Method and device for identifying the operating status of refrigeration equipment based on deep learning

Abstract

The present invention discloses a method and device for identifying the operating status of refrigeration equipment based on deep learning. The method includes: obtaining real-time status data and historical status data of the refrigeration equipment; constructing a state recognition model for the operating status of the refrigeration equipment, the state recognition model including an interactive correlation module, a difference extraction module, a feature fusion module, and a state recognition module; using the real-time status data and the historical status data as training sets to train the constructed state recognition model; the state recognition module completes the identification of the operating status of the refrigeration equipment. The method combines the real-time status data of the collector and the historical status data of the refrigeration equipment and / or the smart terminal connected to the refrigeration equipment to comprehensively obtain information on the operating status, and integrates the interactive correlation features and differential features of the real-time status data and the historical status data. The recognition accuracy is greatly improved, which is beneficial to the operation and efficiency improvement of the refrigeration equipment system.

Description

Technical Field

[0001] The present invention relates to the field of deep learning technology, and in particular to a method and device for identifying the operating status of a refrigeration device based on deep learning. Background Art

[0002] With the development of intelligent home appliances, the demand for monitoring the operating status and fault warnings of intelligent refrigeration equipment is increasing. Traditional methods for identifying refrigeration operating status suffer from insufficient detection and inaccurate results. Previous identification methods include statistical methods and traditional machine learning methods. However, these methods generally suffer from numerous flaws that lead to inaccurate operating status identification. More specifically, insufficient data may lead to overfitting, ineffective dimensionality reduction, and the constructed neural network model has limited spatiotemporal data features. These issues hinder the accurate understanding of refrigeration equipment operating status, and the industry has yet to propose a more effective solution. Summary of the Invention

[0003] In order to solve at least one of the above-mentioned problems in the prior art, an object of the present invention is to provide a recognition method and device that can improve the efficiency of identifying the operating status of intelligent refrigeration equipment.

[0004] To achieve the above-mentioned object of the invention, one embodiment of the present invention provides a method for identifying the operating status of a refrigeration device based on deep learning, comprising the following steps:

[0005] Acquiring real-time status data of a collector of a refrigeration device, and acquiring historical status data of the refrigeration device and / or a smart terminal connected to the refrigeration device;

[0006] Constructing a state recognition model for the operating state of the refrigeration equipment, wherein the state recognition model includes an interactive correlation module, a difference extraction module, a feature fusion module and a state recognition module;

[0007] Using the real-time status data and the historical status data as training sets to train the constructed status recognition model;

[0008] The interactive correlation module extracts interactive correlation features between the real-time status data and the historical status data, the difference extraction module extracts differential features between the real-time status data and the historical status data, the feature fusion module fuses and splices the interactive correlation features and the differential features to obtain a fused feature vector, and the state recognition module calculates a recognition result based on the fused feature vector to complete the recognition of the operating state of the refrigeration equipment;

[0009] Outputs identification information of the operating status.

[0010] As a further improvement of the present invention, the following steps are also included:

[0011] Preprocessing the real-time status data and the historical status data, the preprocessing step comprising:

[0012] Baseline value calculation and deviation calculation are performed on both the real-time status data and the historical status data to form a deviation matrix.

[0013] As a further improvement of the present invention, the interactive association module includes: a spatiotemporal feature extraction module, a semantic information extraction module, and an interactive association feature extraction module connected in sequence;

[0014] The spatiotemporal feature extraction module is a spatiotemporal feature extraction neural network, the semantic information extraction module obtains semantic features of the real-time status data and the historical status data, and the interactive correlation feature extraction module obtains interactive correlation features of the real-time status data and the historical status data.

[0015] As a further improvement of the present invention, the spatiotemporal feature extraction neural network is a dual-channel convolutional neural network model CNN network and a residual network ResNet;

[0016] Also includes the steps:

[0017] Input the deviation matrix of the real-time state data into the convolutional neural network model CNN network and the residual network ResNet of one channel, and input the deviation matrix of the historical state data into the convolutional neural network model CNN network and the residual network ResNet of another channel;

[0018] The temporal and spatial features of the real-time status data and the temporal and spatial features of the historical status data are extracted.

[0019] As a further improvement of the present invention, the semantic information extraction module adopts BiLSTM and Dropout network models;

[0020] Also includes the steps:

[0021] Inputting the spatiotemporal features of the real-time state data and the spatiotemporal features of the historical state data into the corresponding BiLSTM and Dropout network models respectively;

[0022] Context feature vectors of the real-time status data and the historical status data at different times are output respectively, so as to improve the semantic feature representation and fitting capability of the text.

[0023] As a further improvement of the present invention, the interactive correlation feature extraction module includes a self-attention mechanism model and an interactive attention mechanism model in bidirectional attention, and the interactive correlation feature includes a real-time state data attention feature vector and a historical state data attention feature vector;

[0024] Also includes the steps:

[0025] Inputting the context feature vectors of the real-time state data and the historical state data into the self-attention mechanism model and the interactive attention mechanism model respectively;

[0026] Obtaining the real-time state data attention feature vector including the real-time state data's own weight information and the weight information between the context feature vector of the real-time state data and the historical state data;

[0027] Obtain the historical state data attention feature vector including the weight information of the historical state data itself and the weight information between the context feature vector of the historical state data and the real-time state data.

[0028] As a further improvement of the present invention, the following steps are also included:

[0029] Obtaining sentence vector encoding for each sentence in the real-time state data and the historical state data through a pre-trained BERT model;

[0030] Based on the Transformer model, the attention score calculation of the Transformer model is changed to the difference score calculation;

[0031] The text score is calculated for each of the sentence vector codes through the Transformer model to obtain the differential features.

[0032] As a further improvement of the present invention, the following steps are also included:

[0033] The real-time state data attention feature vector, the historical state data attention feature vector, and the differentiated features are mapped together into a unified multimodal vector space for fusion representation to obtain the fused feature vector.

[0034] As a further improvement of the present invention, the following steps are also included:

[0035] After the fused feature vectors are combined through a fully connected layer, they are output to a Softmax function to calculate the text semantic scores of the real-time status data and the historical status data and their normalized score results to obtain identification information of the operating status of the refrigeration equipment.

[0036] As a further improvement of the present invention, the following steps are also included:

[0037] Obtain real-time status data collected by sensors of refrigeration equipment;

[0038] Obtain historical status data recorded in the log system of the operating system of the refrigeration equipment APP or terminal PAD.

[0039] To achieve one of the above-mentioned objectives, an embodiment of the present invention provides a device for identifying the operating status of a refrigeration device based on deep learning, comprising:

[0040] A data acquisition module, configured to acquire real-time status data of a collector of a refrigeration device, and historical status data of the refrigeration device and / or a smart terminal connected to the refrigeration device;

[0041] A model building module is used to build a state recognition model of the refrigeration equipment operating state, wherein the state recognition model includes an interactive correlation module, a difference extraction module, a feature fusion module and a state recognition module;

[0042] A training module, configured to train a constructed state recognition model using the real-time state data and the historical state data as training sets;

[0043] a model operation module, configured to cause the interactive correlation module to extract interactive correlation features between the real-time status data and the historical status data, the difference extraction module to extract differential features between the real-time status data and the historical status data, the feature fusion module to fuse and combine the correlation features and the differential features to obtain a fused feature vector, and the state recognition module to calculate a recognition result based on the fused feature vector to complete the recognition of the operating state of the refrigeration equipment;

[0044] Output module, outputs identification information of operating status.

[0045] To achieve one of the above-mentioned objectives, an embodiment of the present invention provides an electronic device, including:

[0046] a storage module storing a computer program;

[0047] The processing module can implement the steps in the above-mentioned method for identifying the operating status of the refrigeration equipment based on deep learning when executing the computer program.

[0048] To achieve one of the above-mentioned purposes of the invention, an embodiment of the present invention provides a readable storage medium storing a computer program, which, when executed by a processing module, can implement the steps in the above-mentioned method for identifying the operating status of a refrigeration device based on deep learning.

[0049] Compared with the prior art, the present invention has the following beneficial effects: the method and device for identifying the operating status of refrigeration equipment based on deep learning combine the real-time status data of the collector and the historical status data of the refrigeration equipment and / or the smart terminal connected to the refrigeration equipment, comprehensively obtain the operating status information, and integrate the interactive correlation features and differentiation features of the real-time status data and the historical status data, so as to obtain the best and most effective text features, which not only improves the generalization ability of the network, but also alleviates the gradient vanishing problem caused by increasing the depth of the deep neural network and improves the prediction accuracy. The recognition accuracy is greatly improved, and the operating status of the refrigeration equipment is accurately identified, thereby improving the operating status monitoring efficiency of the refrigeration equipment, reducing the daily operation and maintenance costs, and benefiting the operation and efficiency improvement of the refrigeration equipment system. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a flow chart of a method for identifying the operating status of a refrigeration device based on deep learning according to an embodiment of the present invention;

[0051] Figure 2 is a flow chart of a training state recognition model according to an embodiment of the present invention;

[0052] Figure 3 are schematic diagrams of various embodiments of the resulting output of the present invention;

[0053] Figure 4 This is a schematic diagram of a module of a device for identifying the operating status of a refrigeration device according to an embodiment of the present invention;

[0054] Figure 5 This is a structural block diagram of a refrigeration device according to an embodiment of the present invention;

[0055] Among them, 100, interactive association module; 10, spatiotemporal feature extraction module; 20, semantic information extraction module; 30, interactive association feature extraction module; 40, difference extraction module; 50, feature fusion module; 60, state recognition module; 70, output module; 200, refrigeration equipment; 201, sensor; 202, log acquisition interface; 203, processing module; 204, storage module; 205, communication bus. DETAILED DESCRIPTION

[0056] The present invention will be described in detail below with reference to the specific embodiments shown in the accompanying drawings. However, these embodiments do not limit the present invention, and any structural, methodological, or functional changes made by those skilled in the art based on these embodiments are all within the scope of protection of the present invention.

[0057] An embodiment of the present invention provides a method and apparatus for improving the efficiency of identifying the operating status of intelligent refrigeration equipment.

[0058] The deep learning-based method for identifying the operating status of refrigeration equipment in this embodiment constructs a state recognition model, which is a deep neural network fusion model composed of CNN with dual-channel ResNet, BiLSTM+dropout, Attention interactive attention and Transformer model.

[0059] The refrigeration device 200 of this embodiment can be a refrigerator, a freezer, an upright refrigerator, a wine cabinet, etc. The following embodiments can be described by taking a refrigerator as an example. The refrigerator includes a cabinet, a storage space with an opening arranged in the cabinet, and a door body covering the opening. Food can be placed in the storage space. A hardware collector, such as sensor 201, can be set in the refrigerator to measure the operating status of the refrigerator in real time. The refrigerator can have an intelligent system on which an APP program runs, or it can be connected to a terminal PAD or mobile phone via a wireless connection. The refrigerator APP or terminal PAD can store data with historical operation records, or historical logs.

[0060] The following combination Figures 1 to 3 , illustrating a method for identifying the operating status of a refrigeration device based on deep learning provided by an embodiment of the present invention. Although the present application provides method operation steps as shown in the following implementation manner or flowchart, based on conventional or no creative labor, the execution order of the steps in the method that do not logically have necessary causal relationships is not limited to the execution order provided in the implementation manner of the present application.

[0061] Specifically, if Figure 2 As shown, the method for identifying the operating status of a refrigeration device based on deep learning in this embodiment includes the following steps:

[0062] Step S10: acquiring real-time status data of a collector of the refrigeration device 200 , and acquiring historical status data of the refrigeration device 200 and / or a smart terminal connected to the refrigeration device 200 .

[0063] Furthermore, real-time status data collected by the sensor 201 of the refrigeration equipment 200 is obtained;

[0064] The historical status data recorded in the log system of the operating system of the refrigeration equipment 200 APP or the terminal PAD is obtained. Here, the historical data of the past year can be selected.

[0065] In addition, the refrigeration equipment 200APP or the terminal PAD can also collect real-time status data, that is, the real-time data is obtained in the form of hardware integration.

[0066] In step S10, for historical status data, the log text data can be cleaned first to filter out abnormal data, and the model training can be performed in combination with the historical data of the log system, which is beneficial to ensure the accuracy of text recognition.

[0067] Specifically, step S10 further includes the following steps:

[0068] The real-time status data and the historical status data are pre-processed.

[0069] The pre-processing step comprises:

[0070] Baseline value calculation and deviation calculation are performed on both the real-time status data and the historical status data to form a deviation matrix.

[0071] Considering that abnormal data will also be stored in the system, abnormal data such as errors, null values, missing values, etc. obtained by hardware such as sensor 201 are filtered out, and data preprocessing is performed. In combination with the refrigerator's operation-dependent hardware, a vectorized matrix composed of text data is used as the model input.

[0072] Step S20: Construct a state recognition model for the refrigeration equipment operating state. Figure 2 As shown, the state recognition model includes an interactive association module 100 , a difference extraction module 40 , a feature fusion module 50 and a state recognition module 60 .

[0073] The interactive association module 100 includes: a spatiotemporal feature extraction module 10, a semantic information extraction module 20, and an interactive association feature extraction module 30 connected in sequence.

[0074] Step S30: using the real-time status data and the historical status data as training sets to train the constructed status recognition model.

[0075] Step S40: The interactive correlation module 100 extracts the interactive correlation features between the real-time status data and the historical status data, the difference extraction module 40 extracts the differential features between the real-time status data and the historical status data, the feature fusion module 50 fuses and splices the interactive correlation features and the differential features to obtain a fused feature vector, and the state recognition module 60 calculates the recognition result based on the fused feature vector to complete the recognition of the operating status of the refrigeration equipment.

[0076] In addition, the spatiotemporal feature extraction module 10 is a spatiotemporal feature extraction neural network, the semantic information extraction module 20 obtains the semantic features of the real-time status data and the historical status data, and the interactive correlation feature extraction module 30 obtains the interactive correlation features of the real-time status data and the historical status data.

[0077] Here, combined with the expansion and Figure 1 , and give a detailed explanation of the specific operation process of the entire state recognition model.

[0078] Spatiotemporal feature extraction module 10

[0079] The spatiotemporal feature extraction neural network is a dual-channel convolutional neural network model CNN network and residual network ResNet, which extracts spatiotemporal features of text data through the CNN network and ResNet. In addition, the spatiotemporal feature extraction module 10 can also include network model adjustment and optimization loss function.

[0080] The following steps are performed in the spatiotemporal feature extraction module 10:

[0081] Input the deviation matrix of the real-time state data into the convolutional neural network model CNN network and the residual network ResNet of one channel, and input the deviation matrix of the historical state data into the convolutional neural network model CNN network and the residual network ResNet of another channel;

[0082] The temporal and spatial features of the real-time status data and the temporal and spatial features of the historical status data are extracted.

[0083] Semantic information extraction module 20

[0084] The semantic information extraction module 20 uses a BiLSTM and Dropout network model to extract long-range features from text data. Dropout can further prevent overfitting, further optimize the obtained text features, select the most effective text feature semantic information, optimize the model, improve the ability to prevent overfitting, and filter redundant information.

[0085] The semantic information extraction module 20 performs the following steps:

[0086] Inputting the spatiotemporal features of the real-time state data and the spatiotemporal features of the historical state data into the corresponding BiLSTM and Dropout network models respectively;

[0087] Context feature vectors of the real-time status data and the historical status data at different times are output respectively, so as to improve the semantic feature representation and fitting capability of the text.

[0088] The BiLSTM model is good at capturing long-distance dependent text sequence information, and the model can obtain the most effective contextual semantic features.

[0089] Interaction correlation feature extraction module 30

[0090] The interactive correlation feature extraction module 30 includes a self-attention mechanism model and an interactive attention mechanism model in bidirectional attention.

[0091] The following steps are executed from the interactive correlation feature extraction module 30:

[0092] Inputting the context feature vectors of the real-time state data and the historical state data into the self-attention mechanism model and the interactive attention mechanism model respectively;

[0093] Obtaining the real-time state data attention feature vector including the real-time state data's own weight information and the weight information between the context feature vector of the real-time state data and the historical state data;

[0094] The historical state data attention feature vector including the weight information of the historical state data itself and the weight information between the context feature vector of the historical state data and the real-time state data is obtained.

[0095] The interactive correlation features include a real-time state data attention feature vector and a historical state data attention feature vector.

[0096] Here, the input context feature vectors of the real-time state data and the context feature vectors of the historical state data are assigned their own weight information through the self-attention mechanism model, thereby obtaining the internal weight information of the textual semantic features of the real-time state data and the historical state data. Furthermore, the input context feature vectors of the real-time state data and the context feature vectors of the historical state data are assigned their mutual association weight information through the mutual attention mechanism model, thereby obtaining the association weight information between the real-time state data and the historical state data. Ultimately, the real-time state data attention feature vector and the historical state data attention feature vector are obtained, thereby enhancing the importance of different parts of the textual semantic feature information and further optimizing the interpretability of the model.

[0097] Because the contextual feature vectors of the real-time status data and historical status data are respectively input into each other's attention mechanism, the interactivity and correlation characteristics within the respective features of the two data are obtained, which strengthens the correlation and interactive contextual characteristics of the two data.

[0098] In other embodiments of the present invention, the text feature enhancement of the real-time status data and the historical status data may be completed based solely on the self-attention mechanism model, or through other algorithm models.

[0099] Difference extraction module 40

[0100] The difference extraction module 40 includes a Bert model and a Transformer model.

[0101] The following steps are performed in the difference extraction module 40:

[0102] The sentence vector encoding of each sentence in the real-time status data and the historical status data is obtained through the pre-trained BERT model; this facilitates computer learning and calculation.

[0103] Based on the Transformer model, the attention score calculation of the Transformer model is changed to difference score calculation; the self-attention mechanism of the Transformer model is optimized.

[0104] The Transformer model is used to calculate the text score of each sentence vector encoding to obtain the differential features. This can obtain the text feature differences between the real-time status data and the historical status data, further enriching the effectiveness and deep-level features of the text semantic information.

[0105] The difference extraction module 40 may first establish a language model of the real-time state data and the historical state data, and then extract differentiated text features of the text data.

[0106] Here is an explanation of the noun.

[0107] BERT (Bidirectional Encoder Representation from Transformers) model: The BERT model further improves the generalization capability of word embedding models, fully describing features at the character level, word level, sentence level, and even inter-sentence relations. It is built on the Transformer model.

[0108] Transformer model: The Transformer model usually consists of multiple encoders stacked together. The encoder not only includes the self-attention layer and the feedforward neural network layer, but also includes the attention layer. The attention layer can help the current node obtain the key content that needs to be paid attention to.

[0109] Feature fusion module 50

[0110] The feature fusion module 50 may adopt a deep neural network fusion model, such as a CNN+BiLSTM+Attention fusion model.

[0111] The following steps are performed in the feature fusion module 50:

[0112] The real-time state data attention feature vector, the historical state data attention feature vector, and the differentiated features are mapped together into a unified multimodal vector space for fusion representation to obtain the fused feature vector.

[0113] The fused feature vector of multimodal fusion incorporates optimal representation capabilities, including contextual information about text semantics, useful feature information, high-level features, the varying importance of useful features, and the differentiated characteristics between real-time and historical status data. This rich semantic feature information enables excellent text representation. This fused feature includes the contribution of historical and real-time status data to text recognition prediction of the smart refrigerator's operating status.

[0114] It should be noted that in current neural network models, there is no clear distinction between multimodal fusion feature representation and multimodal fusion. Therefore, in some embodiments of the present invention, the step within feature fusion module 50 may also include: fusing the real-time state data attention feature vector, the historical state data attention feature vector, and the differential features to obtain a fused feature vector. Both multimodal fusion feature representation and multimodal fusion are intended to combine the real-time state data, the historical state data, and the differential features to better extract and represent their feature information.

[0115] State recognition module 60

[0116] The state recognition module 60 includes a fully connected layer.

[0117] The state recognition module 60 runs the following steps:

[0118] After the fused feature vector is combined through a fully connected layer, it is output to the Softmax function, and the ReLU function is used as the activation function to calculate the text semantic scores of the real-time status data and the historical status data and their normalized score results to obtain the identification information of the operating status of the refrigeration equipment.

[0119] Step S50: Output the identification information of the operating status.

[0120] The refrigeration device 200 further includes an output module 70 , which outputs identification information of the operating status in the form of touch or display.

[0121] like Figure 3 As shown, the output module 70 may include a result access module, text access, voice broadcast, outbound call terminal, SMS delivery, email access, smart speaker, voice wake-up, pop-up UI interface, and push message.

[0122] Compared with the prior art, this embodiment has the following beneficial effects:

[0123] The deep learning-based method and device for identifying the operating status of a refrigeration device combines the real-time status data of a collector and the historical status data of the refrigeration device 200 and / or the intelligent terminal connected to the refrigeration device 200, comprehensively acquires information on the operating status, and integrates the interactive correlation features and differentiation features of the real-time status data and the historical status data, thereby obtaining the best and most effective text features. This not only improves the generalization ability of the network, but also alleviates the gradient vanishing problem caused by increasing the depth of the deep neural network and improves the prediction accuracy. The recognition accuracy is greatly improved, and the operating status of the refrigeration device 200 is accurately identified, thereby improving the operating status monitoring efficiency of the refrigeration device 200, reducing daily operation and maintenance costs, and benefiting the operation and efficiency improvement of the refrigeration device 200 system.

[0124] In one embodiment, a device for identifying the operating status of a refrigeration device based on deep learning is provided. Figure 4 The deep learning-based refrigeration equipment operating status identification device includes modules, and the specific functions of each module are as follows:

[0125] A data acquisition module, configured to acquire real-time status data of a collector of the refrigeration device 200 and historical status data of the refrigeration device 200 and / or a smart terminal connected to the refrigeration device 200;

[0126] A model building module is used to build a state recognition model of the refrigeration equipment operating state, wherein the state recognition model includes an interactive correlation module 100, a difference extraction module 40, a feature fusion module 50 and a state recognition module 60;

[0127] A training module, configured to train a constructed state recognition model using the real-time state data and the historical state data as training sets;

[0128] A model operation module is configured to cause the interactive correlation module 100 to extract interactive correlation features between the real-time status data and the historical status data, the difference extraction module 40 to extract differential features between the real-time status data and the historical status data, the feature fusion module 50 to fuse and combine the correlation features and the differential features to obtain a fused feature vector, and the state recognition module 60 to calculate a recognition result based on the fused feature vector to complete the recognition of the operating state of the refrigeration equipment;

[0129] The output module 70 outputs identification information of the operating status.

[0130] It should be noted that for details not disclosed in the apparatus for identifying the operating status of a refrigeration device according to an embodiment of the present invention, please refer to the details disclosed in the method for identifying the operating status of a refrigeration device according to an embodiment of the present invention.

[0131] Those skilled in the art will understand that the module schematic diagram is merely an example of a device for identifying the operating status of a refrigeration device, and does not constitute a limitation on the terminal device of the device for identifying the operating status of the refrigeration device. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the device for identifying the operating status of the refrigeration device may also include input and output devices, network access devices, buses, etc.

[0132] The device for identifying the operating status of the refrigeration equipment may also include a computing device such as a computer, a notebook, a PDA, and a cloud server, as well as a processing module 203, a storage module 204, and a computer program stored in the storage module 204 and executable on the processing module 203, such as the aforementioned method for identifying the operating status of the refrigeration equipment. When the processing module 203 executes the computer program, the steps of each of the aforementioned methods for identifying the operating status of the refrigeration equipment are implemented, such as Figure 1 and 2 Steps shown.

[0133] The structural block diagram of the refrigeration equipment 200 is as follows: Figure 5 As shown, the refrigeration equipment 200 may include a sensor 201, a log acquisition interface 202, a processing module 203, a storage module 204, and a communication bus 205. The sensor 201 directly acquires real-time status data, the log acquisition interface 202 can acquire data from the refrigerator app, or data from the terminal PAD. The log acquisition interface 202 and the external terminal PAD can transmit data via a wireless connection, such as Bluetooth, Wi-Fi, ZigBee, etc. The communication bus 205 is used to establish a connection between the sensor 201, the log acquisition interface 202, the processing module 203, and the storage module 204. The communication bus 205 may include a path to transmit information between the above-mentioned sensor 201, the log acquisition interface 202, the processing module 203, and the storage module 204.

[0134] In addition, the present invention also proposes an electronic device, which includes a storage module 204 and a processing module 203. When the processing module 203 executes the computer program, it can implement the steps in the above-mentioned method for identifying the operating status of the refrigeration equipment, that is, implement the steps in any one of the technical solutions in the above-mentioned method for identifying the operating status of the refrigeration equipment.

[0135] The electronic device may be a part of a device integrated into the operating status identification device of the refrigeration equipment, or a local terminal device, or a part of a cloud server.

[0136] The processing module 203 can be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processing module 203 is the control center of the refrigeration equipment operating status identification device, and utilizes various interfaces and circuits to connect the various components of the refrigeration equipment operating status identification device.

[0137] The storage module 204 can be used to store the computer programs and / or modules. The processing module 203 implements the various functions of the refrigeration equipment operating status identification device by running or executing the computer programs and / or modules stored in the storage module 204 and accessing the data stored in the storage module 204. The storage module 204 may primarily include a program storage area and a data storage area. The program storage area may store an operating system, at least one application required for a function, and the like. Furthermore, the storage module 204 may include high-speed random access memory (RAM) and non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0138] For example, the computer program may be divided into one or more modules / units, which are stored in the storage module 204 and executed by the processing module 203 to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program in the device for identifying the operating status of a refrigeration appliance.

[0139] Furthermore, an embodiment of the present invention provides a readable storage medium storing a computer program. When the computer program is executed by the processing module 203, it can implement the steps in the above-mentioned method for identifying the operating status of the refrigeration equipment, that is, implement the steps in any one of the technical solutions in the above-mentioned method for identifying the operating status of the refrigeration equipment.

[0140] If the module integrated with the method for identifying the operating status of the refrigeration equipment is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention can also implement all or part of the processes in the above-mentioned method embodiments by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When executed by the processing module 203, the computer program can implement the steps of each of the above-mentioned method embodiments.

[0141] The computer program includes computer program code, which may be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium may include any entity or device capable of carrying the computer program code, a recording medium, a disk, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electric carrier signal, a telecommunication signal, and a software distribution medium. It should be noted that the content of the computer-readable medium may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0142] It should be understood that although this specification is described in terms of implementation methods, not every implementation method contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each implementation method can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

[0143] The series of detailed descriptions listed above are only specific descriptions of feasible implementation methods of the present invention. They are not intended to limit the scope of protection of the present invention. Any equivalent implementation methods or changes that do not deviate from the technical spirit of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for identifying the operating status of refrigeration equipment based on deep learning, characterized in that: The steps include: Acquiring real-time status data of a collector of a refrigeration device, and acquiring historical status data of the refrigeration device and / or a smart terminal connected to the refrigeration device; Constructing a state recognition model for the operating state of the refrigeration equipment, wherein the state recognition model includes an interactive association module, a difference extraction module, a feature fusion module, and a state recognition module, wherein the interactive association module includes: a spatiotemporal feature extraction module, a semantic information extraction module, and an interactive association feature extraction module connected in sequence; the spatiotemporal feature extraction module is a spatiotemporal feature extraction neural network, the semantic information extraction module obtains semantic features of the real-time state data and the historical state data, and the interactive association feature extraction module obtains interactive association features of the real-time state data and the historical state data; Using the real-time status data and the historical status data as training sets to train the constructed status recognition model; The interactive correlation module extracts interactive correlation features between the real-time status data and the historical status data, the difference extraction module extracts differential features between the real-time status data and the historical status data, the feature fusion module fuses and splices the interactive correlation features and the differential features to obtain a fused feature vector, and the state recognition module calculates a recognition result based on the fused feature vector to complete the recognition of the operating state of the refrigeration equipment; Output identification information of the operating status; The identification method further comprises the steps of: Obtaining sentence vector encoding for each sentence in the real-time state data and the historical state data through a pre-trained BERT model; Based on the Transformer model, the attention score calculation of the Transformer model is changed to the difference score calculation; The text score is calculated for each of the sentence vector codes through the Transformer model to obtain the differential features.

2. The method for identifying the operating status of a refrigeration device according to claim 1, characterized in that: Also includes the steps: Preprocessing the real-time status data and the historical status data, the preprocessing step comprising: Baseline value calculation and deviation calculation are performed on both the real-time status data and the historical status data to form a deviation matrix.

3. The method for identifying the operating status of a refrigeration device according to claim 1, characterized in that: The spatiotemporal feature extraction neural network is a dual-channel convolutional neural network model CNN network and a residual network ResNet; Also includes the steps: Input the deviation matrix of the real-time state data into the convolutional neural network model CNN network and the residual network ResNet of one channel, and input the deviation matrix of the historical state data into the convolutional neural network model CNN network and the residual network ResNet of another channel; The temporal and spatial features of the real-time status data and the temporal and spatial features of the historical status data are extracted.

4. The method for identifying the operating status of a refrigeration device according to claim 3, characterized in that: The semantic information extraction module adopts BiLSTM and Dropout network models; Also includes the steps: Inputting the spatiotemporal features of the real-time state data and the spatiotemporal features of the historical state data into the corresponding BiLSTM and Dropout network models respectively; Context feature vectors of the real-time status data and the historical status data at different times are output respectively, so as to improve the semantic feature representation and fitting capability of the text.

5. The method for identifying the operating status of a refrigeration device according to claim 4, characterized in that: The interactive correlation feature extraction module includes a self-attention mechanism model and an interactive attention mechanism model in bidirectional attention, and the interactive correlation feature includes a real-time state data attention feature vector and a historical state data attention feature vector; Also includes the steps: Inputting the context feature vectors of the real-time state data and the historical state data into the self-attention mechanism model and the interactive attention mechanism model respectively; Obtaining the real-time state data attention feature vector including the real-time state data's own weight information and the weight information between the context feature vector of the real-time state data and the historical state data; The historical state data attention feature vector including the weight information of the historical state data itself and the weight information between the context feature vector of the historical state data and the real-time state data is obtained.

6. The method for identifying the operating status of a refrigeration device according to claim 1, characterized in that: Also includes the steps: The real-time state data attention feature vector, the historical state data attention feature vector, and the differentiated features are mapped together into a unified multimodal vector space for fusion representation to obtain the fused feature vector.

7. The method for identifying the operating status of a refrigeration device according to claim 1, characterized in that: Also includes the steps: After the fused feature vectors are combined through a fully connected layer, they are output to a Softmax function to calculate the text semantic scores of the real-time status data and the historical status data and their normalized score results to obtain identification information of the operating status of the refrigeration equipment.

8. The method for identifying the operating status of a refrigeration device according to claim 1, characterized in that: Also includes the steps: Obtain real-time status data collected by sensors of refrigeration equipment; Obtain historical status data recorded in the log system of the operating system of the refrigeration equipment APP or terminal PAD.

9. A device for identifying the operating status of refrigeration equipment based on deep learning, characterized in that: include: A data acquisition module, configured to acquire real-time status data of a collector of a refrigeration device, and historical status data of the refrigeration device and / or a smart terminal connected to the refrigeration device; A model construction module is used to construct a state recognition model for the operating state of the refrigeration equipment, wherein the state recognition model includes an interactive association module, a difference extraction module, a feature fusion module, and a state recognition module. The interactive association module includes: a spatiotemporal feature extraction module, a semantic information extraction module, and an interactive association feature extraction module connected in sequence; the spatiotemporal feature extraction module is a spatiotemporal feature extraction neural network, the semantic information extraction module obtains semantic features of the real-time state data and the historical state data, and the interactive association feature extraction module obtains interactive association features of the real-time state data and the historical state data; A training module, configured to train a constructed state recognition model using the real-time state data and the historical state data as training sets; A model operation module is used for the interactive association module to extract the interactive association features between the real-time status data and the historical status data, the difference extraction module to extract the differential features between the real-time status data and the historical status data, the feature fusion module to fuse and splice the association features and the differential features to obtain a fused feature vector, and the state recognition module to calculate the recognition result based on the fused feature vector to complete the recognition of the operating status of the refrigeration equipment. The difference extraction module includes: obtaining the sentence vector encoding of each sentence in the real-time status data and the historical status data through a pre-trained BERT model; based on the Transformer model, changing the attention score calculation of the Transformer model to a difference score calculation; performing text score calculation on each of the sentence vector encodings through the Transformer model to obtain the differential features; Output module, outputs identification information of operating status.

10. An electronic device, characterized in that: include: a storage module storing a computer program; The processing module can implement the steps of the method for identifying the operating status of a refrigeration device based on deep learning as described in any one of claims 1 to 8 when executing the computer program.

11. A readable storage medium storing a computer program, characterized in that: When the computer program is executed by the processing module, it can implement the steps in the method for identifying the operating status of a refrigeration device based on deep learning as described in any one of claims 1 to 8.

Images