Grain condition data prediction method and equipment

Abstract

The invention discloses a grain condition data prediction method and device, and relates to the technical field of grain, and the method comprises the steps: carrying out the data cleaning processing of original grain condition data, so as to generate a grain condition data optimal attribute set; training the grain condition data optimal attribute set through a radial basis function (RBF) neural network to generate a grain condition data prediction model; and outputting predicted grain condition data within a specified time step length through the grain condition prediction optimal attribute set and the grain condition data prediction model, and performing early warning according to a comparison result of the predicted grain condition data and a preset safe grain storage critical value. According to the invention, the future development or change trend of the grain condition is predicted through the multi-type grain condition data, so that early warning is carried out on the possible dangerous grain condition to ensure the grain storage safety.

Description

Technical Field

[0001] The present invention relates to the field of grain technology, and in particular to a grain condition data prediction method and a grain condition data prediction device. Background Art

[0002] Food security is a "major issue of the country" that concerns the fate of the country and the people's livelihood. It is an important foundation for national security, the primary task of governing the country and maintaining national security, and a major strategic issue of overall importance related to national economic development, social stability and national independence. Grain condition monitoring is an important means of grain storage. Changes in grain conditions will directly affect the quality of grain storage.

[0003] Due to the inaccurate understanding of the grain storage status, problems such as condensation, mildew and insect pests occur, which not only seriously affect the quality of grain, but also cause considerable economic losses. Therefore, the use of effective grain early warning technology is particularly important to prevent these problems and ensure the safe storage of grain.

[0004] Data analysis plays a vital role in grain condition monitoring. With the technological advancement and widespread application of grain condition measurement and control systems, tens of thousands of grain warehouses across the country have accumulated a large amount of grain condition data. By deeply analyzing this data, the state of grain in the warehouse and its changing trends can be predicted in advance, and potential risks can be warned in time to ensure the safety of grain storage. This method has become an important research field for improving grain storage safety.

[0005] At present, the main method for implementing grain condition early warning technology in the market is to use a single temperature measurement to determine whether a single temperature measurement point exceeds a certain threshold to issue an early warning for abnormal temperature points or too fast temperature rise points in the grain pile. However, this early warning method only conducts a simple analysis based on the data currently generated, and the data source for the analysis is relatively single.

[0006] It can be seen that how to predict the future development or change trend of grain conditions through multiple types of grain data, so as to issue early warnings for possible dangerous grain conditions to ensure grain storage safety, has become a technical problem that needs to be urgently solved by technical personnel in this field. Summary of the invention

[0007] The present invention provides a method for predicting grain data. It is used to predict the future development or change trend of grain conditions through multiple types of grain data, so as to give early warning of possible dangerous grain conditions to ensure grain storage safety. The method includes:

[0008] Performing data cleaning on the original grain condition data to generate an optimal attribute set of the grain condition data, wherein the original grain condition data is a data set formed by multiple parameters affecting the grain condition at a specified time;

[0009] The optimal attribute set of the grain condition data is trained by a radial basis function (RBF) neural network to generate a grain condition data prediction model, wherein the center, width and connection weight of the RBF neural network are determined based on a particle swarm optimization (PSO) algorithm;

[0010] Outputting predicted grain condition data within a specified time step through the optimal attribute set for grain condition prediction and the grain condition data prediction model, and issuing an early warning based on a comparison result between the predicted grain condition data and a preset critical value for safe grain storage;

[0011] Among them, the time corresponding to the predicted grain condition data is consistent with the specified time.

[0012] Preferably, the original grain condition data is cleaned to generate the optimal attribute set of grain condition data, specifically:

[0013] The original grain condition data is normalized, and outliers in the normalized original grain condition data are removed;

[0014] Based on time series characteristics, environmental conditions and biological factors, feature screening is performed on the original grain condition data with the outliers removed to generate an optimal attribute set for the grain condition data.

[0015] Preferably, based on time series characteristics, environmental conditions and biological factors, the original grain condition data with the outliers removed is subjected to feature screening to generate the optimal attribute set of the grain condition data, specifically:

[0016] Obtaining the correlation coefficient between each feature in the original grain condition data and the target variable of the predicted grain condition data, and selecting the optimal attribute set of the grain condition data from each feature based on the correlation coefficient;

[0017] Or, the features with the smallest impact on the grain condition data prediction model are deleted from the original grain condition data in sequence, and the remaining features are used as the optimal attribute set of the grain condition data.

[0018] Preferably, the types of the time series features include, but are not limited to: time series features, lag features, and sliding window features;

[0019] The types of environmental conditions include, but are not limited to: temperature, humidity, gas concentration;

[0020] The types of biological factors include but are not limited to: pest index characteristics.

[0021] Preferably, it also includes:

[0022] If there are missing data in the original grain condition data, the missing data is supplemented by a data interpolation algorithm;

[0023] The type of the data interpolation algorithm at least includes time series interpolation based on the time series characteristics.

[0024] Preferably, the fitness function of the PSO algorithm is mean square error, and further includes:

[0025] The center is concentrated near the typical value of the characteristic corresponding to the main influencing factor of the original grain condition data;

[0026] The width is generated based on the variance of the original grain condition data, or is dynamically generated based on the local data distribution around the center;

[0027] The connection weights are generated by a least square method, or are generated by iterative updating according to a gradient descent method.

[0028] Preferably, an early warning is given based on the comparison result between the predicted grain condition data and the preset critical value of safe grain storage, specifically:

[0029] Comparing the predicted grain condition data of each natural day within the specified time step with the critical value of safe grain storage in sequence;

[0030] If the predicted grain condition data of each natural day does not exceed the critical value of safe grain storage, safety information is generated;

[0031] If there is predicted grain condition data that exceeds the critical value for safe grain storage, different types of early warning information are generated according to the range and amount exceeded by the predicted grain condition data.

[0032] Preferably, different types of warning information are generated according to the range and amount exceeded by the predicted grain condition data, specifically:

[0033] When a single slight excess or slight over-limit occurs, a slight warning message will be issued;

[0034] When a single moderate excess or moderate excess of the standard occurs, a moderate warning message will be issued;

[0035] When a single serious excess or serious over-limit occurs, a serious warning message will be issued;

[0036] The range of the slight excess is 110% to 120% of the critical value of safe grain storage;

[0037] The range of moderate excess is 120% to 150% of the critical value of safe grain storage;

[0038] The serious excess range is more than 150% of the critical value of safe grain storage;

[0039] The minor excess refers to the minor excess occurring 3 times or more within the specified time step;

[0040] The moderate excess is defined as the occurrence of 2 or more moderate excesses within the specified time step;

[0041] The serious exceeding of the standard means that the serious exceeding of the standard occurs once or more within the specified time step.

[0042] Preferably, it also includes:

[0043] Regularly sending the predicted grain condition data of each natural day within the time step and the corresponding change trend chart of the grain condition data to the administrator;

[0044] When an early warning message appears, information including the predicted grain condition data exceeding the critical value of safe grain storage and the corresponding natural day is sent to the administrator.

[0045] On the other hand, the present invention provides a computer device, including a memory and a processor, characterized in that a computer program is stored in the memory, and the processor is configured to run the computer program to execute the above-mentioned grain condition data prediction method.

[0046] The present application discloses a method for predicting grain condition data. The method performs data cleaning processing on the original grain condition data to generate an optimal attribute set of grain condition data; trains the optimal attribute set of grain condition data through a radial basis function (RBF) neural network to generate a grain condition data prediction model; outputs predicted grain condition data within a specified time step through the optimal attribute set of grain condition prediction and the grain condition data prediction model, and issues an early warning based on the comparison result between the predicted grain condition data and a preset critical value for safe grain storage. By applying the technical solution of the present application, the future development or change trend of grain condition can be predicted through multiple types of grain condition data, thereby issuing an early warning for possible dangerous grain conditions to ensure grain storage safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solution of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0048] Figure 1 A schematic diagram of a process of predicting grain data of the present invention;

[0049] Figure 2 A schematic diagram of the principle of a grain condition data prediction method of the present invention;

[0050] Figure 3 It is a structural schematic diagram of a grain condition data prediction device of the present invention;

[0051] Figure 4 It is a schematic structural diagram of the electronic device of the present invention. DETAILED DESCRIPTION

[0052] As described in the background technology, using a single temperature measurement point in a single temperature measurement to determine whether it exceeds a certain threshold to issue an early warning for abnormal temperature points or points where the temperature rises too quickly in the grain pile is merely a simple analysis based on the data currently generated. Moreover, the data source for the analysis is relatively single and cannot predict the future development or changing trend of the grain situation.

[0053] In order to solve the above problems, an embodiment of the present application proposes a grain condition data prediction method, which predicts the future development or change trend of grain conditions through multiple types of grain condition data, thereby issuing early warnings for possible dangerous grain conditions to ensure grain storage safety.

[0054] In order to further explain the technical idea of ​​the present invention, the technical solution of the present invention is now described in combination with specific application scenarios.

[0055] On the one hand, the present invention provides a method for predicting grain data, such as Figure 1-2 ,include:

[0056] Step S101: performing data cleaning processing on the original grain condition data to generate an optimal attribute set of the grain condition data, wherein the original grain condition data is a data set formed by multiple parameters affecting the grain condition at a specified time;

[0057] The optimal attribute set refers to a subset selected from all available attributes (features) in machine learning and data mining tasks, which can make the model perform best on a specific task. Parameters affecting grain condition refer to quantitative indicators of various factors that can affect the state (grain condition) of grain during storage, such as temperature, humidity, gas composition, and microbial conditions.

[0058] In this step, sufficient types of original grain condition data are obtained to predict the future development or change trend of grain condition through multiple types of grain condition data. Therefore, in this step, multiple parameters affecting grain condition are collected to increase the types of grain condition data in the data set.

[0059] In specific implementation, multiple parameters affecting grain conditions are collected at specified times (such as 0:00, 8:00, 16:00, etc. every day), a data set formed by the collected parameters is obtained, and data cleaning processing is performed on the data in the data set to obtain the optimal attribute set of grain condition data.

[0060] As an optional embodiment, the step S101 is specifically:

[0061] Step S1011: normalizing the original grain condition data, and removing abnormal values ​​from the normalized original grain condition data;

[0062] Step S1012: Based on time series characteristics, environmental conditions and biological factors, feature screening is performed on the original grain condition data with the abnormal values ​​removed to generate an optimal attribute set for the grain condition data.

[0063] In step S1011-step S1012, normalization processing is used to accelerate model convergence and improve model accuracy. Removing outliers after normalization processing can improve data representativeness, stabilize the model training process, and improve model generalization ability; feature screening based on time series characteristics, environmental conditions, and biological factors to generate the optimal attribute set of the grain data can improve the accuracy of grain data analysis.

[0064] In the specific implementation, the original grain condition data is first normalized to the range of [0,1] or [-1,1] to ensure that different features are on the same scale, then outliers are selected and removed, and finally the data is screened according to time series characteristics, environmental conditions and biological factors. The screened features form the optimal attribute set of grain condition data.

[0065] As another optional embodiment, the step S1012 is specifically as follows:

[0066] Obtaining the correlation coefficient between each feature in the original grain condition data and the target variable of the predicted grain condition data, and selecting the optimal attribute set of the grain condition data from each feature based on the correlation coefficient;

[0067] Or, the features with the smallest impact on the grain condition data prediction model are deleted from the original grain condition data in sequence, and the remaining features are used as the optimal attribute set of the grain condition data.

[0068] In this step, feature selection can reduce noise and make the model more focused on important features, thereby improving prediction accuracy. At the same time, reducing the number of features can speed up model training and prediction and reduce the demand for memory and computing resources.

[0069] In specific implementation, the above-mentioned original grain condition data can be used to generate the optimal attribute set of grain condition data through filtering method, packaging method or embedded method.

[0070] Filtering method: Calculate the correlation coefficient between each feature in the original grain condition data and the target variable of the predicted grain condition data, filter out the required amount of data from high to low based on the correlation, and obtain the optimal attribute set of the grain condition data.

[0071] Wrapper approach: Start with the full raw grain data set and gradually remove features that have the least impact on model performance until the required number of features is reached.

[0072] Embedded method: By adding an L1 regularization term to the loss function, the coefficients of some features become zero, thereby achieving feature selection.

[0073] It should be noted that starting from an empty set and gradually adding features that have the greatest improvement on the model performance until the stopping condition is reached also falls within the scope of protection of this application.

[0074] As another optional embodiment, the types of the time series features include but are not limited to: time series features, lag features, and sliding window features;

[0075] The types of environmental conditions include, but are not limited to: temperature, humidity, gas concentration;

[0076] The types of biological factors include but are not limited to: pest index characteristics.

[0077] In this step, the timestamp of each data point should be taken into consideration when processing time series features. To handle inconsistent time intervals, resampling and interpolation methods can be used to convert data into fixed time intervals. Lag features are used to capture dynamic changes in time series. Sliding window features are created through moving average, moving standard deviation, etc. Temperature and humidity include the temperature and humidity of the grain itself and the temperature and humidity of the environment in which the grain is located. Gas concentration mainly refers to the content of oxygen and carbon dioxide in the environment in which the grain is located. Pest index features are created through pest data.

[0078] In this way, through the selection of the above features, it is possible to analyze the contribution of each feature in the model and understand which factors have the greatest impact on grain condition changes, thereby improving the accuracy of grain condition monitoring and optimizing grain condition data storage and processing by reducing data redundancy (reducing unnecessary data).

[0079] As another optional embodiment, step S101 further includes:

[0080] Step S1013: if there are missing data in the original grain condition data, the missing data is supplemented by a data interpolation algorithm;

[0081] The type of the data interpolation algorithm at least includes time series interpolation based on the time series characteristics.

[0082] In this step, the original grain condition data is first checked for missing data. If there is missing data, the missing value is supplemented by a data interpolation algorithm. Since the grain condition data involves multiple time series characteristics, such as temperature, humidity, gas concentration, etc., the type of data interpolation algorithm in this step at least includes time series interpolation based on the time series characteristics.

[0083] In this way, data integrity and continuity can be maintained, thereby improving the accuracy of data analysis and model training, and optimizing warehouse management decisions.

[0084] In specific implementation, if the time intervals of the data are inconsistent, resampling can be performed first to convert the data into a fixed time interval; if the time intervals of the data are consistent, forward filling (using the value of the previous time point) or backward filling (using the value of the next time point) can be used for the time characteristics of the data. If the data has obvious seasonality, seasonal interpolation methods can also be used.

[0085] Step S102: training the optimal attribute set of the grain condition data through a radial basis function RBF neural network to generate a grain condition data prediction model, wherein the center, width and connection weight of the RBF neural network are determined based on a particle swarm optimization PSO algorithm;

[0086] In this step, the grain condition data is predicted by a model with strong nonlinear mapping ability and fast training speed. Therefore, in this step, the optimal attribute set of grain condition data is trained by radial basis function RBF neural network to generate a grain condition data prediction model.

[0087] In specific implementation, the optimal attribute set of grain condition data obtained through data cleaning can be used to train the radial basis function (RBF) neural network to generate a grain condition data prediction model.

[0088] As an optional embodiment, the fitness function of the PSO algorithm is a mean square error, and further includes:

[0089] The center is concentrated near the typical value of the characteristic corresponding to the main influencing factor of the original grain condition data;

[0090] The width is generated based on the variance of the original grain condition data, or is dynamically generated based on the local data distribution around the center;

[0091] The connection weights are generated by a least square method, or are generated by iterative updating according to a gradient descent method.

[0092] In this step, the PSO algorithm will try to adjust the model parameters during the optimization process to reduce the mean square error so that the model can better fit all data including outliers. This helps to improve the robustness of the model, that is, the model can still maintain relatively stable performance when facing noise and anomalies in the data.

[0093] Selecting the mean square error function as the fitness function of the PSO algorithm not only meets the goal of minimizing the error, but also can take into account the errors of all data points.

[0094] In specific implementation, the K-means algorithm is used to cluster the data, and the cluster center is selected as the RBF center or some data points are randomly selected as the RBF center; the variance of each RBF center is calculated, and a suitable fixed width is selected or an adaptive width is used to dynamically adjust the width according to the distribution of the data; the least squares method is used to directly calculate the connection weights of the output layer or the gradient descent method (such as stochastic gradient descent SGD, Adam, etc.) is used to iteratively update the connection weights.

[0095] Step S103: outputting predicted grain condition data within a specified time step through the grain condition prediction optimal attribute set and the grain condition data prediction model, and issuing an early warning based on a comparison result between the predicted grain condition data and a preset safe grain storage critical value;

[0096] Among them, the time corresponding to the predicted grain condition data is consistent with the specified time.

[0097] In this step, the predicted value is compared with the critical value of safe grain storage to predict whether there may be dangerous grain conditions. Therefore, this step outputs the predicted grain condition data within the specified time step through the optimal attribute set for grain condition prediction and the grain condition data prediction model, and issues an early warning based on the comparison result between the predicted grain condition data and the preset critical value of safe grain storage.

[0098] In specific implementation, the optimal attribute set for grain condition prediction is input into the grain condition data prediction model to obtain the predicted grain condition data within a specified time step, and the predicted grain condition data is compared with the preset critical value for safe grain storage to obtain the comparison result and issue an early warning.

[0099] As an optional embodiment, the step S103 is specifically:

[0100] Step S1031: comparing the predicted grain condition data of each natural day within the specified time step with the safe grain storage critical value in sequence;

[0101] In this step, in order to make the prediction results accurate, you can use the rolling prediction method, predict one time step at a time, then add the prediction results to the known data, and continue to predict the next time step. The specified time step can be flexibly set according to needs, such as 1 week, 1 month, or 1 quarter.

[0102] In specific implementation, starting from the first natural day within the specified time step, each indicator (such as temperature, humidity, moisture content, etc.) in the predicted grain condition data of the natural day is compared with the corresponding critical value of safe grain storage. For example, for the predicted temperature data of the first day, check whether it exceeds the critical value of safe temperature for grain storage (such as 25°C); similarly, for the predicted humidity data of the first day, check whether it exceeds the critical value of safe humidity (such as 75%).

[0103] Step S1032: if the predicted grain condition data of each natural day does not exceed the critical value of safe grain storage, safety information is generated;

[0104] In this step, when each indicator in the predicted grain situation data of the natural day does not exceed the lowest critical value in the critical value of safe grain storage, it means that there is no trend of dangerous situation, and safety information is generated and output.

[0105] Step S1033: If there is predicted grain condition data exceeding the critical value for safe grain storage, different types of warning information are generated according to the range and amount exceeded by the predicted grain condition data.

[0106] The above steps S1031 to S1033, through layered early warning, assist in accurate decision-making, reasonably allocate resources, improve resource utilization efficiency, and avoid resource waste.

[0107] As another optional embodiment, the step S1033 is specifically:

[0108] When a single slight excess or slight over-limit occurs, a slight warning message will be issued;

[0109] When a single moderate excess or moderate excess of the standard occurs, a moderate warning message will be issued;

[0110] When a single serious excess or serious over-limit occurs, a serious warning message will be issued;

[0111] The range of the slight excess is 110% to 120% of the critical value of safe grain storage;

[0112] The range of moderate excess is 120% to 150% of the critical value of safe grain storage;

[0113] The serious excess range is more than 150% of the critical value of safe grain storage;

[0114] The minor excess refers to the minor excess occurring 3 times or more within the specified time step;

[0115] The moderate excess is defined as the occurrence of 2 or more moderate excesses within the specified time step;

[0116] The serious exceeding of the standard means that the serious exceeding of the standard occurs once or more within the specified time step.

[0117] In this step, the early warning can be divided into two situations, a single exceedance and multiple exceedance (exceeding the standard). For a single exceedance, it is divided into three levels: slight, moderate and severe according to the degree to which the grain data exceeds the critical value of safe grain storage. The slight exceedance range is 110%-120% of the critical value of safe grain storage, the moderate exceedance range is 120%-150% of the critical value of safe grain storage, and the severe exceedance range is 150% of the critical value of safe grain storage. Each time the grain data is monitored to exceed the corresponding level range once, the corresponding level of early warning information is issued, that is, a slight exceedance will issue a slight warning, a moderate exceedance will issue a moderate warning, and a severe exceedance will issue a severe warning.

[0118] For multiple exceedances, the cumulative number of exceedances is considered within the specified time step. A minor exceedance is defined as 3 or more minor exceedances, a moderate exceedance is 2 or more moderate exceedances, and a severe exceedance is 1 or more severe exceedances. When these multiple exceedances occur, warning information of the corresponding level (minor, moderate, severe) is also issued.

[0119] In this way, different types of warnings can enable staff to quickly understand the urgency of the problem, so that they can get proper handling methods, respond in time and reduce the risk of losses.

[0120] During specific implementation, the critical values ​​of safe grain storage of various parameters can be flexibly set according to actual needs.

[0121] As another optional embodiment, the method further includes:

[0122] Regularly sending the predicted grain condition data of each natural day within the time step and the corresponding change trend chart of the grain condition data to the administrator;

[0123] When this step is implemented, a separate trend chart is generated for each grain data indicator. Taking the grain temperature trend chart as an example, the horizontal axis can represent time (natural day) and the vertical axis represents the temperature value. The predicted temperature data for each natural day is marked on the chart and connected with a broken line to clearly show the trajectory of temperature changes throughout the time step. Annotations can be added to the chart, such as the dates when the highest and lowest temperatures occur, and the slope of the temperature change, so that administrators can intuitively understand the temperature fluctuations. Similarly, for other indicators such as humidity and moisture content, their own trend charts are also generated. In the humidity trend chart, the fluctuations in humidity are displayed, and special dates that may affect humidity changes (such as natural days corresponding to periods of heavy rainfall) are marked to help administrators better understand the causes and trends of humidity changes.

[0124] In addition to single-indicator trend graphs, you can also generate comprehensive trend graphs to show the relationship between different grain data indicators. For example, create a dual-axis line graph containing temperature and humidity, with one axis representing temperature and the other representing humidity. In this way, administrators can simultaneously observe the changing trends of temperature and humidity and their correlation on the same graph. If you find a pattern that the humidity increases as the temperature rises, you can better judge the overall changes in grain conditions. For more complex relationships, such as the three-dimensional relationship between temperature, humidity, and pest density, you can use three-dimensional visualization technology (such as a three-dimensional scatter plot or surface plot) to present it.

[0125] When an early warning message appears, information including the predicted grain condition data exceeding the critical value of safe grain storage and the corresponding natural day is sent to the administrator.

[0126] In this step, when the warning is triggered, the predicted grain data that exceeds the critical value of safe grain storage and the corresponding natural day information are accurately extracted from the massive grain data. For each grain condition indicator that triggers the warning, such as temperature, humidity, pest density, etc., obtain its specific data value that exceeds the critical value. For example, if it is a temperature warning, extract the temperature forecast value that exceeds the critical value. If the temperature critical value is 25°C and the actual predicted temperature is 30°C, then record the data of 30°C. At the same time, clarify the natural day corresponding to the data, such as November 10, 2024, so that the administrator can clearly know the time when the problem occurred.

[0127] During the specific implementation, early warning information is sent to the administrator through multiple channels to ensure that the administrator can receive it in time. Common channels include e-mail, text messages, and instant message push within the grain management system. In addition to the above detailed report, some brief descriptions can also be added, such as "This warning includes temperature and humidity abnormalities, please check and take measures in time."

[0128] In this way, by regularly sending the predicted grain condition data and change trend charts for each natural day within the time step to the administrator, the administrator can understand the dynamic changes of the grain condition in advance. When an early warning message appears, the predicted grain condition data and the corresponding natural day information that exceed the critical value of safe grain storage are sent to the administrator, which can ensure that the administrator grasps the key grain condition crisis point in the first place. After receiving the regular grain condition data and change trend charts, the administrator can conduct a comprehensive analysis of the grain condition in combination with historical data and experience. The detailed content in the early warning information can help the administrator accurately locate the problem. By analyzing the past changes in grain conditions and the effects of response measures, the administrator can continuously optimize the grain condition management strategy, quickly determine the possible causes of pest breeding, and formulate targeted pest control plans.

[0129] In summary, a method for predicting grain condition data of the present invention generates an optimal attribute set of grain condition data by performing data cleaning processing on the original grain condition data; trains the optimal attribute set of grain condition data through a radial basis function RBF neural network to generate a grain condition data prediction model; outputs predicted grain condition data within a specified time step through the grain condition prediction optimal attribute set and the grain condition data prediction model, and issues an early warning based on the comparison result between the predicted grain condition data and the preset critical value of safe grain storage. The present invention can predict the future development or change trend of grain condition through multiple types of grain condition data, thereby issuing an early warning for possible dangerous grain conditions to ensure grain storage safety.

[0130] On the other hand, the present invention provides a grain data prediction device, such as Figure 3 As shown, including:

[0131] The processing module 11 is used to perform data cleaning processing on the original grain condition data to generate an optimal attribute set of the grain condition data, wherein the original grain condition data is a data set formed by multiple parameters affecting the grain condition at a specified time;

[0132] A generating module 12 is used to train the optimal attribute set of the grain condition data through a radial basis function RBF neural network to generate a grain condition data prediction model, wherein the center, width and connection weight of the RBF neural network are determined based on a particle swarm optimization PSO algorithm;

[0133] The output module 13 is used to output the predicted grain condition data within a specified time step through the grain condition prediction optimal attribute set and the grain condition data prediction model, and to issue an early warning based on the comparison result between the predicted grain condition data and the preset safe grain storage critical value;

[0134] Among them, the time corresponding to the predicted grain condition data is consistent with the specified time.

[0135] The device of this embodiment can be used to perform Figure 1The technical solution of the method embodiment shown has similar implementation principles and technical effects, which will not be repeated here.

[0136] Preferably, the processing module 11 includes:

[0137] A normalization submodule, used for normalizing the original grain condition data and removing abnormal values ​​in the original grain condition data after the normalization process;

[0138] The screening submodule is used to perform feature screening on the original grain condition data with the abnormal values ​​removed based on time series characteristics, environmental conditions and biological factors to generate an optimal attribute set for the grain condition data.

[0139] Preferably, the screening submodule comprises:

[0140] A screening unit, used for obtaining a correlation coefficient between each feature in the original grain condition data and a target variable of the predicted grain condition data, and screening out an optimal attribute set of the grain condition data from each feature based on the correlation coefficient;

[0141] Or, a deletion unit is used to delete the features with the smallest impact on the grain condition data prediction model from the original grain condition data in sequence, and use the remaining features as the optimal attribute set of the grain condition data.

[0142] Preferably, the types of the time series features include, but are not limited to: time series features, lag features, and sliding window features;

[0143] The types of environmental conditions include, but are not limited to: temperature, humidity, gas concentration;

[0144] The types of biological factors include but are not limited to: pest index characteristics.

[0145] Preferably, the processing module 11 further includes:

[0146] A supplement submodule is used to supplement the missing data by using a data interpolation algorithm if there is missing data in the original grain condition data;

[0147] The type of the data interpolation algorithm at least includes time series interpolation based on the time series characteristics.

[0148] Preferably, the fitness function of the PSO algorithm is mean square error, and the generating module 12 further includes:

[0149] The center is concentrated near the typical value of the characteristic corresponding to the main influencing factor of the original grain condition data;

[0150] The width is generated based on the variance of the original grain condition data, or is dynamically generated based on the local data distribution around the center;

[0151] The connection weights are generated by a least square method, or are generated by iterative updating according to a gradient descent method.

[0152] Preferably, the output module 13 includes:

[0153] A comparison submodule is used to compare the predicted grain condition data of each natural day within the specified time step with the safe grain storage critical value in sequence;

[0154] A safety information generation submodule, for generating safety information if the predicted grain condition data of each natural day does not exceed the critical value of safe grain storage;

[0155] The early warning information generation submodule is used to generate different types of early warning information according to the range and quantity exceeded by the predicted grain condition data if there is predicted grain condition data exceeding the critical value of safe grain storage.

[0156] Preferably, the warning information generation submodule includes:

[0157] A minor warning unit is used to issue a minor warning message when a single minor excess or minor over-limit occurs;

[0158] Moderate warning unit, used to issue moderate warning information when a single moderate excess or moderate excess occurs;

[0159] Serious warning unit, used to issue serious warning information when a single serious excess or serious excess occurs;

[0160] The range of the slight excess is 110% to 120% of the critical value of safe grain storage;

[0161] The range of moderate excess is 120% to 150% of the critical value of safe grain storage;

[0162] The serious excess range is more than 150% of the critical value of safe grain storage;

[0163] The minor excess refers to the minor excess occurring 3 times or more within the specified time step;

[0164] The moderate excess is defined as the occurrence of 2 or more moderate excesses within the specified time step;

[0165] The serious exceeding of the standard means that the serious exceeding of the standard occurs once or more within the specified time step.

[0166] Preferably, the device further comprises:

[0167] The first sending module is used to regularly send the predicted grain condition data of each natural day within the time step and the change trend chart corresponding to the grain condition data to the administrator;

[0168] The second sending module is used to send information containing predicted grain condition data exceeding the critical value of safe grain storage and the corresponding natural day to the administrator when early warning information appears.

[0169] The present invention also provides an electronic device, such as Figure 4 As shown, it includes a memory 110 for storing executable program code; a processor 120, which runs a program corresponding to the executable program code by reading the executable program code stored in the memory 110, so as to execute the method described in any of the above method embodiments.

[0170] For details on the specific execution process of the above steps by the processor 120 and the steps further performed by the processor 120 by running the executable program code, please refer to the present invention. Figure 1 The description of the illustrated embodiment will not be repeated here.

[0171] This electronic device exists in many forms, including but not limited to:

[0172] (1) Mobile communication devices: These devices are characterized by their mobile communication functions and their main purpose is to provide voice and data communications. These terminals include: smart phones (such as iPhones), multimedia phones, functional phones, and low-end phones.

[0173] (2) Ultra-mobile personal computer devices: These devices belong to the category of personal computers, have computing and processing functions, and generally also have mobile Internet access features. These terminals include: PDA, M ID and UMPC devices, such as iPad.

[0174] (3) Portable entertainment devices: These devices can display and play multimedia content. They include audio and video players (such as iPods), handheld game consoles, e-books, smart toys, and portable car navigation devices.

[0175] (4) Server: A device that provides computing services. The server consists of a processor, hard disk, memory, system bus, etc. The server is similar to a general computer architecture, but because it needs to provide highly reliable services, it has higher requirements in terms of processing power, stability, reliability, security, scalability, and manageability.

[0176] (5) Other electronic devices with data interaction functions.

[0177] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method steps described in any of the above method embodiments are implemented.

[0178] The present invention also provides an application program, which is executed to implement the method provided by any method embodiment of the present invention.

[0179] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0180] Each embodiment in this specification is described in a related manner, and the same and similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. For the convenience of description, the description of the above device is divided into various units / modules according to their functions and described separately. Of course, when implementing the present invention, the functions of each unit / module can be implemented in the same or one or more software and / or hardware.

[0181] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for predicting grain data, characterized in that: include: Performing data cleaning on the original grain condition data to generate an optimal attribute set of the grain condition data, wherein the original grain condition data is a data set formed by multiple parameters affecting the grain condition at a specified time; The optimal attribute set of the grain condition data is trained by a radial basis function (RBF) neural network to generate a grain condition data prediction model, wherein the center, width and connection weight of the RBF neural network are determined based on a particle swarm optimization (PSO) algorithm; Outputting predicted grain condition data within a specified time step through the optimal attribute set for grain condition prediction and the grain condition data prediction model, and issuing an early warning based on a comparison result between the predicted grain condition data and a preset critical value for safe grain storage; Among them, the time corresponding to the predicted grain condition data is consistent with the specified time.

2. The method according to claim 1, characterized in that The original grain data is cleaned to generate the optimal attribute set of grain data, specifically: The original grain condition data is normalized, and outliers in the normalized original grain condition data are removed; Based on time series characteristics, environmental conditions and biological factors, feature screening is performed on the original grain condition data with the outliers removed to generate an optimal attribute set for the grain condition data.

3. The method according to claim 2, characterized in that Based on time series characteristics, environmental conditions and biological factors, the original grain condition data without the abnormal values ​​is subjected to feature screening to generate the optimal attribute set of the grain condition data, specifically: Obtaining the correlation coefficient between each feature in the original grain condition data and the target variable of the predicted grain condition data, and selecting the optimal attribute set of the grain condition data from each feature based on the correlation coefficient; Or, the features with the smallest impact on the grain condition data prediction model are deleted from the original grain condition data in sequence, and the remaining features are used as the optimal attribute set of the grain condition data.

4. The method according to claim 2, characterized in that The types of time series features include but are not limited to: time series features, lag features, and sliding window features; The types of environmental conditions include, but are not limited to: temperature, humidity, gas concentration; The types of biological factors include but are not limited to: pest index characteristics.

5. The method according to any one of claims 3 or 4, characterized in that Also includes: If there are missing data in the original grain condition data, the missing data is supplemented by a data interpolation algorithm; The type of the data interpolation algorithm at least includes time series interpolation based on the time series characteristics.

6. The method according to claim 1, characterized in that The fitness function of the PSO algorithm is the mean square error, and also includes: The center is concentrated near the typical value of the characteristic corresponding to the main influencing factor of the original grain condition data; The width is generated based on the variance of the original grain condition data, or is dynamically generated based on the local data distribution around the center; The connection weights are generated by a least square method, or are generated by iterative updating according to a gradient descent method.

7. The method according to claim 1, characterized in that According to the comparison result between the predicted grain situation data and the preset critical value of safe grain storage, an early warning is given, specifically: Comparing the predicted grain condition data of each natural day within the specified time step with the critical value of safe grain storage in sequence; If the predicted grain condition data of each natural day does not exceed the critical value of safe grain storage, safety information is generated; If there is predicted grain condition data that exceeds the critical value for safe grain storage, different types of early warning information are generated according to the range and amount exceeded by the predicted grain condition data.

8. The method according to claim 7, characterized in that Different types of early warning information are generated according to the range and amount exceeded by the predicted grain data, specifically: When a single slight excess or slight over-limit occurs, a slight warning message will be issued; When a single moderate excess or moderate excess of the standard occurs, a moderate warning message will be issued; When a single serious excess or serious over-limit occurs, a serious warning message will be issued; The range of the slight excess is 110% to 120% of the critical value of safe grain storage; The range of moderate excess is 120% to 150% of the critical value of safe grain storage; The serious excess range is more than 150% of the critical value of safe grain storage; The minor excess refers to the minor excess occurring 3 times or more within the specified time step; The moderate excess is defined as the occurrence of 2 or more moderate excesses within the specified time step; The serious exceeding of the standard means that the serious exceeding of the standard occurs once or more within the specified time step.

9. The method according to claim 8, characterized in that Also includes: Regularly sending the predicted grain condition data of each natural day within the time step and the corresponding change trend chart of the grain condition data to the administrator; When an early warning message appears, information including the predicted grain condition data exceeding the critical value of safe grain storage and the corresponding natural day is sent to the administrator.

10. A computer device comprising a memory and a processor, characterized in that The memory stores a computer program, and the processor is configured to run the computer program to execute the grain condition data prediction method described in any one of claims 1 to 9.

Images