimension of the five-dimensional equipment feature vector is an equipment latitude and longitude value; and a fifth dimension of the five-dimensional equipment feature vector is a seasonal yield value of an aquaculture farm to which the equipment belongs.

[0086] K-means clustering is performed on all five-dimensional device feature vectors to obtain three clusters; among them, the five-dimensional device feature vectors in the first cluster correspond to the first type of breeding equipment; the five-dimensional device feature vectors in the second cluster correspond to the second type of breeding equipment; and the five-dimensional device feature vectors in the third cluster correspond to the third type of breeding equipment.

[0087] The first dimension of the five-dimensional device feature vector is the device type value (for example: 1 = water quality monitoring device, 2 = feed feeding device, 3 = waste treatment device, etc.). The second dimension of the five-dimensional device feature vector is the device working environment value (for example: temperature, humidity, pH value, etc.). The third dimension of the five-dimensional device feature vector is the device working time value (for example: daily working time, working time in the past week, etc.). The fourth dimension of the five-dimensional device feature vector is the device latitude and longitude value (for example: the geographical position coordinates of the device). The fifth dimension of the five-dimensional device feature vector is the quarterly yield value of the breeding farm to which the device belongs (for example: the yield in the last quarter).

[0088] K-means clustering is performed on all five-dimensional device feature vectors. Choose an appropriate number of clusters K, usually determined by elbow rule or silhouette coefficient method. Run the K-means algorithm to get different cluster centers and device lists of the clusters. In this example, K = 3, which means the devices are divided into three clusters. According to the clustering results, the devices can be divided into the first type of breeding equipment, the second type of breeding equipment and the third type of breeding equipment.

[0089] Through K-means clustering, devices can be classified according to similarity, which is convenient for subsequent resource allocation and management. In terms of optimizing resource allocation: based on the clustering results, the roles of different types of devices in breeding can be better understood, so as to optimize resource allocation. In terms of improving breeding efficiency: through classification of devices, devices can be managed more effectively, and breeding efficiency can be improved.

[0090] Exemplarily, the sorting result of the importance coefficient values of each breeding area obtained by the preset RFE model specifically includes:

[0091] Collect data from different breeding areas to form a breeding data set, and take the breeding yield as the target variable.

[0092] The breeding data set is divided into a breeding training set and a breeding test set.

[0093] The preset logistic regression model is trained using the breeding training set.

[0094] The importance coefficient ranking results of the aquaculture areas are obtained by predicting the aquaculture test set through the logistic regression model, and each importance coefficient decreases in size in turn.

[0095] By using the recursive feature elimination (RFE) model, it can be determined which aquaculture area has a greater impact on the aquaculture yield, so as to identify the key aquaculture area. Based on the importance coefficient ranking results of the aquaculture area, the aquaculture area with higher importance is given priority, which can ensure that the aquaculture activities in these areas are fully supported, thereby improving the aquaculture efficiency.

[0096] Exemplarily, the data of different aquaculture areas are collected to form an aquaculture data set, which specifically includes:

[0097] The remaining aquaculture period, latitude and longitude, dissolved oxygen value, pH value, temperature, feed type and degree of manual intervention of different aquaculture areas are collected.

[0098] The missing values and outliers in the samples are processed.

[0099] The numerical features are normalized to form an aquaculture data set.

[0100] When collecting the remaining aquaculture period, latitude and longitude, dissolved oxygen value, pH value, temperature, feed type and degree of manual intervention of different aquaculture areas, it can be obtained by field investigation, historical records or sensor collection, etc.

[0101] In the process of processing missing values and outliers in the samples, appropriate filling methods are needed to process missing values, such as using average, median or mode filling. In the process of identifying and processing outliers, IQR method or other statistical methods can be used to identify and process outliers. The normalization of numerical features is to unify different scale features to the same range, and the methods that can be used include minimum-maximum normalization or Z-score standardization.

[0102] By collecting relevant data of different aquaculture areas and performing appropriate preprocessing, an aquaculture data set can be effectively formed. This approach helps to improve the accuracy of the logistic regression model, so as to better determine the importance coefficient ranking results of the aquaculture area, optimize resource allocation, and improve the aquaculture efficiency.

[0103] Exemplarily, the operation period and operation power of the first type of aquaculture equipment, the second type of aquaculture equipment and the third type of aquaculture equipment are adjusted in turn according to the device type of each aquaculture device, the aquaculture area where each aquaculture device is located, the predicted wave energy yield and the importance coefficient value ranking results, which specifically includes:

[0104] The statistical first type of breeding equipment, the second type of breeding equipment and the third type of breeding equipment need wave energy in the next working cycle.

[0105] According to the importance coefficient value ranking of each breeding area and the importance coefficient value ranking result, the importance coefficient value ranking of each first type of breeding equipment, the importance coefficient value ranking of each second type of breeding equipment and the importance coefficient value ranking of each third type of breeding equipment are confirmed.

[0106] If the wave energy needed by all the first type of breeding equipment is greater than the predicted wave energy production, according to the importance coefficient value ranking of each first type of breeding equipment, the running cycle and the running power provided by each first type of breeding equipment are adjusted in turn.

[0107] If the wave energy needed by all the first type of breeding equipment is less than or equal to the predicted wave energy production, and the wave energy needed by all the first type of breeding equipment and all the second type of breeding equipment is greater than the predicted wave energy production, after meeting the running cycle requirement and the running power requirement of all the first type of breeding equipment, according to the importance coefficient value ranking of each second type of breeding equipment, the running cycle and the running power provided by each second type of breeding equipment are adjusted in turn.

[0108] If the wave energy needed by all the first type of breeding equipment and all the second type of breeding equipment is less than or equal to the predicted wave energy production, and the wave energy needed by all the first type of breeding equipment, all the second type of breeding equipment and all the third type of breeding equipment is greater than the predicted wave energy production, after meeting the running cycle requirement and the running power requirement of all the first type of breeding equipment and all the second type of breeding equipment, according to the importance coefficient value ranking of each third type of breeding equipment, the running cycle and the running power provided by each third type of breeding equipment are adjusted in turn.

[0109] Suppose the total wave energy needed by the first type of breeding equipment is 80 kilowatt hours, the total wave energy needed by the second type of breeding equipment is 60 kilowatt hours, and the total wave energy needed by the third type of breeding equipment is 40 kilowatt hours. Then suppose the predicted wave energy production is 150 kilowatt hours. According to the importance coefficient value ranking of each breeding area, the importance coefficient value ranking of the first type of breeding equipment is A > B > C, the importance coefficient value ranking of the second type of breeding equipment is D > E > F, and the importance coefficient value ranking of the third type of breeding equipment is G > H > I.

[0110] Then when adjusting the equipment running parameters, the following methods can be referred to:

[0111] Since the total wave energy required by all the first type of aquaculture equipment is 80 kWh, which is less than the predicted wave energy production 150 kWh. According to the importance coefficient value ranking, the operation cycle and operation power of equipment A, B, C are adjusted in turn.

[0112] Since the total wave energy required by all the first type of aquaculture equipment and the second type of aquaculture equipment is 140 kWh, which is less than the predicted wave energy production 150 kWh. After meeting the requirements of the first type of aquaculture equipment, according to the importance coefficient value ranking, the operation cycle and operation power of equipment D, E, F are adjusted in turn.

[0113] Since the total wave energy required by all the first type of aquaculture equipment, the second type of aquaculture equipment and the third type of aquaculture equipment is 180 kWh, which is greater than the predicted wave energy production 150 kWh. After meeting the requirements of the first type of aquaculture equipment and the second type of aquaculture equipment, according to the importance coefficient value ranking, the operation cycle and operation power of equipment G, H, I are adjusted in turn until the wave energy is allocated. If the remaining wave energy production is not enough after adjusting the operation cycle and operation power of equipment G, keep the operation cycle and operation power of equipment H, I as 0.

[0114] Exemplarily, the first type of aquaculture equipment includes water quality monitoring equipment, feed feeding equipment and waste treatment equipment; the second type of aquaculture equipment includes underwater camera monitoring equipment, water pump, filtration equipment and disease prevention equipment; the automated control equipment, greenhouse and incubation equipment and aquatic processing equipment.

[0115] It should be noted that the first type of aquaculture equipment is essential to maintain the basic operation of the aquaculture area, and any failure may cause significant economic loss or ecological disaster; the second type of aquaculture equipment is important for daily operation, but their failure will affect the efficiency of aquaculture, but will not immediately cause disastrous consequences. The third type of aquaculture equipment belongs to auxiliary equipment, although it is not essential, but plays a positive role in improving the efficiency of aquaculture and reducing labor intensity, etc.

[0116] Exemplarily, before the operation cycle and operation power of the first type of aquaculture equipment, the second type of aquaculture equipment and the third type of aquaculture equipment are adjusted in turn, it further includes:

[0117] Adjust the operation cycle and operation power of all safety equipment, including life-saving equipment, fire extinguishing equipment and emergency power supply; the power supply power of the emergency power supply is greater than the total rated power of all the first type of aquaculture equipment.

[0118] Compared with the prior art, the embodiment of the application provides a wave energy control and distribution method for offshore aquaculture. In a first aspect, a prediction model is established to predict the wave energy yield in the next working period in advance. In this way, the operation plan of the aquaculture equipment can be arranged according to the prediction result to avoid waste or shortage of resources. On the other hand, the remaining aquaculture periods of each sub-aquaculture area are sorted, and then different working periods are divided according to the sorting result. In this way, the demand of the sub-aquaculture area with an ending aquaculture period can be preferentially met when the wave energy yield is high, thereby improving the aquaculture efficiency. On the other hand, the operation period and operation power of different types of aquaculture equipment are adjusted according to the predicted wave energy yield and the importance coefficient sorting result of the aquaculture area. For example, the operation power of the equipment can be appropriately increased during the period with high wave energy yield; and the operation period and power of the equipment can be appropriately reduced during the period with low yield to save energy.

[0119] An embodiment of the present application provides a wave energy control and distribution system for offshore aquaculture, comprising: a working period module 201, a model setting module 202, a data conversion module 203, a model training module 204, a wave energy prediction module 205, an importance sorting module 206, and a device adjustment module 207.

[0120] The working period module 201 is used to obtain the aquaculture periods of each sub-aquaculture area in the offshore aquaculture area, sort the remaining aquaculture periods of each sub-aquaculture area from small to large, and obtain a plurality of working periods according to the sorting result.

[0121] The model setting module 202 is used to set the input layer of a preset neural network model to three dimensions, select an LSTM layer and a GRU layer to constitute a hidden layer of the preset neural network model, set a plurality of first neurons and a second neuron in the output layer, each first neuron corresponds to a group of sensors deployed on the sea, and the output of the second neuron is the sum of the outputs of all the first neurons. Each group of sensors is responsible for information collection of a marine sub-area.

[0122] The data conversion module 203 is used to convert historical marine data into a three-dimensional tensor; the first dimension of the three-dimensional tensor is the number of batch samples, which is equal to the number of groups of sensors deployed on the sea; the second dimension of the three-dimensional tensor is the time step, which is equal to the time span of the historical marine data; and the third dimension of the three-dimensional tensor is the number of features.

[0123] The model training module 204 is used to input the three-dimensional tensor into the preset neural network model for training.

[0124] The wave energy prediction module 205 is used to obtain the predicted wave energy yield in the next working period through the preset neural network model.

[0125] The importance ranking module 206 is configured to obtain the importance coefficient value ranking result of each cultivation area by using the preset RFE model.

[0126] The device adjustment module 207 is configured to adjust the operation period and operation power of the first type of cultivation device, the second type of cultivation device and the third type of cultivation device in sequence according to the device type of each cultivation device, the cultivation area where each cultivation device is located, the predicted wave energy yield and the importance coefficient value ranking result. It can be clearly understood by those skilled in the art that, for the convenience and brevity of description, the specific working process of the identification system described above can refer to the corresponding process in the foregoing method embodiments, which will not be described here 。

[0127] Compared with the prior art, the embodiment of the present application provides a wave energy control and distribution system for offshore cultivation. In the first aspect, a prediction model is established to predict the wave energy yield in the next working period in advance. In this way, the operation plan of the cultivation device can be arranged according to the prediction result, so as to avoid waste or shortage of resources. On the other hand, the remaining cultivation periods of each sub-cultivation area are ranked, and then different working periods are divided according to the ranking result. In this way, it can be ensured that the demand of the sub-cultivation area which is about to end the cultivation period is met preferentially when the wave energy yield is high, so as to improve the cultivation efficiency. On the other hand, the operation period and operation power of different types of cultivation devices are adjusted according to the predicted wave energy yield and the importance coefficient ranking result of the cultivation area. For example, the operation power of the device can be appropriately increased during the period when the wave energy yield is high, and the operation period and power of the device can be appropriately reduced during the period when the yield is low, so as to save energy.

[0128] The above is the preferred embodiment of the present application. It should be noted that for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which are also considered within the protection scope of the present application.

Claims

1. A method of wave energy control distribution for offshore farming, characterized by, The method comprises the following steps: The input layer of the preset neural network model is set to three dimensions, an LSTM layer and a GRU layer are selected to constitute the hidden layer of the preset neural network model, a plurality of first neurons and a second neuron are set in the output layer, each first neuron corresponds to a group of sensors deployed on the sea, the output of the second neuron is the sum of the outputs of all first neurons, and each group of sensors is responsible for information collection of a marine sub-region; The historical marine data is converted into a three-dimensional tensor, the first dimension of the three-dimensional tensor is the number of batch samples, the number of batch samples is equal to the number of groups of sensors deployed on the sea, the second dimension of the three-dimensional tensor is the time step, and the time step is equal to the time span of the historical marine data, and the third dimension of the three-dimensional tensor is the number of features; The three-dimensional tensor is input into the preset neural network model for training; The predicted wave energy yield of the next working period is obtained through the preset neural network model; An importance coefficient value ranking result of each cultivation area is obtained through a preset recursive feature elimination model, specifically including: collecting data of different cultivation areas to form a cultivation data set, and taking the cultivation yield as a target variable; the cultivation data set is divided into a cultivation training set and a cultivation test set; the preset logistic regression model is trained using the cultivation training set; the cultivation test set is predicted through the logistic regression model to obtain an importance coefficient ranking result of a plurality of cultivation areas, and each importance coefficient decreases in size; the size of each importance coefficient reflects the size of the influence of the cultivation area on the cultivation yield; Historical operation data of each cultivation equipment is collected, feature extraction is performed on the historical operation data of each cultivation equipment to obtain a five-dimensional equipment feature vector corresponding to different cultivation equipment, the first dimension of the five-dimensional equipment feature vector is a device type value, the second dimension of the five-dimensional equipment feature vector is a device working environment value, the third dimension of the five-dimensional equipment feature vector is a device working time value, the fourth dimension of the five-dimensional equipment feature vector is a device latitude and longitude value, and the fifth dimension of the five-dimensional equipment feature vector is a quarterly yield value of a cultivation farm to which the device belongs; K-means clustering is performed on all five-dimensional equipment feature vectors to obtain three clusters; wherein the five-dimensional equipment feature vectors in the first cluster correspond to first-type cultivation equipment, the five-dimensional equipment feature vectors in the second cluster correspond to second-type cultivation equipment, and the five-dimensional equipment feature vectors in the third cluster correspond to third-type cultivation equipment; The running period and running power of the first-type cultivation equipment, the second-type cultivation equipment and the third-type cultivation equipment are adjusted in sequence according to the device type of each cultivation equipment, the cultivation area where each cultivation equipment is located, the predicted wave energy yield and the importance coefficient value ranking result, specifically including: ​ The statistical first type of breeding equipment, the second type of breeding equipment and the third type of breeding equipment need wave energy in the next working cycle; the first type of breeding equipment includes water quality monitoring equipment, feed feeding equipment and waste treatment equipment; the second type of breeding equipment includes underwater camera monitoring equipment, water pump, filter equipment and disease prevention equipment; the third type of breeding equipment includes automatic control equipment, greenhouse and incubation equipment and aquatic product processing equipment; According to the importance coefficient value ranking of each breeding area and the importance coefficient value ranking result, the importance coefficient value ranking of each first type of breeding equipment, the importance coefficient value ranking of each second type of breeding equipment and the importance coefficient value ranking of each third type of breeding equipment are confirmed; If the wave energy required by all the first type of breeding equipment is greater than the predicted wave energy yield, the running cycle and running power of each first type of breeding equipment are adjusted in turn according to the importance coefficient value ranking of each first type of breeding equipment; If the wave energy required by all the first type of breeding equipment is less than or equal to the predicted wave energy yield, and the wave energy required by all the first type of breeding equipment and all the second type of breeding equipment is greater than the predicted wave energy yield, after meeting the running cycle requirements and running power requirements of all the first type of breeding equipment, the running cycle and running power of each second type of breeding equipment are adjusted in turn according to the importance coefficient value ranking of each second type of breeding equipment; If the wave energy required by all the first type of breeding equipment and all the second type of breeding equipment is less than or equal to the predicted wave energy yield, and the wave energy required by all the first type of breeding equipment, all the second type of breeding equipment and all the third type of breeding equipment is greater than the predicted wave energy yield, after meeting the running cycle requirements and running power requirements of all the first type of breeding equipment and all the second type of breeding equipment, the running cycle and running power of each third type of breeding equipment are adjusted in turn according to the importance coefficient value ranking of each third type of breeding equipment.

2. The method of claim 1, wherein, The remaining breeding cycles of each sub-breeding area are sorted from small to large, and a plurality of working cycles are obtained according to the sorting result, which specifically includes: The remaining breeding cycles of each sub-breeding area are counted; The plurality of remaining breeding cycles are sorted from small to large, and the first remaining breeding cycle is taken as the next working cycle; For the remaining breeding cycles other than the first remaining breeding cycle, the end day of the last remaining breeding cycle is taken as the start day of the corresponding working cycle, and the end day of the current remaining breeding cycle is taken as the end day of the corresponding working cycle.

3. The method of claim 1, wherein the wave energy is controlled and distributed for offshore farming. The number of features is five, and the elements of each array in the three-dimensional tensor are wave height value, wave period, wave direction angle, wind speed and wind direction angle in turn.

4. The method of claim 1, wherein the wave energy is controlled and distributed for offshore farming. The data of different breeding areas are collected to form a breeding data set, which specifically includes: The remaining breeding cycles, latitude and longitude, dissolved oxygen value, pH value, temperature, feed type and degree of artificial intervention of different breeding areas are collected; Missing values and abnormal values in the sample are processed; Numerical features are normalized to obtain a plurality of samples to form a breeding data set.

5. The method of claim 1, wherein, Before the operation cycle and operation power of the first type of breeding equipment, the second type of breeding equipment and the third type of breeding equipment are adjusted in sequence, comprising: Adjusting the operation cycle and operation power of all safety equipment, including life-saving equipment, fire extinguishing equipment and emergency power supply; the power supply power of the emergency power supply is greater than the total rated power of all the first type of breeding equipment.

6. A wave energy control distribution system for offshore farming, c h a r a c t e r i s e d in that Comprising: A working cycle module for obtaining the breeding cycle of each sub-breeding area in the offshore breeding area, sorting the remaining breeding cycle of each sub-breeding area from small to large, and obtaining a plurality of working cycles according to the sorting result; A model setting module for setting the input layer of a preset neural network model to three dimensions, selecting an LSTM layer and a GRU layer to constitute a preset neural network model hidden layer; setting a plurality of first neurons and a second neuron in the output layer, each first neuron corresponding to a group of sensors deployed on the sea; the output of the second neuron is the sum of the outputs of all first neurons; each group of sensors is responsible for information collection of a marine sub-region; A data conversion module for converting historical marine data into a three-dimensional tensor; the first dimension of the three-dimensional tensor is the number of batch samples, which is equal to the number of groups of sensors deployed on the sea; the second dimension of the three-dimensional tensor is the time step, which is equal to the time span of the historical marine data; the third dimension of the three-dimensional tensor is the number of features; A model training module for inputting the three-dimensional tensor into the preset neural network model for training; A wave energy prediction module for obtaining the predicted wave energy yield of the next working cycle through the preset neural network model; An importance sorting module for obtaining an importance coefficient value sorting result of each breeding area through a preset recursive feature elimination model; A device adjustment module for collecting historical operation data of each breeding device; performing feature extraction on the historical operation data of each breeding device to obtain a five-dimensional device feature vector corresponding to different breeding devices; the first dimension of the five-dimensional device feature vector is a device type value, the second dimension of the five-dimensional device feature vector is a device working environment value, the third dimension of the five-dimensional device feature vector is a device working time value; the fourth dimension of the five-dimensional device feature vector is a device latitude and longitude value; the fifth dimension of the five-dimensional device feature vector is a quarterly yield value of the breeding farm to which the device belongs; K-means clustering is performed on all five-dimensional device feature vectors to obtain three clusters; wherein the five-dimensional device feature vector in the first cluster corresponds to the first type of breeding equipment; the five-dimensional device feature vector in the second cluster corresponds to the second type of breeding equipment; the five-dimensional device feature vector in the third cluster corresponds to the third type of breeding equipment; according to the device type of each breeding device, the breeding area where each breeding device is located, the predicted wave energy yield and the importance coefficient value sorting result, the operation cycle and operation power of the first type of breeding equipment, the second type of breeding equipment and the third type of breeding equipment are adjusted in sequence, specifically comprising: The statistical first type of breeding equipment, the second type of breeding equipment and the third type of breeding equipment need wave energy in the next working cycle; the first type of breeding equipment includes water quality monitoring equipment, feed feeding equipment and waste treatment equipment; the second type of breeding equipment includes underwater camera monitoring equipment, water pump, filter equipment and disease prevention equipment; the third type of breeding equipment includes automatic control equipment, greenhouse and incubation equipment and aquatic processing equipment; According to the importance coefficient value ranking of each breeding area and the importance coefficient value ranking result, the importance coefficient value ranking of each first type of breeding equipment, the importance coefficient value ranking of each second type of breeding equipment and the importance coefficient value ranking of each third type of breeding equipment are confirmed; If the wave energy required by all the first type of breeding equipment is greater than the predicted wave energy production, the operation cycle and operation power of each first type of breeding equipment are adjusted in turn according to the importance coefficient value ranking of each first type of breeding equipment; If the wave energy required by all the first type of breeding equipment is less than or equal to the predicted wave energy production, and the wave energy required by all the first type of breeding equipment and all the second type of breeding equipment is greater than the predicted wave energy production, after meeting the operation cycle requirement and operation power requirement of all the first type of breeding equipment, the operation cycle and operation power of each second type of breeding equipment are adjusted in turn according to the importance coefficient value ranking of each second type of breeding equipment; If the wave energy required by all the first type of breeding equipment and all the second type of breeding equipment is less than or equal to the predicted wave energy production, and the wave energy required by all the first type of breeding equipment, all the second type of breeding equipment and all the third type of breeding equipment is greater than the predicted wave energy production, after meeting the operation cycle requirement and operation power requirement of all the first type of breeding equipment and all the second type of breeding equipment, the operation cycle and operation power of each third type of breeding equipment are adjusted in turn according to the importance coefficient value ranking of each third type of breeding equipment.

Images