CN114021832A - LSTM algorithm-based grain mildew prediction method and system - Google Patents

LSTM algorithm-based grain mildew prediction method and system Download PDF

Info

Publication number
CN114021832A
CN114021832A CN202111346834.4A CN202111346834A CN114021832A CN 114021832 A CN114021832 A CN 114021832A CN 202111346834 A CN202111346834 A CN 202111346834A CN 114021832 A CN114021832 A CN 114021832A
Authority
CN
China
Prior art keywords
mildew
grain
granary
probability
data
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202111346834.4A
Other languages
Chinese (zh)
Inventor
张红伟
许景
李聪聪
尤赛赛
马雪迪
李晓辉
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Anhui University
Original Assignee
Anhui University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Anhui University filed Critical Anhui University
Priority to CN202111346834.4A priority Critical patent/CN114021832A/en
Publication of CN114021832A publication Critical patent/CN114021832A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/02—Agriculture; Fishing; Forestry; Mining

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Strategic Management (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Molecular Biology (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Tourism & Hospitality (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Agronomy & Crop Science (AREA)
  • Primary Health Care (AREA)
  • Mining & Mineral Resources (AREA)
  • Marine Sciences & Fisheries (AREA)
  • Animal Husbandry (AREA)
  • Development Economics (AREA)
  • Game Theory and Decision Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention relates to a grain mildew prediction method and system based on an LSTM algorithm. The grain mildew prediction method comprises the steps of firstly establishing a grain mildew prediction model based on an LSTM algorithm, then preprocessing collected detection data, and dividing a monitoring data set into a training set and a testing set. And then predicting the temperature data and the humidity data in a future preset time period II in the granary by using the trained grain mildew prediction model so as to predict the mildew probability of the grain. And finally, judging whether the mildew probability is greater than a preset probability, and predicting the mildew of the grain inside the granary when the mildew probability is greater than the preset probability. And when the mildew probability is smaller than the preset probability, predicting that the grains in the granary do not mildew. The grain mildew prediction method can effectively predict the mildew probability of the grain in the granary, and can save manpower and material resources compared with the traditional technology.

Description

LSTM algorithm-based grain mildew prediction method and system
Technical Field
The invention relates to the technical field of neural networks, in particular to a grain mildew prediction method and system based on an LSTM algorithm; LSTM (long short-term memory), namely: long and short term memory neural networks.
Background
Grain is an important article related to the survival of the national people, and a large amount of sufficient grain is stored to play a vital role in stabilizing the development of national economy. The granary is basic equipment for storing grains, and has certain requirements on heat insulation, moisture and water resistance, insect prevention, ventilation conditions and the like during the design of the granary. And the grain mildew is an important factor influencing the grain storage safety.
At present, most grain processing enterprises mainly rely on the naked eyes to observe whether grains are mildewed or not, and manually pick out mildewed grains. The chemical analysis method has higher accuracy and reliability in detecting the grain mildew, and the existing color sorter and other equipment also detect the grain mildew. However, neither the chemical analysis method nor the color sorter needs to consume a large amount of manpower and material resources when the grain mildew detection is carried out, and the grain mildew prediction in the granary cannot be effectively carried out.
Disclosure of Invention
Based on this, the invention provides a grain mildew prediction method and system based on an LSTM algorithm, aiming at the problem that the prior art can not effectively predict the mildew of grains in a granary.
The invention discloses a grain mildew prediction method based on an LSTM algorithm, which comprises the following steps of S1-S6.
And S1, establishing a grain mildew prediction model based on an LSTM algorithm.
And S2, preprocessing the acquired monitoring data to obtain a monitoring data set. The monitoring data comprises temperature data and humidity data of the interior of the granary within a preset time period in the past.
And S3, dividing the monitoring data set into a training set and a testing set.
And S4, setting initial parameters of the grain mildew prediction model, and training the grain mildew prediction model according to the data of the training set.
And S5, predicting the temperature data and the humidity data in a future preset time period II in the granary by using the trained grain mildew prediction model so as to predict the mildew probability of the grain.
And S6, judging whether the mildew probability is greater than a preset probability, and predicting the mildew of the grain inside the granary when the mildew probability is greater than the preset probability. And when the mildew probability is smaller than the preset probability, predicting that the grains in the granary do not mildew.
In one embodiment, the grain mildew prediction model includes an input layer, an output layer, and LSTM cell units. The LSTM cell unit includes an input gate, a forgetting gate, and an output gate.
In one embodiment, the forgetting door ftFor deciding the information to discard from the cell state, the expression of the process is:
ft＝σ(Wf·[ht-1,xt]+bf)
where σ is Sigmoid function and the output range is [0,1 ]]An output of 0 indicates forgetting, and an output of 1 indicates leaving. WfThe forgetting gate weight value is obtained. bfTo forget the door bias. h ist-1And predicting output for the last moment. x is the number oftThe feature amount is input for the present time.
In one embodiment, the input gate itFor deciding the value to be updated by the Sigmoid layer, the expression formula of the process is:
it＝σ(Wi·[ht-1,xt]+bi)
in the formula, WiRepresenting the input gate weight. biBiasing the input gate.
In one of the embodiments, the output gate otFor determining the output state at the present moment, the processThe expression formula is:
ot＝σ(Wo·[ht,xt]+bo)
in the formula, WoRepresenting the output gate weight. boIs an output gate bias. h istAnd predicting and outputting the current time.
In one embodiment, in step S2, the preprocessing is: and (5) carrying out abnormal value elimination and standardization processing on the monitoring data. The method for removing the abnormal values of the monitoring data comprises the following steps:
suppose that n times of monitoring is carried out in the granary, and the obtained ith monitoring value is MiWherein i is 1,2, … n. The accumulated continuous 3 times of monitoring values are respectively Ti-1，TiAnd Ti+1Wherein i is 2,3, … n-1. The ith monitored jitter characteristic h is calculatedi：
hi＝|2×Mi-(Ti-1+Ti+1)|
According to the ith monitored jump characteristic hiSequentially calculating the average value of the jitter characteristics of the ith monitoring
Figure BDA0003354475670000035
And a beat eigenvalue mean square error δ.
Calculating the relative difference Q of the ith monitoringi. If QiIf the value is more than 3, the data monitored in the ith time is taken as an abnormal value to be removed.
In one embodiment, the average value of the ith monitored jitter characteristics
Figure BDA0003354475670000031
The expression formula of (a) is:
Figure BDA0003354475670000032
in one embodiment, the expression formula of the mean square error δ of the ith monitored jitter characteristic value is as follows:
Figure BDA0003354475670000033
in one embodiment, the relative difference Q of the i-th monitoringiThe expression formula of (a) is:
Figure BDA0003354475670000034
the invention also discloses a grain mildew prediction system which adopts any grain mildew prediction method based on the LSTM algorithm. This grain prediction system that mildenes and rot includes:
the data acquisition module is used for acquiring monitoring data in the granary in real time. The monitoring data consists of temperature data and humidity data in a granary within a past preset time period.
The data processing module is used for preprocessing the monitoring data to obtain a monitoring data set. The data processing module is further configured to divide the monitoring data set into a training set and a test set.
The initial parameter setting module is used for setting initial parameters of the grain mildew prediction model.
The training module is used for training the grain mildew prediction model according to the data of the training set.
The prediction module is used for predicting the temperature data and the humidity data in a future preset time period II in the granary by using the trained grain mildew prediction model so as to predict the mildew probability of the grain. The processing module is further configured to determine whether the probability of mildew is greater than a predetermined probability. And when the mildew probability is greater than the preset probability, predicting that the grain in the granary is mildewed. And when the mildew probability is smaller than the preset probability, predicting that the grains in the granary do not mildew.
And the alarm module is used for generating alarm information when the grains in the granary are predicted to mildew.
Compared with the prior art, the grain mildew prediction method and system based on the LSTM algorithm have the following beneficial effects:
1. according to the grain mildew prediction method, the activity state of microorganisms is monitored by utilizing the temperature and humidity changes of the granary, the grain mildew prediction model based on the LSTM algorithm is established, historical monitoring data of the temperature and the humidity in the granary are utilized, the grain mildew prediction model is effectively trained, the grain mildew probability of the granary is effectively predicted by utilizing the trained grain mildew prediction model, and therefore the cost of manpower and material resources consumed when grain mildew is directly detected by means of a chemical analysis method or a color sorter and the like is saved.
2. The grain mildew prediction method predicts the mildew probability of the grain in the granary, can provide data reference for a manager of the granary, and can enable the manager to pay more attention to the granary when the mildew probability is higher, so that the granary management is facilitated, the mildew occurrence of the grain is reduced, and the loss of grain enterprises due to large-scale mildew is avoided.
3. This grain prediction system that mildenes and rot adopts and to carry out real-time detection and automatic recording to the inside humiture data of granary through data acquisition module, and the monitoring data who reachs the collection is handled to rethread data processing module, then the grain after the application training of prediction module mildenes and rot prediction model predicts the humiture of granary inside in future predetermined time quantum to the probability of mildenes and rot of the inside grain of granary is predicted out, and then whether the inside grain of prediction granary takes place mildenes and rot. The system can automatically collect, process and analyze the temperature and humidity in the granary, finally automatically predict whether grain in the granary mildews or not, and remind granary managers through the alarm system, so that the granary managers can be concerned and managed intensively, and further the granary can be prevented from suffering in the bud and loss caused by the granary is reduced.
Drawings
FIG. 1 is a flow chart of a grain mildew prediction method based on an LSTM algorithm in embodiment 1 of the present invention;
FIG. 2 is a schematic diagram of the principle of the LSTM algorithm in example 1 of the present invention;
fig. 3 is a block diagram of a system for predicting food mildew in embodiment 2 of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the term "or/and" includes any and all combinations of one or more of the associated listed items.
Referring to fig. 1, the present invention discloses a grain mildew prediction method based on LSTM algorithm, which includes the following steps, i.e., steps S1 to S6.
And S1, establishing a grain mildew prediction model based on an LSTM algorithm.
In this embodiment, the grain mildew prediction model includes an input layer, an output layer, and LSTM cell units. The LSTM cell unit may include an input gate, a forgetting gate, and an output gate.
A Recurrent Neural Network (RNN) is a Neural Network that takes sequence data as input, each hidden layer neuron includes a feedback input, and the input of the hidden layer neuron in the Recurrent Neural Network includes both the input at the current time and the input over a period of time. Therefore, the recurrent neural network is characterized by having "memory".
Referring to fig. 2, the LSTM algorithm is an improvement of the recurrent neural network, and has a unique memory and forgetting module, which can learn the sample timing characteristics. The LSTM algorithm solves the problems of gradient loss and explosion of the RNN in the training process of a Back Propagation Through Time (BPTT), and can fully utilize the Time dependency of historical sample information.
The LSTM algorithm determines the information to discard from the cell state through a forget gate, and the process is expressed as:
ft＝σ(Wf·[ht-1,xt]+bf)
wherein σ is Sigmoid function, and the output range is [0,1 ]]An output of 0 indicates forgetting, and an output of 1 indicates leaving. WfThe forgetting gate weight value is obtained. bfTo forget the door bias. h ist-1And predicting output for the last moment. x is the number oftThe feature amount is input for the present time. The input gate determines the value to be updated through the Sigmoid layer, and the tanh layer creates a new candidate value vector C't，C′tWill be added to the state (tanh refers to mapping real numbers to [ -1,1 [)]Hyperbolic tangent function). Through the above two layers of new information being stored in the unit state, the process expression is:
it＝σ(Wi·[ht-1,xt]+bi)
C′t＝tanh(Wc·[ht-1,xt]+bc)
in the formula, WiRepresenting the input gate weight. biBiasing the input gate. WcIs the tan h layer weight. bcIs configured for a tanh layer.
Ct-1Is updated to CtIs operated as Ct-1And ftMultiplication, new candidate value is it×C′tThe expression of the operation to update the state of the old cell is:
Ct＝ft×Ct-1+it×C′t
finally, an output gate otFor determining the output state at the present time, the cell state C at the present time is first determinedtExcited by a tanh layer, ht-1And xtAnd obtaining the updated weight value through a Sigmoid layer. The result is that the current cell state is weighted after being excited by the tanh layer to obtain the state at the current moment. Detailed description of the preferred embodimentThe expression of (a) is:
ot＝σ(Wo·[ht,xt]+bo)
ht＝ot·tanh(Ct)
in the formula, WoIs the output gate weight. boIs an output gate bias. h istAnd predicting and outputting the current time. As can be seen from fig. 2, the forgetting gate and the input gate respectively determine the unit state forgetting part and the candidate value part through the Sigmoid function, that is, the output range of the Sigmoid function is [0,1 ]]An output of 0 indicates forgetting, and an output of 1 indicates leaving.
In this embodiment, the steps of the LSTM training algorithm may be as follows:
(1) forward computing neurons ft、it、ct、ot、htThe output value of (d);
(2) calculating error terms at each moment through backward propagation along time, and then propagating the error terms to an upper layer;
(3) the gradient of each weight is calculated according to the corresponding error term.
And S2, preprocessing the acquired monitoring data to obtain a monitoring data set. The monitoring data comprises temperature data and humidity data of the interior of the granary within a preset time period in the past.
The humiture of grain is an important index for guaranteeing the grain safe storage, and the loss of grain in the storage process can be reduced to the maximum extent only by timely and accurately measuring the temperature measurement data of each layer of the grain pile, analyzing the grain storage condition according to the detected temperature data, making a decision and taking measures. In this embodiment, the humiture of grain in the granary can be recorded by adopting a plurality of existing humiture recorders. Every humiture record appearance can include the probe, through burying the probe and inserting in the grain heap to monitor the humiture of grain. It should be noted here that, because plants breathe to consume oxygen and decompose organic matters, and release energy, the respiration of grains is different at different depths and levels of the grain heap, so that the temperature and humidity of grains at different depths and levels can be different. Therefore, when the humiture is monitored on grain, the probes of the humiture recorders can be respectively embedded into different depth levels of the grain heap according to the actual grain storage condition, so that the temperature and the humidity of the grain at different depth levels can be measured, and then the preliminary calculation processing can be performed on a plurality of groups of humiture data at different depth levels, for example, the average value processing or the weighted average value processing is performed, so that the average temperature of the temperature in the grain bin can be accurately obtained.
In this embodiment, since the data is derived from actual data measured in the granary, the acquired data needs to be preprocessed, and the preprocessing is: and (5) carrying out abnormal value elimination and standardization processing on the monitoring data. The data exception eliminating treatment of the acquired data by a '3 delta' method before training comprises the following steps:
suppose that a monitoring point is monitored n times, and the obtained ith monitoring value is MiWherein i is 1,2, … n. The accumulated continuous 3 times of monitoring values are respectively Ti-1，TiAnd Ti+1Wherein i is 2,3, … n-1. The ith monitored jitter characteristic h is calculatedi：
hi＝|2×Mi-(Ti-1+Ti+1)|
According to the ith monitored jump characteristic hiSequentially calculating the average value of the jitter characteristics of the ith monitoring
Figure BDA0003354475670000071
And a beat eigenvalue mean square error δ. In the embodiment, the average value of the ith monitored jitter characteristics
Figure BDA0003354475670000072
The expression of (d) may be:
Figure BDA0003354475670000073
the expression formula of the mean square error delta of the jitter characteristic value of the ith monitoring can be as follows:
Figure BDA0003354475670000081
thereby calculating the relative difference Q of the ith monitoringi. If QiIf the value is more than 3, the data monitored in the ith time is taken as an abnormal value to be removed. In the present embodiment, the relative difference Q of the i-th monitoringiThe expression of (d) may be:
Figure BDA0003354475670000082
and S3, dividing the monitoring data set into a training set and a testing set.
And S4, setting initial parameters of the grain mildew prediction model, and training the grain mildew prediction model according to the data of the training set.
And S5, predicting the temperature data and the humidity data in a future preset time period II in the granary by using the trained grain mildew prediction model so as to predict the mildew probability of the grain.
In this embodiment, the main reason why the food is mildewed is that a large amount of minute bacteria are metabolized and decomposed and are propagated in a large amount in a suitable air environment. The micro-bacteria can also be called as microorganisms, and generally have the characteristics of large quantity and small size, and also have very tenacious adaptability and very fast propagation and metabolism speed. Therefore, in the process from the growth of grains to the harvest, processing, transportation and storage of grains, microorganisms of different varieties can be propagated through various paths from different sources, and then are continuously gathered on the surfaces of the grains. When the granary is in an environment with proper temperature and humidity, the microorganisms can accelerate nutrition metabolism activities, and further, the grains are mildewed.
Various microorganisms can generate heat when carrying out metabolic activity, so that the temperature of the granary is increased, the activity state of the microorganisms is monitored by utilizing the temperature change of the granary and combining the humidity change, and the purpose of predicting the grain mildew is achieved by establishing a grain mildew prediction model based on an LSTM algorithm by utilizing monitoring data.
And S6, judging whether the mildew probability is greater than a preset probability, and predicting the mildew of the grain inside the granary when the mildew probability is greater than the preset probability. And when the mildew probability is smaller than the preset probability, predicting that the grains in the granary do not mildew.
The prediction probability may be set according to an empirical value, and the prediction probability of the present embodiment may be preferably set to 50%, that is: and when the mildew probability is more than 50%, judging that the granary is mildewed.
In summary, compared with the prior art, the grain mildew prediction method based on the LSTM algorithm provided by the invention has the following advantages:
1. according to the grain mildew prediction method, the activity state of microorganisms is monitored by utilizing the temperature and humidity changes of the granary, the grain mildew prediction model is effectively trained by establishing the LSTM algorithm-based grain mildew prediction model and utilizing the historical monitoring data of the temperature and the humidity in the granary, and then the grain mildew probability of the granary is effectively predicted by utilizing the trained grain mildew prediction model.
2. The grain mildew prediction method predicts the mildew probability of the grain in the granary, can provide data reference for a manager of the granary, and can enable the manager to pay more attention to the granary when the mildew probability is higher, so that the granary management is facilitated, the mildew occurrence of the grain is reduced, and the loss of grain enterprises due to large-scale mildew is avoided.
Example 2
Referring to fig. 3, the present embodiment provides a system for predicting grain mildew, which employs the method for predicting grain mildew based on LSTM algorithm in embodiment 1. This grain prediction system that mildenes and rot includes:
the data acquisition module is used for acquiring monitoring data in the granary in real time. The monitoring data consists of temperature data and humidity data in a granary within a past preset time period. In this embodiment, the data acquisition module may include a plurality of detection devices. The detection devices can adopt the existing sensing devices, are arranged in the granary and can detect temperature information and humidity information in the granary.
The data processing module is used for preprocessing the monitoring data to obtain a monitoring data set. The data processing module is further configured to divide the monitoring data set into a training set and a test set.
The initial parameter setting module is used for setting initial parameters of the grain mildew prediction model.
The training module is used for training the grain mildew prediction model according to the data of the training set.
The prediction module is used for predicting the temperature data and the humidity data in a future preset time period II in the granary by using the trained grain mildew prediction model so as to predict the mildew probability of the grain. The processing module is further configured to determine whether the probability of mildew is greater than a predetermined probability. And when the mildew probability is greater than the preset probability, predicting that the grain in the granary is mildewed. And when the mildew probability is smaller than the preset probability, predicting that the grains in the granary do not mildew.
The alarm module is used for generating alarm information when predicting that the grains in the granary are mildewed. The alarm module in the embodiment can comprise an acousto-optic alarm lamp, and the acousto-optic alarm lamp can be arranged around the working environment of a granary manager and can effectively remind the granary manager in time, so that the attention to and the management of the granary are strengthened.
The technical features of the embodiments described above may be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the embodiments described above are not described, but should be considered as being within the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above examples are merely illustrative of several embodiments of the present invention, and the description thereof is more specific and detailed, but not to be construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention. Therefore, the protection scope of the present invention should be subject to the appended claims.

Claims (10)

1. A grain mildew prediction method based on an LSTM algorithm is characterized by comprising the following steps:
s1, establishing a grain mildew prediction model based on an LSTM algorithm;
s2, preprocessing the collected monitoring data to obtain a monitoring data set; the monitoring data comprise temperature data and humidity data in a granary within a past preset time period I;
s3, dividing the monitoring data set into a training set and a testing set;
s4, setting initial parameters of the grain mildew prediction model, and training the grain mildew prediction model according to the data of the training set;
s5, predicting temperature data and humidity data in a future preset time period II in the granary by using the trained grain mildew prediction model so as to predict the mildew probability of grains;
s6, judging whether the mildew probability is greater than a preset probability, and predicting the mildew of the grain inside the granary when the mildew probability is greater than the preset probability; and when the mildew probability is smaller than the preset probability, predicting that the grains in the granary do not mildew.
2. The LSTM algorithm based grain mildew prediction method of claim 1 wherein the grain mildew prediction model comprises an input layer, an output layer, and LSTM cell units; the LSTM cell unit comprises an input gate, a forgetting gate and an output gate.
3. The LSTM algorithm based grain mildew prediction method of claim 2, wherein the forgetting gate ftFor deciding the information to discard from the cell state, the expression of the process is:
ft＝σ(Wf·[ht-1,xt]+bf)
where σ is Sigmoid function and the output range is [0,1 ]]The output is 0 to indicate forgetting, and the output is 1 to indicate retention; wfThe forgetting gate weight value is obtained; bfBiasing for a forget gate; h ist-1Predicting output for the last moment; x is the number oftThe feature amount is input for the present time.
4. The LSTM algorithm based grain mildew prediction method of claim 3, wherein input gate itFor deciding the value to be updated by the Sigmoid layer, the expression formula of the process is:
it＝σ(Wi·[ht-1,xt]+bi)
in the formula, WiRepresenting the input gate weight; biBiasing the input gate.
5. The LSTM algorithm based grain mildew prediction method of claim 4, wherein the output gate otFor determining the output state at the current moment, the expression formula of the process is as follows:
ot＝σ(Wo·[ht,xt]+bo)
in the formula, WoRepresenting the weight value of the output gate; boBiasing the output gate; h istAnd predicting and outputting the current time.
6. The LSTM algorithm-based grain mildew prediction method according to claim 1, wherein in step S2, the preprocessing is: sequentially removing abnormal values and standardizing the monitoring data; the method for removing the abnormal values of the monitoring data comprises the following steps:
assuming that n times of monitoring is carried out in the granary, the obtained ith monitoring value is MiWherein i is 1,2, … n; the accumulated continuous 3 times of monitoring values are respectively Ti-1，TiAnd Ti+1Wherein i is 2,3, … n-1; the ith monitor is calculatedRun-out characteristic h ofi：
hi＝|2×Mi-(Ti-1+Ti+1)|
According to the ith monitored jump characteristic hiSequentially calculating the average value of the jitter characteristics of the ith monitoring
Figure FDA0003354475660000024
And the mean square error δ of the jitter eigenvalue;
calculating the relative difference Q of the ith monitoringi(ii) a If QiIf the value is more than 3, the data monitored in the ith time is taken as an abnormal value to be removed.
7. The LSTM algorithm based grain mildew prediction method of claim 6, wherein the ith monitored beat feature mean value
Figure FDA0003354475660000021
The expression formula of (a) is:
Figure FDA0003354475660000022
8. the LSTM algorithm-based grain mildew prediction method of claim 7, wherein the ith monitored jitter eigenvalue mean square error δ is expressed by the formula:
Figure FDA0003354475660000023
9. the LSTM algorithm based grain mildew prediction method of claim 8, wherein the ith monitored relative difference QiThe expression formula of (a) is:
Figure FDA0003354475660000031
10. a system for predicting grain mildew, which adopts the LSTM algorithm-based grain mildew prediction method according to any one of claims 1 to 9; the grain mildew prediction system comprises:
the data acquisition module is used for acquiring monitoring data in the granary in real time; the monitoring data consists of temperature data and humidity data in a granary within a past preset time period I;
the data processing module is used for preprocessing the monitoring data to obtain a monitoring data set; the data processing module is also used for dividing the monitoring data set into a training set and a testing set;
the initial parameter setting module is used for setting initial parameters of the grain mildew prediction model;
the training module is used for training the grain mildew prediction model according to the data of the training set;
the prediction module is used for predicting the temperature data and the humidity data in the granary within a preset time period II in the future by using the trained grain mildew prediction model so as to predict the mildew probability of the grain; the processing module is also used for judging whether the mildew probability is greater than a preset probability; when the mildew probability is larger than the preset probability, predicting that the grains in the granary mildew; when the mildew probability is smaller than the preset probability, predicting that the grains in the granary do not mildew; and
and the alarm module is used for generating alarm information when the grains in the granary are predicted to mildew.
CN202111346834.4A 2021-11-15 2021-11-15 LSTM algorithm-based grain mildew prediction method and system Pending CN114021832A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111346834.4A CN114021832A (en) 2021-11-15 2021-11-15 LSTM algorithm-based grain mildew prediction method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111346834.4A CN114021832A (en) 2021-11-15 2021-11-15 LSTM algorithm-based grain mildew prediction method and system

Publications (1)

Publication Number Publication Date
CN114021832A true CN114021832A (en) 2022-02-08

Family

ID=80064030

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111346834.4A Pending CN114021832A (en) 2021-11-15 2021-11-15 LSTM algorithm-based grain mildew prediction method and system

Country Status (1)

Country Link
CN (1) CN114021832A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116993273A (en) * 2023-09-27 2023-11-03 中国标准化研究院 Grain depot operation and maintenance management system based on data analysis
CN117035203A (en) * 2023-10-10 2023-11-10 中国标准化研究院 Grain depot inventory management system based on big data

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116993273A (en) * 2023-09-27 2023-11-03 中国标准化研究院 Grain depot operation and maintenance management system based on data analysis
CN116993273B (en) * 2023-09-27 2023-12-15 中国标准化研究院 Grain depot operation and maintenance management system based on data analysis
CN117035203A (en) * 2023-10-10 2023-11-10 中国标准化研究院 Grain depot inventory management system based on big data
CN117035203B (en) * 2023-10-10 2023-12-29 中国标准化研究院 Grain depot inventory management system based on big data

Similar Documents

Publication Publication Date Title
CN110263846B (en) Fault diagnosis method based on fault data deep mining and learning
CN112731159B (en) Method for pre-judging and positioning battery faults of battery compartment of energy storage power station
WO2022126526A1 (en) Battery temperature predication method and system
CN112488395B (en) Method and system for predicting line loss of power distribution network
CN114021832A (en) LSTM algorithm-based grain mildew prediction method and system
CN107992968B (en) Electric energy meter metering error prediction method based on integrated time series analysis technology
CN108304970B (en) Method and system for rapidly predicting storage period of apples and controlled atmosphere storage monitoring system
CN111813084A (en) Mechanical equipment fault diagnosis method based on deep learning
CN112685950B (en) Method, system and equipment for detecting abnormality of ocean time sequence observation data
CN111210128B (en) Wetland early warning method based on artificial intelligence and random self-adaptive threshold
CN112818604A (en) Wind turbine generator risk degree assessment method based on wind power prediction
CN112116162A (en) Power transmission line icing thickness prediction method based on CEEMDAN-QFAOA-LSTM
CN112329990A (en) User power load prediction method based on LSTM-BP neural network
CN110119758A (en) A kind of electricity consumption data abnormality detection and model training method, device
CN116843071B (en) Transportation network operation index prediction method and device for intelligent port
CN111191855B (en) Water quality abnormal event identification and early warning method based on pipe network multi-element water quality time sequence data
CN114519923A (en) Intelligent diagnosis and early warning method and system for power plant
CN117193222A (en) Intelligent quality control system based on industrial Internet of things and big data and control method thereof
CN112990585A (en) Hen laying rate prediction method based on LSTM-Kalman model
CN116930609A (en) Electric energy metering error analysis method based on ResNet-LSTM model
CN111898673A (en) Dissolved oxygen content prediction method based on EMD and LSTM
CN117077870B (en) Water resource digital management method based on artificial intelligence
CN113742178A (en) Network node health state monitoring method based on LSTM
CN115586321A (en) Method, system, memory and equipment for identifying online monitoring data of dissolved gas in oil
CN113868957B (en) Residual life prediction and uncertainty quantitative calibration method under Bayes deep learning

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination