Hermetia illucens manure treatment system based on intelligent control

Abstract

The invention discloses a hermetia illucens feces treatment system based on intelligent control, and relates to the technical field of environmental protection, and the system comprises a central control unit which controls the preset number of larvae and sends the preset number of larvae to a treatment capacity prediction module; the sensor module is connected with the central control unit to collect temperature and humidity data, a temperature and humidity range is preset in the central control unit, and larva growth temperature and humidity are adjusted based on the preset temperature and humidity range and sent to the handling capacity prediction module; the disease detection module monitors the body color change of larvae in real time, calculates the disease index of the larvae, and sends the disease index and image data to the handling capacity prediction module through the central control unit; the treatment capacity prediction module predicts the feces treatment capacity based on the larva disease index, the temperature, the humidity and the quantity, and the central control unit performs feces putting in the next putting period based on the initial feces putting quantity and the feces treatment capacity. According to the invention, larva feces treatment capacity evaluation is carried out through temperature and humidity, larva quantity and larva disease conditions, and automatic feces treatment is realized.

Description

Technical Field

[0001] The invention relates to the technical field of environmental protection, and more particularly to a black soldier fly excrement treatment system based on intelligent control. Background Art

[0002] The black soldier fly (Hermetia illucens) is an insect with broad application prospects in waste treatment and organic waste degradation. The larvae of black soldier flies can efficiently convert organic waste into biomass protein and fertilizer, so they have received increasing attention in agriculture, aquaculture and waste management. At present, many farms and organic waste treatment facilities have begun to introduce black soldier flies to treat organic waste such as feces. Traditional black soldier fly feces treatment technology mainly relies on manual management, with limited treatment efficiency and treatment capacity, and has high requirements for environmental conditions (such as temperature, humidity, etc.).

[0003] Although the application of black soldier flies in manure and sewage treatment has achieved certain results, the existing technology has several obvious shortcomings. First, the level of automation in the treatment process is low, and it usually relies on manual monitoring and operation, making it difficult to achieve large-scale and efficient treatment. Secondly, the monitoring and regulation of the growth environment of black soldier flies is relatively extensive, resulting in large fluctuations in treatment efficiency, especially when environmental conditions such as temperature and humidity are unstable, the treatment capacity of black soldier flies may drop significantly. In addition, the existing system lacks intelligent management methods in terms of waste placement, monitoring of the treatment process, and separation of manure and sewage, and problems such as waste of resources and low efficiency are prone to occur during the treatment process.

[0004] Therefore, how to automatically process black soldier fly feces is a problem that technicians in this field urgently need to solve. Summary of the invention

[0005] In view of this, the present invention provides a black soldier fly feces treatment system based on intelligent control, which accurately evaluates the larval feces treatment volume through temperature and humidity, larval number, and larval disease conditions, thereby achieving efficient feces treatment and being able to automatically perform feces treatment.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] A black soldier fly feces and sewage treatment system based on intelligent control, comprising: a central control unit, a temperature and humidity control module, a feces and sewage delivery module, a sensor module, a larva delivery module, a disease detection module, and a processing volume prediction module, wherein the central control unit controls the larva delivery module to deliver a preset number of larvae, and sends the number to the processing volume prediction module; the sensor module comprises a temperature sensor and a humidity sensor, and the temperature sensor and the humidity sensor are connected to the central control unit, and a temperature and humidity range is preset in the central control unit. Based on the preset temperature and humidity range, the temperature and humidity control module is controlled to adjust the larvae growth temperature and humidity, and the number is sent to the processing volume prediction module; the disease detection module monitors the larvae body color changes in real time, calculates the larvae disease index, and sends the image data to the processing volume prediction module through the central control unit; the processing volume prediction module predicts the feces and sewage treatment volume based on the disease index, the larvae growth temperature and humidity, and the number of larvae, and sends the result to the central control unit; the central control unit delivers feces and sewage in the next delivery cycle based on the initial feces and sewage delivery volume and the feces and sewage treatment volume.

[0008] Preferably, the disease detection module specifically includes:

[0009] The feature value extraction module extracts the feature value of the disease at each collection moment within a set time period;

[0010] The disease index calculation module determines the disease index of the disease based on the characteristic value of the disease; the determination formula of the disease index of the disease is:

[0011]

[0012] Among them, F t represents the disease index at time t, w j represents the weight corresponding to the characteristic value of the jth disease, f j Represents the characteristic value of the jth disease, j = 1, 2, ..., m.

[0013] Preferably, the processing volume prediction module specifically includes:

[0014] The model building module uses a deep convolutional neural network and a time series information storage network to integrate the temperature and humidity of larvae growth, disease index and environmental attributes of black soldier fly feces treatment at different times, and constructs a feces treatment volume prediction model;

[0015] Model training module: Joint training of the time series information storage network and the deep convolutional neural network. The image data of the number of larvae released in history is used as the training sample of the deep convolutional neural network, and the temperature and humidity of larvae growth and the disease index are used as the training sample of the time series information storage network. The two are jointly trained.

[0016] The model prediction module obtains the growth temperature and humidity, disease index and number of larvae to be tested, obtains the corresponding manure and sewage treatment volume prediction distribution through the manure and sewage treatment volume prediction model, and selects the treatment volume with the maximum probability of exceeding a preset threshold as the manure and sewage treatment volume corresponding to the data to be tested.

[0017] Preferably, the model building module specifically includes:

[0018] The joint learning module models the characteristics of black soldier fly feces treatment data and selects the larval growth temperature and humidity, disease index, and larval number that affect the amount of black soldier fly feces treatment for time-domain joint learning;

[0019] The prediction distribution module stores the final hidden layer state h of the network unit in the time series information of multiple iterations t The softmax function is used to estimate the probability distribution y of the amount of manure treated. t , the formula is as follows:

[0020] y t =softmax(W*h t + b);

[0021] Among them, W and b are weight and bias terms.

[0022] Preferably, the model training module specifically includes:

[0023] Initialization module, which performs end-to-end training of deep convolutional neural networks, uses image data to pre-train the model and initialize it;

[0024] Vector value calculation module, forward calculation training of the time series information storage network, using the pre-trained deep convolutional neural network to extract features from image data obtained in different time periods, and forward calculate the output value of each neuron in sequence through the time series information storage network, the time series information storage network calculates i t 、c t 、f t , o t 、h t The values ​​of the five vectors;

[0025] The weight gradient calculation module fine-tunes all parameters of the deep convolutional neural network and the temporal information storage network at the same time, and reversely calculates the error term δ value of each neuron; the back propagation of the error term of the temporal information storage network includes two directions: one is the back propagation along time, starting from the current moment t, and calculating the error term at each moment; the other is to propagate the error term to the upper layer of the feature extraction network, and calculate the gradient of each weight according to the corresponding error term.

[0026] Preferably, the joint learning module specifically includes:

[0027] The attribute setting module sets the environmental attribute elements that are constant in the time domain. K and L represent the black soldier fly species and feces type respectively; it sets the attribute elements that change in the time domain. From t = 1 to t = T, there is a temperature C that changes with time in the time domain. t , humidity S t and disease index F t ;

[0028] In the rule setting module, the timing information storage network update rules are set as follows:

[0029] x 0 =T 1 *A 1 ;

[0030] x t =T 2 *{F,A 2} t ,t∈{1,...,T};

[0031] Among them, A 1 is the time-invariant environmental attribute element {K, L}, A 2 is the attribute element that changes in the time domain {C t ,S t ,F t}, T 1 and T 2 They are the conversion matrices of time-invariant attributes and time-varying attributes respectively. The time-invariant environmental attribute factor feature set of the time-invariant information storage network is {x 0 ,x 1 ,...,x T};

[0032] The forward calculation setting module normalizes the feature data, controls the data dimension between 0 and 1, sets the number of layers of the temporal information storage network model to be consistent with the time point, and each layer has input and output. The first layer of the temporal information storage network model inputs the temporal invariant environmental attribute factor feature x 0 , and its impact results are continuously passed on to the prediction of each moment. The input of the n+1th layer of the temporal information storage network model layer, the image passes through the deep feature extraction network and is fused with the temporal variation environment attribute factors as the input of the n+1th layer of the temporal information storage network model, and is input into the layers of the temporal information storage network model in the order of different time series shooting, setting the forward calculation of the temporal information storage network unit model.

[0033] Preferably, the forward calculation setting module specifically includes:

[0034] Long-term storage unit c t-1 Through the forgetting unit ft To forget and discard information, set the forget unit f t Affected by the external input x at the current moment t , the short-term storage output h of the previous moment t-1 , the long-term storage c of the previous moment t-1 The control expression is as follows:

[0035] f t =σ(W xf x t +W hf h t-1 +W cf c t-1 +b f );

[0036] W xf , W hf , W cf , b f Respectively represent the weight of external input, the weight of short-term storage, the weight of long-term storage and the bias;

[0037] The external input x at the current time t and the short-term storage output h of the previous moment t-1 Calculate the new information at the current moment

[0038]

[0039] By input unit i t Control the new information at the current moment Write long-term storage unit to generate new long-term storage c t , which is expressed as follows:

[0040] i t =σ(W xi x t +W hi h t-1 +W ci c t-1 +b i );

[0041]

[0042] Among them, i t Subject to x t 、h t-1 、c t-1 control, i t 、c t 、f t , o tThey are the input unit, long-term storage unit, forgetting unit, and output unit on the temporal information storage network unit, respectively; where σ(·) represents the S-shaped activation function, · represents the component multiplication, and W is the weight matrix connecting different units;

[0043] Activate long-term storage unit c t , by the output unit o t Control the currently accumulated storage c t Select some relevant storage to generate the storage h that you are interested in at this moment t , and then store this part to output y t , o t =σ(W xo x t +W ho h t-1 +W co c t-1 +b o ), h t =o t ·tanh(c t );

[0044] Among them, the output unit o t Subject to x t 、h t-1 and the current time long storage c t control.

[0045] It can be seen from the above technical solutions that compared with the prior art, the present invention discloses a black soldier fly feces treatment system based on intelligent control, which realizes intelligent control of the entire system through a central control unit, including larval placement, temperature and humidity control, feces placement, etc., thereby improving the automation and operation efficiency of the system. Using black soldier flies to treat feces and sewage, combined with modules such as temperature and humidity control, disease detection and processing volume prediction, can more effectively treat feces and sewage and reduce environmental pollution; the processing volume prediction module predicts based on multiple factors such as disease index, larval growth temperature and humidity, and larval number, which improves the accuracy of feces and sewage treatment volume prediction and helps optimize feces and sewage placement strategies; the sensor module and the disease detection module can monitor the larval growth environment and health status in real time, and the central control unit adjusts the system parameters in time according to this information to ensure the growth of black soldier flies and the feces treatment effect; the joint training of the time series information storage network and the deep convolutional neural network can fully integrate the multi-time series attribute elements of black soldier fly feces and sewage treatment, and improve the performance and prediction ability of the model; according to different environmental attributes and black soldier fly species and other factors, it can be adjusted and optimized to adapt to different application scenarios and needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0047] Figure 1 A schematic diagram of the structure provided by the present invention;

[0048] Figure 2 Schematic diagram of the joint learning module provided by the present invention. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] The embodiment of the present invention discloses a black soldier fly feces treatment system based on intelligent control, such as Figure 1 As shown, it includes: a central control unit, a temperature and humidity control module, a manure delivery module, a sensor module, a larva delivery module, a disease detection module, and a processing volume prediction module. The central control unit controls the larva delivery module to deliver a preset number of larvae, and sends them to the processing volume prediction module; the sensor module includes a temperature sensor and a humidity sensor, the temperature sensor and the humidity sensor are connected to the central control unit, the temperature and humidity range is preset in the central control unit, and the temperature and humidity control module is controlled based on the preset temperature and humidity range to adjust the larval growth temperature and humidity, and send them to the processing volume prediction module; the disease detection module monitors the larval body color changes in real time, calculates the larval disease index, and sends it to the processing volume prediction module together with the image data through the central control unit; the processing volume prediction module predicts the manure treatment volume based on the disease index, the larval growth temperature and humidity, and the larval number, and sends it to the central control unit; the central control unit delivers manure in the next delivery cycle based on the initial manure delivery amount and the manure treatment amount.

[0051] In a specific embodiment, the disease detection module uses spectral analysis technology or machine learning algorithms, both of which are existing technologies, and uses a spectral sensor to illuminate the larvae to obtain the spectral information reflected or emitted by the larvae. Different body colors will cause changes in spectral characteristics. By analyzing spectral data, changes in larval body color can be monitored in real time. Combined with computer vision technology, machine learning algorithms, such as convolutional neural networks (CNNs) in deep learning, are used to train a large number of larval images so that the model can automatically identify and judge changes in larval body color.

[0052] In a specific embodiment, the disease detection module specifically includes:

[0053] The feature value extraction module extracts the feature value of the disease at each collection moment within a set time period;

[0054] The disease index calculation module determines the disease index of the disease based on the characteristic value of the disease; the formula for determining the disease index of the disease is:

[0055]

[0056] Among them, F t represents the disease index at time t, w j represents the weight corresponding to the characteristic value of the jth disease, f j Represents the characteristic value of the jth disease, j = 1, 2, ..., m.

[0057] In a specific embodiment, the processing volume prediction module specifically includes:

[0058] The model building module uses a deep convolutional neural network and a time series information storage network to integrate the temperature and humidity of larvae growth, disease index and environmental attributes of black soldier fly feces treatment at different times, and constructs a feces treatment volume prediction model;

[0059] Model training module: Joint training of the time series information storage network and the deep convolutional neural network. The image data of the number of larvae released in history is used as the training sample of the deep convolutional neural network, and the temperature and humidity of larvae growth and the disease index are used as the training sample of the time series information storage network. The two are jointly trained.

[0060] The model prediction module obtains the growth temperature and humidity, disease index and number of larvae to be tested, obtains the corresponding manure and sewage treatment volume prediction distribution through the manure and sewage treatment volume prediction model, and selects the treatment volume with the maximum probability of exceeding the preset threshold as the manure and sewage treatment volume corresponding to the data to be tested.

[0061] In a specific embodiment, the model building module specifically includes:

[0062] The joint learning module models the characteristics of black soldier fly feces treatment data and selects the larval growth temperature and humidity, disease index, and larval number that affect the amount of black soldier fly feces treatment for time-domain joint learning;

[0063] The prediction distribution module stores the final hidden layer state h of the network unit in the time series information of multiple iterations t The softmax function is used to estimate the probability distribution y of the amount of manure treated. t , the formula is as follows:

[0064] y t =softmax(W*h t + b);

[0065] Among them, W and b are weight and bias terms.

[0066] A feces treatment volume prediction model was constructed by integrating the temperature and humidity of larval growth, disease index and environmental attributes of black soldier fly feces treatment at different times using deep convolutional neural networks and time series information storage networks. Convolutional neural networks are highly adaptable and very suitable for processing data with statistical stationarity and local correlation. They implicitly learn the features of regions of different shapes from the captured training image data. Time series information storage networks have great advantages in learning long-term dependencies and time series in higher-level feature sequences, so the two are combined to model feces treatment volume from time and space.

[0067] In a specific embodiment, the model training module specifically includes:

[0068] Initialization module, which performs end-to-end training of deep convolutional neural networks, uses image data to pre-train the model and initialize it;

[0069] Vector value calculation module, forward calculation training of the time series information storage network, using the pre-trained deep convolutional neural network to extract features from image data obtained in different time periods, and forward calculate the output value of each neuron in sequence through the time series information storage network, the time series information storage network calculates i t 、c t 、f t , o t 、h t The values ​​of the five vectors;

[0070] The weight gradient calculation module fine-tunes all parameters of the deep convolutional neural network and the temporal information storage network at the same time, and reversely calculates the error term δ value of each neuron; the back propagation of the error term of the temporal information storage network includes two directions: one is the back propagation along time, starting from the current moment t, and calculating the error term at each moment; the other is to propagate the error term to the upper layer of the feature extraction network, and calculate the gradient of each weight according to the corresponding error term.

[0071] In a specific embodiment, Figure 2 As shown in Figure 2, the joint learning module specifically includes:

[0072] The attribute setting module sets the environmental attribute elements that do not change in the time domain. K and L represent the black soldier fly species and feces type respectively; it sets the attribute elements that change in the time domain. From t = 1 to t = T, there is a temperature C that changes with time in the time domain. t , humidity S t and disease index F t ;

[0073] In the rule setting module, the timing information storage network update rules are set as follows:

[0074] x 0 =T 1 *A 1 ;

[0075] x t =T 2 *{F,A 2} t ,t∈{1,...,T};

[0076] Among them, A 1 is the time-invariant environmental attribute element {K, L}, A 2 is the attribute element that changes in the time domain {C t ,S t ,F t}, T 1 and T 2 They are the conversion matrices of time-invariant attributes and time-varying attributes respectively. The time-invariant environmental attribute factor feature set of the time-invariant information storage network is {x 0 ,x 1 ,...,x T}; F is the disease index set.

[0077] The forward calculation setting module normalizes the feature data and controls the data dimension between 0 and 1. Since the dimensions of several parameters in the feature data are different, it is necessary to normalize the data first, control the data dimension between 0 and 1, and convert it into a dimensionless expression, which is conducive to eliminating the dimensional influence between dimensions. The number of layers of the temporal information storage network model is set to be consistent with the time point, and each layer has input and output. Here, each layer of the temporal information storage network model is equivalent to the staged process of healthy black soldier flies developing into diseased black soldier flies. The first layer of the temporal information storage network model inputs the time-invariant environmental attribute factor feature x 0 , and its impact results are continuously passed on to the prediction of each moment, so that the prediction of the second layer can be affected by the first layer, and the prediction of the third layer can be affected by the second layer. At the same time, the temporal information storage network model contains a storage unit that attempts to store information for a longer time. According to the temporal order, the impact of all pictures and environmental attributes can be accumulated and stored in sequence, which is beneficial to the final prediction. The input of the n+1 layer of the temporal information storage network model layer, the image after passing through the deep feature extraction network and the fusion of the temporal changing environmental attribute factors are used as the input of the n+1 layer of the temporal information storage network model, and are input into the layers of the temporal information storage network model in the order of different time series shooting, and the forward calculation of the temporal information storage network unit model is set.

[0078] In a specific embodiment, the forward calculation setting module specifically includes:

[0079] Long-term storage unit c t-1 Through the forgetting unit f t To forget and discard information, set the forget unit f t Affected by the external input x at the current moment t , the short-term storage output h of the previous moment t-1 , the long-term storage c of the previous moment t-1 The control expression is as follows:

[0080] f t =σ(W xf x t +W hf h t-1 +W cf c t-1 +b f );

[0081] W xf , W hf , W cf , b f Respectively represent the external input weight, short-term storage weight, long-term storage weight and bias of the forgetting unit;

[0082] The external input x at the current timet and the short-term storage output h of the previous moment t-1 Calculate the new information at the current moment

[0083]

[0084] Among them, W xc is the external input weight of the new information, W hc is the short-term storage weight of the new information, b c Bias for long-term storage of new information.

[0085] By input unit i t Control the new information at the current moment Write long-term storage unit to generate new long-term storage c t , which is expressed as follows:

[0086] i t =σ(W xi x t +W hi h t-1 +W ci c t-1 +b i );

[0087]

[0088] Among them, i t Subject to x t 、h t-1 、c t-1 control, i t 、c t 、f t , o t are the input unit, long-term storage unit, forgetting unit, and output unit of the temporal information storage network unit, respectively; where σ(·) represents the S-shaped activation function, · represents the component multiplication, and W is the weight matrix connecting different units; W xi is the external input weight of the input unit, W hi is the short-term storage weight of the input unit, b i Long-term storage bias for the input unit.

[0089] Activate long-term storage unit c t , by the output unit o t Control the currently accumulated storage c t Select some relevant storage to generate the storage h that you are interested in at this moment t , and then store this part to output y t , o t =σ(W xo xt +W ho h t-1 +W co c t-1 +b o ), h t =o t ·tanh(c t );t is the current moment, t-1 is the previous moment.

[0090] Among them, the output unit o t Subject to x t 、h t-1 and the current time long storage c t Control of W xo is the external input weight of the output unit, W ho is the short-term storage weight of the output unit, b o Long-term storage bias for the output unit.

[0091] In a specific embodiment, the preset temperature and humidity range in the central control unit specifically includes:

[0092]

[0093] Among them, Q opt is the optimal temperature setting value, H opt is the optimal humidity setting value, Q 当前 and H 当前 is the current temperature and humidity, Q 目标 and H 目标 This is the most suitable temperature and humidity range for black soldier flies.

[0094] In a specific embodiment, a communication module is also included, which enables managers to monitor the processing process in real time through a remote terminal and receive alarm information in time when an abnormality occurs. This function makes the management of the system more convenient and flexible, and is particularly suitable for large farms or remote operation scenarios. This advantage comes from the combination of the communication module and the central control module, which realizes wireless transmission of data and remote monitoring.

[0095] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0096] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A black soldier fly feces treatment system based on intelligent control, characterized in that: include: A central control unit, a temperature and humidity control module, a manure delivery module, a sensor module, a larva delivery module, a disease detection module, and a processing volume prediction module, wherein the central control unit controls the larva delivery module to deliver a preset number of larvae, and sends the result to the processing volume prediction module; the sensor module includes a temperature sensor and a humidity sensor, and the temperature sensor and the humidity sensor are connected to the central control unit. A temperature and humidity range is preset in the central control unit, and the temperature and humidity control module is controlled to adjust the larval growth temperature and humidity based on the preset temperature and humidity range, and the result is sent to the processing volume prediction module; the disease detection module monitors the changes in larval body color in real time, calculates the disease index of the larvae, and sends the result together with the image data to the processing volume prediction module through the central control unit; the processing volume prediction module predicts the manure delivery volume based on the disease index, the larval growth temperature and humidity, and the number of larvae, and sends the result to the central control unit; the central control unit delivers manure in the next delivery cycle based on the initial manure delivery volume and the manure delivery volume.

2. According to claim 1, a black soldier fly feces treatment system based on intelligent control is characterized in that: The disease detection module specifically includes: The feature value extraction module extracts the feature value of the disease at each collection moment within a set time period; The disease index calculation module determines the disease index of the disease based on the characteristic value of the disease; the determination formula of the disease index of the disease is: Among them, F t represents the disease index at time t, w j represents the weight corresponding to the characteristic value of the jth disease, f j Represents the characteristic value of the jth disease, j = 1, 2, ..., m.

3. The black soldier fly feces treatment system based on intelligent control according to claim 1 is characterized in that: The processing volume prediction module specifically includes: The model building module uses a deep convolutional neural network and a time series information storage network to integrate the temperature and humidity of larvae growth, disease index and environmental attributes of black soldier fly feces treatment at different times, and constructs a feces treatment volume prediction model; Model training module: Joint training of the time series information storage network and the deep convolutional neural network. The image data of the number of larvae released in history is used as the training sample of the deep convolutional neural network, and the temperature and humidity of larvae growth and the disease index are used as the training sample of the time series information storage network. The two are jointly trained. The model prediction module obtains the growth temperature and humidity, disease index and number of larvae to be tested, obtains the corresponding manure and sewage treatment volume prediction distribution through the manure and sewage treatment volume prediction model, and selects the treatment volume with the maximum probability of exceeding a preset threshold as the manure and sewage treatment volume corresponding to the data to be tested.

4. The black soldier fly feces treatment system based on intelligent control according to claim 3 is characterized in that: The model building module specifically includes: The joint learning module models the characteristics of black soldier fly feces treatment data and selects the larval growth temperature and humidity, disease index, and larval number that affect the amount of black soldier fly feces treatment for time-domain joint learning; The prediction distribution module stores the final hidden layer state h of the network unit in the time series information of multiple iterations t The softmax function is used to estimate the probability distribution y of the amount of manure treated. t , the formula is as follows: y t =softmax(W*h t +b) Among them, W and b are weight and bias terms.

5. The black soldier fly feces treatment system based on intelligent control according to claim 3 is characterized in that: The model training module specifically includes: Initialization module, which performs end-to-end training of deep convolutional neural networks, uses image data to pre-train the model and initialize it; Vector value calculation module, forward calculation training of the time series information storage network, using the pre-trained deep convolutional neural network to extract features from image data obtained in different time periods, and forward calculate the output value of each neuron in sequence through the time series information storage network, the time series information storage network calculates i t 、c t 、f t , o t 、h t The values ​​of the five vectors; The weight gradient calculation module fine-tunes all parameters of the deep convolutional neural network and the time series information storage network at the same time, and reversely calculates the error term δ value of each neuron; the back propagation of the error term of the time series information storage network includes two directions: one is the back propagation along time, starting from the current moment t, and calculating the error term at each moment; the other is to propagate the error term to the upper layer of the feature extraction network, and calculate the gradient of each weight according to the corresponding error term.

6. The black soldier fly feces treatment system based on intelligent control according to claim 4 is characterized in that: The joint learning module specifically includes: The attribute setting module sets the environmental attribute elements that are constant in the time domain. K and L represent the black soldier fly species and feces type respectively; it sets the attribute elements that change in the time domain. From t = 1 to t = T, there is a temperature C that changes with time in the time domain. t , humidity S t and disease index F t ; In the rule setting module, the timing information storage network update rules are set as follows: x0=T1*A1; x t =T2*{F,A2} t ,t∈{1,...,T}; Among them, A1 is the environmental attribute element {K, L} that does not change in the time domain, and A2 is the attribute element {C t ,S t ,F t }, T1 and T2 are the transformation matrices of the time-invariant attribute and the time-varying attribute respectively. The feature set of the time-invariant environmental attribute factor of the time-invariant information storage network is {x0,x1,...,x T }; The forward calculation setting module normalizes the feature data and controls the data dimension between 0 and 1. The number of layers of the temporal information storage network model is set to be consistent with the time point, and each layer has input and output. The first layer of the temporal information storage network model inputs the time-invariant environmental attribute factor feature x0, and its impact result is continuously transmitted to the prediction of each moment. The input of the n+1th layer of the temporal information storage network model layer, the image passes through the deep feature extraction network and is fused with the time-varying environmental attribute factor as the input of the n+1th layer of the temporal information storage network model, and is input into the layers of the temporal information storage network model in the order of different time series shooting, and the forward calculation of the temporal information storage network unit model is set.

7. The black soldier fly feces treatment system based on intelligent control according to claim 6 is characterized in that: The forward calculation setting module specifically includes: Long-term storage unit c t-1 Through the forgetting unit f t To forget and discard information, set the forget unit f t Affected by the external input x at the current moment t , the short-term storage output h of the previous moment t-1 , the long-term storage c of the previous moment t-1 The control expression is as follows: f t =σ(W xf x t +W hf h t-1 +W cf c t-1 +b f ); W xf , W hf , W cf 、b f Respectively represent the external input weight, short-term storage weight, long-term storage weight and forgetting unit bias of the forgetting unit; The external input x at the current time t and the short-term storage output h of the previous moment t-1 Calculate the new information at the current moment By input unit i t Control the new information at the current moment Write long-term storage unit to generate new long-term storage c t , which is expressed as follows: i t =σ(W xi x t +W hi h t-1 +W ci c t-1 +b i ); Among them, i t Subject to x t 、h t-1 、c t-1 control, i t 、c t 、f t , o t They are the input unit, long-term storage unit, forgetting unit, and output unit on the temporal information storage network unit, respectively; where σ(·) represents the S-shaped activation function, · represents the component multiplication, and W is the weight matrix connecting different units; Activate long-term storage unit c t , by the output unit o t Control the currently accumulated storage c t Select some relevant storage to generate the storage h that you are interested in at this moment t , and then store this part to output y t , o t =σ(W xo x t +W ho h t-1 +W co c t-1 +b o ), h t =o t ·tanh(c t ); Among them, the output unit o t Subject to x t 、h t-1 and the current time long storage c t control.

Images