Temperature and humidity prediction method and prediction system for tobacco processing environment

Abstract

The invention provides a temperature and humidity prediction method and system for a tobacco processing environment. The method comprises the following steps: acquiring a first data set of a workshop of a tobacco processing environment, wherein the first data set comprises workshop external meteorological data, workshop air conditioner operation parameters, operation conditions of process section equipment in the workshop, first temperature and humidity data outside the workshop within a first preset duration, and second temperature and humidity data corresponding to a process section in the workshop; the first data set is input into a hybrid prediction model subjected to training testing, a prediction result output by the hybrid prediction model is obtained, and the prediction result comprises the temperature and humidity value of the workshop in the tobacco processing environment within a second preset duration after the current moment. According to the scheme, fusion analysis of the time sequence features and the spatial features can be carried out, and the influence of each factor on temperature and humidity prediction is more fully considered, so that the prediction precision of the workshop temperature and humidity in the tobacco processing environment can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of tobacco shredded tobacco environment monitoring technology, and in particular to a temperature and humidity prediction method and prediction system for a tobacco processing environment. Background Art

[0002] In the tobacco production process, moisture content is a key process parameter, directly impacting the quality of the finished tobacco. During tobacco processing, the temperature and humidity of the processing environment are factors that influence moisture content. Therefore, accurate monitoring of this temperature and humidity, as well as forecasting of future future environmental temperature and humidity levels, is essential.

[0003] Currently, the prediction of temperature and humidity in the processing environment mainly relies on weather forecasts and temperature and humidity sensors in the factory. The temperature and humidity of the tobacco processing environment in the future are estimated based on manual experience. This requires high professional skills of the estimation operators and is not suitable for ordinary operators. In addition, manual estimation is still prone to errors. Summary of the Invention

[0004] In view of this, the purpose of the embodiments of the present application is to provide a method and system for predicting temperature and humidity in a tobacco processing environment, which can improve the problem of inaccurate temperature and humidity prediction in a tobacco processing environment.

[0005] To achieve the above technical objectives, the technical solutions adopted in this application are as follows:

[0006] In a first aspect, an embodiment of the present application provides a method for predicting temperature and humidity in a tobacco processing environment, the method comprising:

[0007] Obtaining a first data set of a workshop in a tobacco processing environment, the first data set including meteorological data outside the workshop, operating parameters of air conditioning in the workshop, operating conditions of equipment in a process section in the workshop, first temperature and humidity data outside the workshop within a first preset time period, and second temperature and humidity data inside the workshop corresponding to the process section;

[0008] The first data set is input into a trained and tested hybrid prediction model to obtain a prediction result output by the hybrid prediction model, wherein the prediction result includes the temperature and humidity values of the workshop in the tobacco processing environment within a second preset time period after the current moment. The hybrid prediction model includes a time series feature processing module, a spatial feature processing module, a feature fusion module and a regression prediction module, wherein the time series feature processing module is used to extract a time series feature vector from the first data set; the spatial feature processing module is used to extract a spatial feature vector from the first data set; the feature fusion module is used to fuse the time series feature vector and the spatial feature vector to obtain a fused feature vector; and the regression prediction module is used to perform prediction based on the fused feature vector to obtain the prediction result.

[0009] In conjunction with the first aspect, in some optional implementations, inputting the first data set into a trained hybrid prediction model to obtain a prediction result output by the hybrid prediction model includes:

[0010] The time series feature processing module in the hybrid prediction model extracts features from the workshop external meteorological data, workshop air conditioning operating parameters, operating conditions of process section equipment in the workshop, first temperature and humidity data outside the workshop within a first preset time period, and second temperature and humidity data corresponding to the process section in the workshop in the first data set to obtain a time series feature vector. The time series feature processing module is a bidirectional LSTM module.

[0011] A spatial feature processing module in the hybrid prediction model is used to extract features from the second temperature and humidity data corresponding to the process section in the workshop in the first data set, the location topology of the process section equipment in the workshop, and the location distribution of the sensors that collect the second temperature and humidity data to obtain a spatial feature vector. The spatial feature processing module is a graph convolutional network module.

[0012] By means of a feature fusion module in the hybrid prediction model, the temporal feature vector and the spatial feature vector are subjected to feature fusion based on an attention mechanism to obtain a fused feature vector, wherein the feature fusion module is an attention network module;

[0013] The fused feature vector is mapped into temperature and humidity values through the regression prediction module in the hybrid prediction model to obtain the prediction result, and the regression prediction module is a fully connected layer network module.

[0014] In conjunction with the first aspect, in some optional embodiments, before obtaining the first data set of a workshop in the tobacco processing environment, the method further includes:

[0015] Acquire a second data set of a workshop in a tobacco processing environment, the second data set including weather data outside the workshop, operating parameters of air conditioning in the workshop, operating conditions of equipment in a process section in the workshop, first temperature and humidity data outside the workshop, and second temperature and humidity data inside the workshop corresponding to the process section;

[0016] Inputting the second data set into the constructed hybrid prediction model, and performing segmented training on the hybrid prediction model;

[0017] The trained hybrid prediction model is tested and optimized to obtain a trained and tested hybrid prediction model.

[0018] In conjunction with the first aspect, in some optional implementations, inputting the second data set into the constructed hybrid prediction model and performing segmented training on the hybrid prediction model includes:

[0019] Based on the training set in the second data set, the bidirectional LSTM module in the hybrid prediction model is trained using the MSE loss function, and the other modules in the hybrid prediction model except the time series feature processing module are frozen;

[0020] All modules in the hybrid prediction model are unfrozen, and based on the training set, all modules of the hybrid prediction model are trained using a hybrid loss function to obtain a trained hybrid prediction model.

[0021] In combination with the first aspect, in some optional implementations, the MSE loss function L1 is:

[0022]

[0023] Wherein, N is the number of training samples of the second data set; α is the temperature prediction weight coefficient; β is the humidity prediction weight coefficient; i is an integer ranging from 1 to N; represents the i-th measured temperature value; represents the i-th predicted temperature value; represents the i-th measured humidity value; represents the i-th predicted humidity value;

[0024] The hybrid loss function L2 is:

[0025]

[0026] Where γ1 refers to the weight coefficient of the mean absolute error; y ture Refers to the measured values of temperature and humidity; y pred Refers to the predicted value of temperature and humidity; w t Refers to the time weight coefficient; t refers to the current time step; T refers to the total time step; ΔH refers to the humidity change rate; the subscript ture represents the measured value, and the subscript pred represents the predicted value; wherein, the hybrid loss function L2 satisfies the thermodynamic constraint equation and is expressed as:

[0027]

[0028] In the formula, k1 represents the temperature weight coefficient; k2 represents the air conditioning refrigeration weight coefficient; k3 represents the equipment heat generation weight coefficient; C ext Indicates the ambient temperature outside the workshop; C in Indicates the temperature inside the workshop; Q ac Indicates the air conditioning cooling capacity in the workshop air conditioning operating parameters; P eq Indicates the heat generated by the equipment in the process section of the workshop during its operation.

[0029] In conjunction with the first aspect, in some optional implementations, the trained hybrid prediction model is tested and optimized to obtain a trained and tested hybrid prediction model, including:

[0030] Based on the test set in the second data set, the trained hybrid prediction model is tested and parameters are optimized to obtain a hybrid prediction model after training and testing, wherein the optimized parameters include the learning rate, batch size and regularization parameter of the hybrid prediction model.

[0031] In conjunction with the first aspect, in some optional implementations, the method further includes:

[0032] If the temperature and humidity values in the predicted result are not within the pre-planned temperature and humidity range, the temperature and humidity control system is controlled to adjust the temperature and humidity values corresponding to the process section in the workshop so that the temperature and humidity values in the predicted result re-predicted based on the adjusted temperature and humidity values are within the temperature and humidity range.

[0033] In conjunction with the first aspect, in some optional implementations, the method further includes:

[0034] If the temperature and humidity values in the prediction result are not within the pre-planned temperature and humidity range, an early warning prompt will be issued.

[0035] In a second aspect, an embodiment of the present application further provides a prediction system, which includes a processor and a memory coupled to each other, wherein a computer program is stored in the memory. When the computer program is executed by the processor, the prediction system executes the method as described above.

[0036] The invention adopting the above technical solution has the following advantages:

[0037] In the technical solution provided by this application, a temporal feature processing module extracts temporal feature vectors from the first dataset, thereby capturing the lagged effects of meteorological parameters and temperature and humidity. A spatial feature processing module extracts spatial feature vectors from the first dataset, thereby capturing the gradient conduction of temperature and humidity across different spatial regions. Furthermore, the first dataset contains a wide variety of data types that are highly correlated with the predicted temperature and humidity. This allows for a more comprehensive consideration of the impact of various factors on the predicted temperature and humidity, thereby improving the accuracy of temperature and humidity predictions for tobacco processing workshops. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The present application may be further illustrated by the non-limiting embodiments provided in the accompanying drawings. It should be understood that the following drawings illustrate only certain embodiments of the present application and are therefore not to be construed as limiting the scope of the present application. It is understood that a person skilled in the art can derive other relevant drawings from these drawings without inventive effort.

[0039] Figure 1 A schematic flow chart of a method for predicting temperature and humidity in a tobacco processing environment provided in an embodiment of the present application.

[0040] Figure 2 A schematic diagram of the network structure of the hybrid prediction model provided in an embodiment of the present application. DETAILED DESCRIPTION

[0041] The present application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that similar or identical parts in the drawings or descriptions are numbered the same. Implementations not shown or described in the drawings are known to those of ordinary skill in the art. In the description of this application, the terms "first," "second," etc. are used solely to distinguish descriptions and are not to be construed as indicating or implying relative importance.

[0042] Please refer to Figure 1 and Figure 2 The present invention provides a method for predicting the temperature and humidity of a tobacco processing environment. The method may include the following steps:

[0043] Step 110, obtaining a first data set of a tobacco processing workshop environment, the first data set including weather data outside the workshop, operating parameters of the workshop air conditioning, operating conditions of equipment in a process section of the workshop, first temperature and humidity data outside the workshop within a first preset time period, and second temperature and humidity data inside the workshop corresponding to the process section;

[0044] Step 120: input the first data set into a hybrid prediction model that has been trained and tested to obtain a prediction result output by the hybrid prediction model, wherein the prediction result includes the temperature and humidity values of the workshop in the tobacco processing environment within a second preset time period after the current moment. The hybrid prediction model includes a time series feature processing module, a spatial feature processing module, a feature fusion module and a regression prediction module, wherein the time series feature processing module is used to extract a time series feature vector from the first data set; the spatial feature processing module is used to extract a spatial feature vector from the first data set; the feature fusion module is used to fuse the time series feature vector and the spatial feature vector to obtain a fused feature vector; and the regression prediction module is used to perform prediction based on the fused feature vector to obtain the prediction result.

[0045] In the above-described embodiment, the temporal feature processing module extracts temporal feature vectors from the first dataset, thereby capturing the lagged effects of meteorological parameters and temperature and humidity. The spatial feature processing module extracts spatial feature vectors from the first dataset, thereby capturing the gradient conduction of temperature and humidity across different spatial regions. Furthermore, the first dataset contains a wide variety of data types, which are highly correlated with the predicted temperature and humidity. This allows for a more comprehensive consideration of the impact of various factors on the predicted temperature and humidity, thereby improving the accuracy of temperature and humidity predictions for tobacco processing workshops. This eliminates the need for manual predictions and improves the accuracy of inaccurate temperature and humidity predictions for tobacco processing environments.

[0046] The following is a detailed description of the various steps in the temperature and humidity prediction method for the tobacco processing environment:

[0047] In step 110, the method for obtaining the various types of data in the first dataset can be flexibly determined based on actual circumstances. For example, weather data outside the workshop can be obtained from a weather station or weather website. This weather data may include weather conditions, temperature, humidity, and other data at the current geographic location within a certain period of time in the future.

[0048] Workshop air conditioning operating parameters can include, but are not limited to, air conditioning cooling capacity Q, supply air volume R, return air enthalpy h1, and supply air enthalpy h2, where Q = 1.2 × R × (h1 - h2). Supply air volume R can be acquired through the air conditioning PLC system; return air enthalpy h1 can be calculated using a temperature and humidity sensor and a psychrometric chart; and h2 can be calculated using air conditioning setting parameters. In other words, workshop air conditioning operating parameters are acquired using conventional methods.

[0049] The operating status of the process section equipment in the workshop may include parameters such as the start and stop status of the equipment, the heat generation of the equipment in the process section, etc., and the method for obtaining various parameters is conventional and will not be repeated here.

[0050] The first temperature and humidity data outside the workshop may include temperature data and humidity data, which may be collected by temperature and humidity sensors deployed in the environment outside the workshop.

[0051] The second temperature and humidity data corresponding to the process section in the workshop may include temperature data and humidity data, which can be collected by temperature and humidity sensors deployed at the location of the process section in the workshop. The number of temperature and humidity sensors can be multiple and arranged at intervals, and the specific number can be flexibly set according to the data situation. The first preset time length is the time length at one end before the current moment, which can be flexibly set according to the actual situation. For example, the first preset time length can be 1 minute, 5 minutes, 30 minutes, etc. before the current moment, which is not specifically limited here.

[0052] In the first dataset, all data types have undergone preprocessing, including data cleaning, missing value filling, and normalization. Data cleaning uses the 3σ principle to eliminate outliers and mark data during equipment downtime as invalid, effectively filtering out abnormal values.

[0053] Missing values can be filled using linear interpolation (for time series data) and KNN (K-Nearest Neighbors) algorithm.

[0054] Normalization can use Min-Max normalization of external meteorological data, various temperature and humidity data, and Z-Score standardization of equipment parameters.

[0055] In addition, the first data set can be sliced into time series using a sliding window. The window length can be flexibly set according to actual conditions, such as 10 seconds, 30 seconds, 1 minute, etc.

[0056] In step 120, the second preset time period may be 5 minutes, 10 minutes, 30 minutes, etc., and may be flexibly set according to actual conditions. The prediction result may include temperature and humidity data and status information indicating whether the temperature and humidity data are abnormal.

[0057] Temperature and humidity in the predicted results are factors that influence tobacco moisture content. If these predictions indicate anomalies, the tobacco processing environment's temperature and humidity can be adjusted in advance. This prevents these anomalies from causing abnormal tobacco moisture content, which can affect tobacco quality. Furthermore, workshop temperature and humidity can be used as control parameters for tobacco drying operations, improving control accuracy.

[0058] In this embodiment, in step 120, the first data set is input into the trained hybrid prediction model to obtain a prediction result output by the hybrid prediction model, including:

[0059] A time series feature processing module in the hybrid prediction model is used to extract features from the workshop external meteorological data, workshop air conditioning operating parameters, operating conditions of process section equipment in the workshop, first temperature and humidity data outside the workshop within a first preset time period, and second temperature and humidity data corresponding to the process section in the workshop, in the first data set, to obtain a time series feature vector. The time series feature processing module is a bidirectional LSTM (Long Short-Term Memory) module.

[0060] A spatial feature processing module in the hybrid prediction model is used to extract features from the second temperature and humidity data corresponding to the process section in the workshop in the first data set, the location topology of the process section equipment in the workshop, and the location distribution of the sensors that collect the second temperature and humidity data to obtain a spatial feature vector. The spatial feature processing module is a graph convolutional network (GCN) module.

[0061] The feature fusion module in the hybrid prediction model performs feature fusion on the temporal feature vector and the spatial feature vector based on the attention mechanism to obtain a fused feature vector, wherein the feature fusion module is an attention network module, and the attention network module can be a multi-head attention network module, such as a four-head attention network module;

[0062] The fused feature vector is mapped into temperature and humidity values through the regression prediction module in the hybrid prediction model to obtain the prediction result, and the regression prediction module is a fully connected layer network module.

[0063] In this embodiment, the bidirectional LSTM module is a double-layer stacked LSTM, each layer of which can have 64 hidden units, which is used to capture the long-term and short-term time series dependencies of time-related data such as external meteorological changes, equipment operating status, first temperature and humidity data outside the workshop, and second temperature and humidity data corresponding to the process section in the workshop.

[0064] The GCN module can be a three-layer graph convolutional network, where the number of sensors collecting temperature and humidity is the number of nodes, and the inverse of the spatial distance between nodes is the edge weight. The spatial feature processing module / GCN module can be used to analyze the secondary temperature and humidity data corresponding to the process section within the workshop, the location topology of the process section equipment within the workshop, and the location distribution of temperature and humidity sensors. It can model the spatial topology of the workshop sensor network (such as heat conduction between upstream and downstream process sections) and output a spatial feature matrix.

[0065] The feature fusion module can be a 4-head attention network module, which can dynamically allocate the weights of external meteorological influences and internal spatial conduction in the workshop, and output a fused feature vector.

[0066] The regression prediction module is a fully connected layer network module that can convert the fused feature vector into temperature and humidity values, which are the predicted temperature and humidity values.

[0067] In this embodiment, before obtaining the first data set of a workshop in a tobacco processing environment, the method further includes:

[0068] Acquire a second data set of a workshop in a tobacco processing environment, the second data set including weather data outside the workshop, operating parameters of air conditioning in the workshop, operating conditions of equipment in a process section in the workshop, first temperature and humidity data outside the workshop, and second temperature and humidity data inside the workshop corresponding to the process section;

[0069] Inputting the second data set into the constructed hybrid prediction model, and performing segmented training on the hybrid prediction model;

[0070] The trained hybrid prediction model is tested and optimized to obtain a trained and tested hybrid prediction model.

[0071] The second data set can be relevant data collected from the tobacco processing environment over a period of time in the past (such as a quarter, a year, etc.). The various types of data in the second data set are preprocessed data, and the preprocessing method is similar to that of the first data set. The window length of the time series slice of the second data set can be 1 hour, 6 hours, etc., which can be larger than the window length of the first data set. The sliding step length of the window can be 10 minutes, 30 minutes, etc., which will not be repeated here. The sample size of the second data set can exceed 10,000, and the specific number can be flexibly determined according to actual conditions.

[0072] In this embodiment, the second data set is input into the constructed hybrid prediction model, and the hybrid prediction model is trained in segments, including:

[0073] Based on the training set in the second data set, the bidirectional LSTM module in the hybrid prediction model is trained using an MSE (Mean Squared Error) loss function, and the other modules in the hybrid prediction model except the time series feature processing module are frozen;

[0074] All modules in the hybrid prediction model are unfrozen, and based on the training set, all modules of the hybrid prediction model are trained using a hybrid loss function to obtain a trained hybrid prediction model.

[0075] In this embodiment, the MSE loss function L1 is:

[0076]

[0077] Wherein, N is the number of training samples of the second data set; α is the temperature prediction weight coefficient; β is the humidity prediction weight coefficient; i is an integer ranging from 1 to N; represents the i-th measured temperature value; represents the i-th predicted temperature value; represents the i-th measured humidity value; represents the i-th predicted humidity value;

[0078] The hybrid loss function L2 is:

[0079]

[0080] Where γ1 refers to the weight coefficient of the mean absolute error; y ture Refers to the measured values of temperature and humidity; y pred Refers to the predicted value of temperature and humidity; w t Refers to the time weight coefficient; t refers to the current time step; T refers to the total time step; ΔH refers to the humidity change rate; the subscript ture represents the measured value, and the subscript pred represents the predicted value; wherein, the hybrid loss function L2 satisfies the thermodynamic constraint equation and is expressed as:

[0081]

[0082] In the formula, k1 represents the temperature weight coefficient; k2 represents the air conditioning refrigeration weight coefficient; k3 represents the equipment heat generation weight coefficient; C ext Indicates the ambient temperature outside the workshop; C in Indicates the temperature inside the workshop; Q ac Indicates the air conditioning cooling capacity in the workshop air conditioning operating parameters; P eq Indicates the heat generated by the equipment in the process section of the workshop during its operation.

[0083] In this embodiment, segmented training is adopted. The bidirectional LSTM module is trained first, and other modules are frozen. The bidirectional LSTM is forced to focus on the extraction and modeling of time series features, avoiding gradient conflicts caused by the joint training of multiple modules in the parameter initialization stage, ensuring that the time series feature processing module obtains a more stable parameter space initialization, which can make the square error amplify the gradient of large deviation samples, accelerate the model to capture the main trend features, and match the temperature-humidity weight distribution with the humidity sensitivity requirements of the silk drying process.

[0084] Using the MSE loss function in the initial training stage can fully demonstrate its adaptability to continuous value prediction tasks and effectively capture the local fluctuation details of time series data; switching to the hybrid loss function in the second stage can simultaneously optimize the point accuracy, trend robustness and probability distribution consistency of the prediction results.

[0085] The staged training strategy decomposes the exploration of high-dimensional parameter space into a sequential optimization process. In the first stage, fixing the parameters of non-critical modules reduces the probability of the model falling into a local optimum. In the second stage, global fine-tuning is performed with good initial parameter values for each module, improving overall convergence stability. The staged unfreezing strategy essentially forms an implicit curriculum learning mechanism. By gradually releasing model complexity, it adheres to a more "simple to complex" learning principle than direct end-to-end training, reducing the risk of overfitting in scenarios with limited training data.

[0086] After the timing module is pre-trained, the unfrozen attention mechanism, convolution module, etc. can perform cross-scale feature fusion based on the established timing representation, enabling the model to simultaneously capture short-term fluctuations, medium- and long-term periodicity, and the coupling of external influencing factors.

[0087] In this embodiment, the trained hybrid prediction model is tested and optimized to obtain a trained and tested hybrid prediction model, including:

[0088] Based on the test set in the second data set, the trained hybrid prediction model is tested and parameters are optimized to obtain a hybrid prediction model after training and testing, wherein the optimized parameters include the learning rate, batch size and regularization parameter of the hybrid prediction model.

[0089] In this embodiment, the optimized hyperparameter (such as learning rate) search method can use Bayesian optimization or grid search. In this way, the optimized hybrid prediction model is conducive to improving the accuracy of prediction.

[0090] In this embodiment, the method further includes:

[0091] If the temperature and humidity values in the predicted result are not within the pre-planned temperature and humidity range, the temperature and humidity control system is controlled to adjust the temperature and humidity values corresponding to the process section in the workshop so that the temperature and humidity values in the predicted result re-predicted based on the adjusted temperature and humidity values are within the temperature and humidity range.

[0092] The temperature and humidity range can be flexibly set according to actual conditions, and no specific restrictions are made here. If the temperature and humidity values are not within the pre-planned temperature and humidity range, the PID (Proportional Integral Derivative) algorithm can be used based on the difference between the temperature and humidity values and the median of the temperature and humidity range to control the temperature and humidity control system to adjust the temperature and humidity values corresponding to the process section in the workshop, so that the adjusted temperature and humidity are within the planned temperature and humidity range. That is, if the estimated temperature and humidity do not meet the production requirements, the temperature and humidity of the processing environment can be adjusted in advance to improve the response speed of the temperature and humidity adjustment, which is conducive to the temperature and humidity of the processing environment to maintain production requirements in the long term and avoid abnormal temperature and humidity in the processing environment in the future.

[0093] The temperature and humidity control system can include air conditioning, humidifiers, dehumidifiers, etc. Of course, humidifiers and dehumidifiers can be integrated into the air conditioning to achieve flexible adjustment of the temperature and humidity of the processing environment.

[0094] In this embodiment, the method further includes:

[0095] If the temperature and humidity values in the prediction result are not within the pre-planned temperature and humidity range, an early warning prompt will be issued.

[0096] As an example, a graded warning system can be used:

[0097]

[0098] In this embodiment, the use of graded warnings can provide more detailed warning content, facilitating subsequent cause analysis and improvement by engineers. The deep closed loop of prediction-control-warning can achieve a leap in intelligent temperature and humidity control in the silk-making workshop.

[0099] An embodiment of the present application also provides a prediction system, which includes a processor and a memory coupled to each other, wherein a computer program is stored in the memory. When the computer program is executed by the processor, the prediction system executes the temperature and humidity prediction method for the tobacco processing environment as described above.

[0100] The prediction system can include a display screen for real-time monitoring of ambient temperature and humidity. This screen displays the ambient temperature and humidity of the outside environment, the workshop, and the process sections, presenting these graphically. This helps silk production and management personnel conduct multi-dimensional analysis of key indicators. The prediction system analyzes the impact of equipment operation, air conditioning status, and external climate on the production environment's temperature and humidity. Data modeling also analyzes the correlation between workshop temperature and humidity and the moisture content of each silk leaf, providing a basis for regulating ambient temperature and humidity.

[0101] It should be noted that those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the prediction system described above can refer to the corresponding processes of each step in the aforementioned method, and will not be elaborated here.

[0102] Through the description of the above implementation methods, those skilled in the art can clearly understand that the present application can be implemented through hardware or by means of software plus a necessary general hardware platform. Based on this understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each implementation scenario of the present application.

[0103] In the embodiments provided in the present application, it should be understood that the disclosed systems and methods can also be implemented in other ways. The system and method embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the systems, methods and computer program products according to the multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of code, and a part of the module, program segment or code includes one or more executable instructions for implementing the specified logical function. It should also be noted that each box in the block diagram and / or flowchart, and the combination of the boxes in the block diagram and / or flowchart can be implemented by a dedicated hardware-based system that performs the specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions. In addition, the functional modules in the various embodiments of the present application can be integrated together to form an independent part, or each module can exist alone, or two or more modules can be integrated to form an independent part.

[0104] The above description is merely an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, various modifications and variations of the present application are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A method for predicting temperature and humidity in a tobacco processing environment, characterized in that: The method comprises: Obtaining a first data set of a workshop in a tobacco processing environment, the first data set including meteorological data outside the workshop, operating parameters of air conditioning in the workshop, operating conditions of equipment in a process section in the workshop, first temperature and humidity data outside the workshop within a first preset time period, and second temperature and humidity data inside the workshop corresponding to the process section; The first data set is input into a trained and tested hybrid prediction model to obtain a prediction result output by the hybrid prediction model, wherein the prediction result includes the temperature and humidity values of the workshop in the tobacco processing environment within a second preset time period after the current moment. The hybrid prediction model includes a time series feature processing module, a spatial feature processing module, a feature fusion module and a regression prediction module, wherein the time series feature processing module is used to extract a time series feature vector from the first data set; the spatial feature processing module is used to extract a spatial feature vector from the first data set; the feature fusion module is used to fuse the time series feature vector and the spatial feature vector to obtain a fused feature vector; and the regression prediction module is used to perform prediction based on the fused feature vector to obtain the prediction result.

2. The method according to claim 1, characterized in that Inputting the first data set into a trained hybrid prediction model to obtain a prediction result output by the hybrid prediction model includes: The time series feature processing module in the hybrid prediction model extracts features from the workshop external meteorological data, workshop air conditioning operating parameters, operating conditions of process section equipment in the workshop, first temperature and humidity data outside the workshop within a first preset time period, and second temperature and humidity data corresponding to the process section in the workshop in the first data set to obtain a time series feature vector. The time series feature processing module is a bidirectional LSTM module. A spatial feature processing module in the hybrid prediction model is used to extract features from the second temperature and humidity data corresponding to the process section in the workshop in the first data set, the location topology of the process section equipment in the workshop, and the location distribution of the sensors that collect the second temperature and humidity data to obtain a spatial feature vector. The spatial feature processing module is a graph convolutional network module. By means of a feature fusion module in the hybrid prediction model, the temporal feature vector and the spatial feature vector are subjected to feature fusion based on an attention mechanism to obtain a fused feature vector, wherein the feature fusion module is an attention network module; The fused feature vector is mapped into temperature and humidity values through the regression prediction module in the hybrid prediction model to obtain the prediction result, and the regression prediction module is a fully connected layer network module.

3. The method according to claim 1, characterized in that Before acquiring the first data set of the workshop of the tobacco processing environment, the method further comprises: Acquire a second data set of a workshop in a tobacco processing environment, the second data set including weather data outside the workshop, operating parameters of air conditioning in the workshop, operating conditions of equipment in a process section in the workshop, first temperature and humidity data outside the workshop, and second temperature and humidity data inside the workshop corresponding to the process section; Inputting the second data set into the constructed hybrid prediction model, and performing segmented training on the hybrid prediction model; The trained hybrid prediction model is tested and optimized to obtain a trained and tested hybrid prediction model.

4. The method according to claim 3, characterized in that Inputting the second data set into the constructed hybrid prediction model, and performing segmented training on the hybrid prediction model, including: Based on the training set in the second data set, the bidirectional LSTM module in the hybrid prediction model is trained using the MSE loss function, and the other modules in the hybrid prediction model except the time series feature processing module are frozen; All modules in the hybrid prediction model are unfrozen, and based on the training set, all modules of the hybrid prediction model are trained using a hybrid loss function to obtain a trained hybrid prediction model.

5. The method according to claim 4, characterized in that The MSE loss function L1 is: Wherein, N is the number of training samples of the second data set; α is the temperature prediction weight coefficient; β is the humidity prediction weight coefficient; i is an integer ranging from 1 to N; represents the i-th measured temperature value; represents the i-th predicted temperature value; represents the i-th measured humidity value; represents the i-th predicted humidity value; The hybrid loss function L2 is: Where γ1 refers to the weight coefficient of the mean absolute error; y ture Refers to the measured values of temperature and humidity; y pred Refers to the predicted value of temperature and humidity; w t Refers to the time weight coefficient; t refers to the current time step; T refers to the total time step; ΔH refers to the humidity change rate; the subscript ture represents the measured value, and the subscript pred represents the predicted value; wherein, the hybrid loss function L2 satisfies the thermodynamic constraint equation and is expressed as: In the formula, k1 represents the temperature weight coefficient; k2 represents the air conditioning refrigeration weight coefficient; k3 represents the equipment heat generation weight coefficient; C ext Indicates the ambient temperature outside the workshop; C in Indicates the temperature inside the workshop; Q ac Indicates the air conditioning cooling capacity in the workshop air conditioning operating parameters; P eq Indicates the heat generated by the equipment in the process section of the workshop during its operation.

6. The method according to claim 3, characterized in that The trained hybrid prediction model is tested and optimized to obtain a trained and tested hybrid prediction model, including: Based on the test set in the second data set, the trained hybrid prediction model is tested and parameters are optimized to obtain a hybrid prediction model after training and testing, wherein the optimized parameters include the learning rate, batch size and regularization parameter of the hybrid prediction model.

7. The method according to claim 1, characterized in that The method further comprises: If the temperature and humidity values in the predicted result are not within the pre-planned temperature and humidity range, the temperature and humidity control system is controlled to adjust the temperature and humidity values corresponding to the process section in the workshop so that the temperature and humidity values in the predicted result re-predicted based on the adjusted temperature and humidity values are within the temperature and humidity range.

8. The method according to claim 7, characterized in that The method further comprises: If the temperature and humidity values in the prediction result are not within the pre-planned temperature and humidity range, an early warning prompt will be issued.

9. A prediction system, characterized in that The prediction system includes a processor and a memory coupled to each other, wherein a computer program is stored in the memory. When the computer program is executed by the processor, the prediction system executes the method according to any one of claims 1 to 8.

Images