Air conditioner frost detection method and device based on neural network model, air conditioner

Abstract

The application relates to the technical field of intelligent household appliances, and discloses an air conditioner frosting detection method and device based on a neural network model and an air conditioner. The method comprises the following steps: acquiring real-time operation data and real-time weather data of the air conditioner under the condition that the air conditioner operates in a heating mode; inputting the acquired real-time operation data into a frosting degree prediction model to obtain a frosting degree prediction result; and performing cumulative analysis on the frosting degree prediction result according to the real-time weather data and a sliding window to determine a defrosting opportunity. The frosting degree prediction model is a neural network model obtained by training based on a data set and an outdoor unit frosting image label, and the data set comprises operation data of the air conditioner under various frosting working conditions. The method performs cumulative judgment from two dimensions of time and environment, avoids interference caused by extreme weather and accidental misjudgment caused by the model. Furthermore, the stability and reliability of the air conditioner system operation are significantly improved.

Description

Technical Field

[0001] This application relates to the field of smart home appliance technology, such as a method and device for detecting air conditioner frost based on a neural network model, and an air conditioner. Background Technology

[0002] When an air conditioner operates in a low-temperature and humid environment for heating, frost easily forms on the surface of the outdoor unit's heat dissipation components. The formation of frost significantly increases the heat exchange resistance of the outdoor unit, leading to a decrease in heating efficiency and even system protective shutdown. Because the frosting process is gradual, the characteristic parameters of thin frost and the critical frost state differ only slightly. Traditional algorithms are prone to misjudging the critical point, causing "oscillations" in defrost control. Furthermore, extreme weather conditions severely affect the frosting process, causing abnormal data changes and increasing the likelihood of misjudgments. This, in turn, leads to unnecessary defrosting operations, making it difficult for traditional algorithms to maintain robustness under extreme conditions.

[0003] The related technology discloses a method for detecting frost on an air conditioner, including: using the self-learning ability of an artificial neural network to establish a correspondence between the operating parameters of the air conditioner in heating mode and the frost state of the condenser; obtaining the current operating parameters of the air conditioner; determining the current frost state corresponding to the current operating parameters through the correspondence; determining the current frost state corresponding to the current operating parameters includes: determining the frost state corresponding to the operating parameters that are the same as the current operating parameters in the correspondence as the current frost state.

[0004] In the process of implementing the embodiments of this disclosure, at least the following problems were found in the related art:

[0005] Although the algorithms of related technologies take into account environmental parameters, they do not take into account weather factors; this leads to a certain degree of misjudgment in the detection of frost under extreme weather conditions.

[0006] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this application, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0007] To provide a basic understanding of some aspects of the disclosed embodiments, a brief summary is given below. This summary is not intended as a general commentary, nor is it intended to identify key / important components or describe the scope of protection of these embodiments, but rather as a prelude to the detailed description that follows.

[0008] This disclosure provides an air conditioner frost detection method and device based on a neural network model, and an air conditioner, to improve the reliability of frost detection.

[0009] In some embodiments, the method includes: acquiring real-time operating data and real-time weather data of the air conditioner when it is operating in heating mode; inputting the acquired real-time operating data into a frost degree prediction model to obtain a frost degree prediction result; and performing cumulative analysis on the frost degree prediction result based on the real-time weather data and a sliding window to determine the defrosting timing; wherein the frost degree prediction model is a neural network model trained based on a dataset and outdoor unit frost image labels, and the dataset includes operating data of the air conditioner under various frost conditions.

[0010] In some embodiments, the apparatus includes a processor and a memory storing program instructions, the processor being configured to execute, when running the program instructions, the air conditioner frost detection method based on a neural network model as described above.

[0011] In some embodiments, the air conditioner includes: an outdoor unit body; and an air conditioner frost detection device based on a neural network model, as described above, installed on the outdoor unit body.

[0012] The air conditioner frost detection method and device based on a neural network model, and the air conditioner provided in this disclosure can achieve the following technical effects:

[0013] Based on real-time operating data of the air conditioner and a frost level prediction model, frost level prediction results are obtained. Then, using sliding window data and real-time weather data, the frost level prediction results are cumulatively analyzed to determine the defrosting timing. This cumulative judgment from both time and environmental dimensions avoids interference from extreme weather and accidental misjudgments from the model. Consequently, the stability and reliability of the air conditioning system are significantly improved.

[0014] The above general description and the description below are exemplary and illustrative only and are not intended to limit this application. Attached Figure Description

[0015] One or more embodiments are illustrated by way of example with reference to the accompanying drawings. These illustrations and drawings do not constitute a limitation on the embodiments. Elements having the same reference numerals in the drawings are shown as similar elements. The drawings are not to be scaled. And wherein:

[0016] Figure 1 This is a schematic diagram of the first air conditioner frost detection method based on a neural network model provided in this disclosure embodiment;

[0017] Figure 2 This is a schematic diagram of the frost degree prediction model training method provided in the embodiments of this disclosure;

[0018] Figure 3This is a schematic diagram of a convolution process provided in an embodiment of this disclosure;

[0019] Figure 4 This is a schematic diagram of the second air conditioner frost detection method based on a neural network model provided in this embodiment of the present disclosure;

[0020] Figure 5 This is a schematic diagram of the third air conditioner frost detection method based on a neural network model provided in this disclosure embodiment;

[0021] Figure 6 This is a schematic diagram of an air conditioner frost detection device based on a neural network model provided in an embodiment of this disclosure;

[0022] Figure 7 This is a schematic diagram of an outdoor unit of an air conditioner provided in an embodiment of this disclosure.

[0023] Figure label:

[0024] 100: Air conditioner frost detection device based on neural network model; 101: Processor; 102: Memory; 103: Communication interface; 104: Bus; 200: Air conditioner outdoor unit body. Detailed Implementation

[0025] To provide a more detailed understanding of the features and technical content of the embodiments of this disclosure, the implementation of the embodiments of this disclosure will be described in detail below with reference to the accompanying drawings. The accompanying drawings are for illustrative purposes only and are not intended to limit the embodiments of this disclosure. In the following technical description, for ease of explanation, several details are used to provide a full understanding of the disclosed embodiments. However, one or more embodiments may still be implemented without these details. In other cases, well-known structures and devices may be simplified in their depiction to simplify the drawings.

[0026] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this disclosure described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion.

[0027] Unless otherwise stated, the term "multiple" means two or more.

[0028] In this embodiment of the disclosure, the character " / " indicates that the objects before and after it are in an "or" relationship. For example, A / B means: A or B.

[0029] The term "and / or" describes an association between objects, indicating that three relationships can exist. For example, A and / or B means: A or B, or A and B.

[0030] The term "correspondence" can refer to an association or binding relationship. The correspondence between A and B means that there is an association or binding relationship between A and B.

[0031] In this embodiment of the disclosure, smart home appliances refer to home appliances formed by introducing microprocessors, sensor technology and network communication technology into home appliances. They have the characteristics of intelligent control, intelligent sensing and intelligent application. The operation of smart home appliances often relies on the application and processing of modern technologies such as the Internet of Things, the Internet and electronic chips. For example, smart home appliances can be connected to electronic devices to enable users to remotely control and manage smart home appliances.

[0032] Combination Figure 1 As shown, this disclosure provides a first air conditioner frost detection method based on a neural network model, including:

[0033] S101, when the air conditioner is in heating mode, the processor obtains real-time operating data of the air conditioner and real-time weather data.

[0034] S102, the processor inputs the acquired real-time running data into the frost degree prediction model to obtain the frost degree prediction result.

[0035] S103, the processor performs cumulative analysis on the frost level prediction results based on real-time weather data and sliding windows to determine the timing of defrosting.

[0036] The frost level prediction model is a neural network model trained on a dataset and outdoor unit frost image labels. The dataset includes air conditioner operating data under various frost conditions.

[0037] Here, the outdoor heat exchanger surface is prone to frost formation when the air conditioner is in heating mode. Therefore, various sensors are used to collect real-time operating data of the air conditioner when it is in heating mode. Real-time operating data includes various parameters that characterize or affect frost formation, such as indoor ambient temperature, and / or indoor coil temperature, and / or outdoor ambient temperature, and / or outdoor coil temperature, and / or compressor operating frequency, and / or electronic expansion valve opening, and / or compressor discharge / suction temperature, and / or compressor discharge / suction pressure. Real-time weather data is also acquired, including weather condition data and outdoor environmental data (such as humidity). Real-time weather data can be obtained from local weather data via a server from the cloud, or it can be obtained from other smart home appliances (such as smart speakers) through linkage between the air conditioner and other smart home devices.

[0038] The collected real-time operating data is input into the frost level prediction model to obtain the frost level prediction result. The frost level prediction model is a trained neural network model, trained on a dataset, and calibrated using outdoor unit frost image labels corresponding to the data in the dataset. The dataset includes operating data of the air conditioner under different frost conditions, including parameters as described above for the real-time operating data. The outdoor unit frost image labels are frost images corresponding to the operating data, with the frost level indicated. For example, the frost level includes two categories: thin frost and severe frost; the frost level in the frost image is then labeled as either thin or severe frost. The frost level prediction result output by the trained neural network model is compared with the actual frost level result corresponding to the image labels, serving as the basis for adjusting the neural network model parameters. This process is repeated to obtain a frost level prediction model that meets the specified criteria.

[0039] After obtaining the frost level prediction results, to improve the stability of frost detection in the air conditioning system under extreme weather conditions, a cumulative analysis of the frost level prediction results is performed, combining real-time weather data and a sliding window, to determine the defrosting timing. On one hand, a sliding window is used to make decisions from a time perspective. On the other hand, real-time weather data is used to make decisions from an environmental perspective. This ensures the sensitivity of frost detection while reducing the interference of extreme weather on the detection results, thereby reducing the possibility of misjudgment. The sliding window parameters (including the sliding window size and threshold) are preferably dynamic values, such as those that can be adjusted based on real-time weather data. This allows for adaptive adjustments based on the environment to better adapt to different conditions.

[0040] The air conditioning frost detection method based on a neural network model provided in this disclosure obtains frost level prediction results based on real-time air conditioning operating data and a frost level prediction model. Then, by using a sliding window and real-time weather data, the frost level prediction results are cumulatively analyzed to determine the defrosting timing. This cumulative judgment from both time and environmental dimensions avoids interference from extreme weather and accidental misjudgments caused by the model. Consequently, the stability and reliability of the air conditioning system are significantly improved.

[0041] Optionally, the frost severity prediction model includes a one-dimensional convolutional layer, a fully connected layer, an activation function, and a classifier;

[0042] The fully connected layer includes a first fully connected layer, a second fully connected layer, and a third fully connected layer. The outputs of the one-dimensional convolutional layer, the first fully connected layer, and the second fully connected layer introduce activation functions, and the output of the third fully connected layer connects to the classifier.

[0043] Here, a lightweight neural network model is used as the frost severity prediction model to ensure the algorithm's real-time inference speed. Specifically, the frost severity prediction model includes a one-dimensional convolutional layer, a fully connected layer, an activation function, and a classifier. The outputs of the convolutional layer and the first two fully connected layers (i.e., the first and second fully connected layers) are all incorporated into an activation function, using ReLU. The last fully connected layer (i.e., the third fully connected layer) is connected to a classifier (which can be a Sigmoid classifier). In this embodiment, the classifier outputs two frost severity results, corresponding to 0 (thin frost) and 1 (severe frost). That is, when the classifier outputs a frost severity prediction result of 0, it represents thin frost. When the frost severity prediction result is 1, it represents severe frost.

[0044] Alternatively, the one-dimensional convolutional layer can be replaced with a Long Short-Term Memory (LSTM) network or a Gated Recurrent Unit (GRU). All three methods can process multi-dimensional data, extract frost features from multi-dimensional data, and uncover hidden correlations between different real-time running data.

[0045] Optionally, combined Figure 2 As shown, in step S102, the frost severity prediction model is trained in the following way:

[0046] S121, the processor normalizes the running data in the dataset.

[0047] S122, the processor inputs the normalized running data into the neural network model to train the model.

[0048] S123, the processor uses the outdoor unit frost image labels to calibrate the output of the neural network model, and uses the calibrated and trained neural network model as the frost degree prediction model.

[0049] Here, because the air conditioner's operational data includes multi-dimensional data with differences in dimensions and orders of magnitude, this embodiment of the disclosure normalizes the operational data in the dataset to ensure the model can fairly learn and capture the relationships between the various data points. This scales all operational data to a fixed range. The normalized operational data is then input into the model for training.

[0050] In detail, the model's convolutional layers process one-dimensional sequence data by sliding a learnable convolutional kernel (also called a filter) across the input one-dimensional sequence. At each position, element-wise multiplication and summation are performed to generate the output sequence (the convolution process is as follows). Figure 3 (As shown). The input variables of the convolutional layer are... The convolution kernel is a one-dimensional vector of length k. During convolution, the convolution kernel slides from top to bottom across the input sequence, moving one stride (typically 1) at a time. At each position, the kernel is multiplied element-wise with the corresponding input subsequence, and the results are summed to obtain an element of the output sequence. More specifically, for the i-th element of the output sequence... y i Its calculation formula is Here, 'i' must meet certain conditions to ensure that the convolution kernel has effective coverage over the input sequence. The stride controls the distance the convolution kernel slides each time, usually denoted by 's'. If zero padding is used, extra zero elements are added to both ends of the input sequence to change the output length of the convolution operation.

[0051] In a fully connected layer of the model, each neuron is connected to all neurons in the previous layer. This means that the output of each neuron in the previous layer is transmitted to every neuron in the current layer. Therefore, the main function of the fully connected layer is to integrate and transform the features extracted from the previous layers, thereby completing complex nonlinear mappings. The input vector of the fully connected layer is... ,in, x i This is the output of the i-th neuron in the previous layer. The weight matrix is ​​W, an m×n matrix. W ij This represents the weights from the j-th neuron in the previous layer to the i-th neuron in the current layer. The bias vector is... , where b i This is the bias of the i-th neuron in the current layer. The linear output of the i-th neuron in the current layer. It can be represented as The matrix form is z = Wx + b. The activation function is calculated using the formula ReLU( x )= max (0, x ).

[0052] After the fully connected layers and activation functions are calculated, the results are classified using a Sigmoid classifier. The Sigmoid calculation formula is as follows: The output determines the probability of belonging to one of the two label classes, with the prediction result ranging from [0,1]. The prediction results are divided into two categories, 0 (thin frost) and 1 (severe frost), with 0.5 as the boundary. The prediction results are compared using outdoor unit frost image labels to ensure the accuracy of the model output.

[0053] Optionally, in step S121, the processor normalizes the running data in the dataset, including:

[0054] The processor removes outliers and missing values ​​from the running data.

[0055] The processor uses max-min normalization to remove processed running data in order to linearly scale the running data to a fixed range.

[0056] Here, the air conditioner's operating data may be partially missing or corrupted during collection or transmission. Therefore, before normalizing the operating data, preprocessing is required to remove outliers and missing values. After preprocessing, the input data is normalized using a max-min normalization method. This maps the original data range to [0,1]. The formula for maximum value normalization is... Here, x' represents the normalized running data, and x represents the preprocessed running data. This ensures that all running data are on the same scale, thereby improving model performance.

[0057] Combination Figure 4 As shown, this disclosure provides a second air conditioner frost detection method based on a neural network model, including:

[0058] S101, when the air conditioner is in heating mode, the processor obtains real-time operating data of the air conditioner and real-time weather data.

[0059] S102, the processor inputs the acquired real-time running data into the frost degree prediction model to obtain the frost degree prediction result.

[0060] S131, the processor determines the initial size and initial threshold of the sliding window based on the frosting conditions and defrosting time error.

[0061] S132, the processor dynamically adjusts the initial size and initial threshold of the sliding window based on real-time weather data.

[0062] S133, the processor stores the frost degree prediction results in a dynamically adjusted sliding window, and when the stored data reaches its maximum value, it performs cumulative analysis on the frost degree prediction results in the dynamically adjusted sliding window to determine the defrosting time.

[0063] The frost level prediction model is a neural network model trained on a dataset and outdoor unit frost image labels. The dataset includes air conditioner operating data under various frost conditions.

[0064] Here, a sliding window is used to accumulate the predicted frost severity from the model output to determine the defrosting timing. The sliding window threshold is a critical value used to define the magnitude of a certain type of frost severity prediction within the sliding window. For example, the sliding window threshold is used to define the magnitude of severe frost among the predicted frost severity results within the sliding window. The sliding window threshold is smaller than the sliding window size. The initial size and initial sliding window threshold depend on the frost condition and defrosting time error. Specifically, when the frost condition indicates severe frost, the initial size and initial sliding window threshold are smaller. When the frost condition indicates relatively light frost, the initial size and initial sliding window threshold are larger. The defrosting time error refers to the range of difference between the expected defrosting time and the actual defrosting time. The stricter the defrosting time error requirement, the smaller the initial size and initial sliding window threshold. Conversely, the broader the defrosting time error requirement, the larger the initial size and initial sliding window threshold.

[0065] Considering the significant impact of weather data on frost formation, this embodiment utilizes real-time weather data to dynamically adjust the initial size and threshold of the sliding window. Specifically, when real-time weather data indicates favorable conditions for frost formation, the initial size and threshold of the sliding window are decreased. Conversely, when real-time weather data indicates unfavorable conditions for frost formation, the initial size and threshold of the sliding window are increased. This avoids the system making incorrect judgments unnecessarily.

[0066] Furthermore, in this embodiment, an adjusted sliding window is used to process the frost severity prediction results. When the data stored in the sliding window accumulates to its maximum value, the number of frost severity prediction results within the window is determined to decide whether to initiate defrosting. Specifically, defrosting is initiated only if the number of severe frost prediction results is greater than or equal to the adjusted initial sliding window threshold. Otherwise, defrosting is not initiated, and the sliding window is updated in real time. This cumulative judgment method based on the sliding window over a time dimension is unaffected by instantaneous anomalies and can filter out interference from extreme weather and accidental misjudgments caused by complex factors. This ensures that the air conditioning system can operate stably and accurately in various complex environments.

[0067] Furthermore, in some embodiments, the number of thin frost particles in the predicted frost level within the sliding window can be determined. If the number of thin frost particles is less than or equal to the adjusted initial sliding window threshold, then defrosting is initiated. Here, the rules for determining the sliding window threshold need to be adaptively adjusted based on the principles described above.

[0068] Optionally, in step S131, the processor determines the initial size and initial sliding window threshold of the sliding window based on the frosting conditions and defrosting time error, including:

[0069] The processor determines the baseline size and baseline threshold of the sliding window based on the frosting conditions.

[0070] The processor predicts the detection cycle of the model based on the defrosting time error and the degree of frost formation, and calculates the maximum sliding window threshold.

[0071] The processor adjusts the baseline size and baseline sliding window threshold of the sliding window based on the maximum sliding window threshold, and uses the adjusted values ​​as the initial size and initial sliding window threshold of the sliding window.

[0072] Here, the baseline size and threshold of the sliding window are first determined based on the frosting condition. The frosting condition can be categorized into low-temperature high-humidity, low-temperature low-humidity, and variable-temperature / variable-humidity conditions, with corresponding sliding window parameters set for each condition. This allows for the determination of the baseline parameters for the sliding window based on the frosting condition. Then, the detection cycle is predicted based on the defrost time error and the degree of frosting, and the maximum sliding window threshold is calculated. The defrost time error is an interval value, with a maximum defrost time error. The quotient of the predicted detection cycle based on the maximum defrost time error and the degree of frosting can be used as the maximum sliding window threshold. Further, the maximum sliding window threshold is compared with the baseline sliding window threshold. If the baseline sliding window threshold is less than or equal to the maximum sliding window threshold, the baseline size and threshold of the sliding window are maintained. If they are greater, the baseline sliding window threshold and the baseline size of the sliding window are adjusted. The ratio of the sliding window threshold to the sliding window size is greater than or equal to a first value. The first value ranges from greater than or equal to 70%.

[0073] For example, the defrost time error is [-3 min, 3 min], meaning that defrost commands can be triggered within three minutes before or after the expected defrost time. Simultaneously, the model detection cycle is 5 seconds, meaning a prediction result is output every 5 seconds. Therefore, the defrost time error model can detect 72 times. The baseline sliding window threshold cannot exceed 72 times. If the baseline sliding window threshold can be 60 times, and the first value is 75%, the sliding window size would be 80 times. In this way, a relatively reasonable sliding window parameter can be determined to meet the operating conditions.

[0074] Optionally, in step S132, the processor dynamically adjusts the initial size and initial threshold of the sliding window based on real-time weather data, including:

[0075] When real-time weather data indicates stable weather conditions, the processor increases the initial size and initial threshold of the sliding window.

[0076] When real-time weather data indicates a sudden change in weather, the processor will reduce the initial size and initial threshold of the sliding window.

[0077] Here, sudden weather changes refer to extreme weather events such as strong winds, torrential rain, or sudden temperature fluctuations. Stable weather is the opposite of sudden weather changes. When the weather is stable, increasing the initial size and threshold of the sliding window allows for a more comprehensive consideration of predictions over a longer period, thus improving the overall completeness and accuracy of the judgment. During sudden weather changes, decreasing the initial size and threshold of the sliding window allows for more timely and sensitive detection of current anomalies, enabling a rapid response. This dynamic adjustment of the sliding window size based on real-time weather data allows the model to better adapt to different weather conditions, effectively improving the reliability of frost detection and the accuracy of defrost control.

[0078] Optionally, the initial sliding window threshold is adjusted 5 times. Then, based on the first value, the adjustment range of the initial size is determined. Here, the initial sliding window threshold is adjusted 5 times whether it is increased or decreased. For example, if the initial sliding window size is 80 and the initial sliding window threshold is 60, then the first value is 75%. If the adjusted initial sliding window threshold is 65, then the adjusted initial size is 86, and the adjustment range of the initial sliding window size is 6.

[0079] Combination Figure 5 As shown, this disclosure provides a third air conditioner frost detection method based on a neural network model, including:

[0080] S101, when the air conditioner is in heating mode, the processor obtains real-time operating data of the air conditioner and real-time weather data.

[0081] S102, the processor inputs the acquired real-time running data into the frost degree prediction model to obtain the frost degree prediction result.

[0082] S103, the processor performs cumulative analysis on the frost level prediction results based on real-time weather data and sliding windows.

[0083] The frost level prediction model is a neural network model trained on a dataset and outdoor unit frost image labels. The dataset includes air conditioner operating data under various frost conditions.

[0084] S204, if the analysis results show that the number of severe frosts in the data within the dynamically adjusted sliding window is greater than the detection threshold, the processor determines to trigger a defrost command.

[0085] The predicted degree of frost includes severe frost and thin frost.

[0086] Here, the detection threshold is the adjusted initial sliding window threshold. After the amount of data within the sliding window reaches its maximum value, the number of severe frost points in the predicted frost level within the sliding window is analyzed. If the number of severe frost points exceeds the detection threshold, a defrosting command is triggered. Otherwise, the sliding window continues to update, and the accumulated results are analyzed in real time to see if they exceed the detection threshold. This significantly reduces the interference of extreme weather on the detection results and decreases the possibility of misjudgments.

[0087] Combination Figure 6 As shown, this disclosure provides an air conditioner frost detection device 100 based on a neural network model, including a processor 101 and a memory 102. Optionally, the device may further include a communication interface 103 and a bus 104. The processor 101, communication interface 103, and memory 102 can communicate with each other via the bus 104. The communication interface 103 can be used for information transmission. The processor 101 can call logical instructions in the memory 102 to execute the air conditioner frost detection method based on the neural network model described in the above embodiment.

[0088] Furthermore, the logical instructions in the aforementioned memory 102 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium.

[0089] The memory 102, as a computer-readable storage medium, can be used to store software programs and computer-executable programs, such as program instructions / modules corresponding to the methods in the embodiments of this disclosure. The processor 101 executes functional applications and data processing by running the program instructions / modules stored in the memory 102, that is, it implements the air conditioner frost detection method based on the neural network model in the above embodiments.

[0090] The memory 102 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the terminal device. Furthermore, the memory 102 may include high-speed random access memory and may also include non-volatile memory.

[0091] Combination Figure 7As shown, this disclosure provides an air conditioner, including: an outdoor unit body 200 and the aforementioned air conditioner frost detection device 100 based on a neural network model. The air conditioner frost detection device 100 based on the neural network model is installed on the outdoor unit body. The installation relationship described herein is not limited to placement inside the outdoor unit body, but also includes installation connections with other components of the outdoor unit, including but not limited to physical connections, electrical connections, or signal transmission connections. Those skilled in the art will understand that the air conditioner frost detection device 100 based on the neural network model can be adapted to feasible outdoor unit bodies to achieve other feasible embodiments.

[0092] This disclosure provides a computer-readable storage medium storing computer-executable instructions configured to execute the above-described air conditioner frost detection method based on a neural network model.

[0093] The technical solutions of this disclosure can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes one or more instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in this disclosure. The aforementioned storage medium can be a non-transitory storage medium, such as a USB flash drive, external hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, etc., and other media capable of storing program code.

[0094] The foregoing description and accompanying drawings fully illustrate embodiments of this disclosure to enable those skilled in the art to practice them. Other embodiments may include structural, logical, electrical, procedural, and other changes. The embodiments represent only possible variations. Individual components and functions are optional unless explicitly required, and the order of operation may vary. Parts and features of some embodiments may be included in or replace parts and features of other embodiments. Moreover, the terminology used in this application is for describing embodiments only and is not intended to limit the claims. As used in the description of embodiments and claims, the singular forms “a,” “an,” and “the” are intended to equally include the plural forms unless the context clearly indicates otherwise. Similarly, the term “and / or” as used in this application means including one or more of the associated listed items and all possible combinations thereof. Additionally, when used in this application, the term "comprise" and its variations "comprises" and / or "comprising" refer to the presence of stated features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof. Without further limitations, an element defined by the phrase "comprises a..." does not exclude the presence of other identical elements in the process, method, or apparatus that includes said element. In this document, each embodiment may focus on the differences from other embodiments, and similar or identical parts between embodiments can be referred to mutually. For methods, products, etc., disclosed in the embodiments, if they correspond to the method section disclosed in the embodiments, the relevant parts can be referred to the description of the method section.

[0095] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the embodiments of this disclosure. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0096] The methods and products disclosed in the embodiments herein (including but not limited to devices and equipment) can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units may be merely a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to implement this embodiment according to actual needs. In addition, the functional units in the embodiments of this disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0097] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than that shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different blocks may also occur in a different order than disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. Each block in a block diagram and / or flowchart, as well as combinations of blocks in a block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

Claims

1. A method for detecting air conditioner frost based on a neural network model, characterized in that, include: When the air conditioner is in heating mode, acquire the real-time operating data of the air conditioner and the real-time weather data; The acquired real-time operating data is input into the frost level prediction model to obtain the frost level prediction result; wherein, the frost level prediction model is a neural network model trained based on a dataset and outdoor unit frost image labels, and the dataset includes the air conditioner's operating data under various frost conditions. Based on the real-time weather data and the sliding window, a cumulative analysis is performed on the frost severity prediction results to determine the timing of defrosting; this includes, The initial size and initial threshold of the sliding window are determined based on the frosting conditions and defrosting time error. Specifically, the smaller the initial size and initial threshold of the sliding window are when the frosting conditions are more severe or the defrosting time error is more stringent. Conversely, the larger the initial size and initial threshold of the sliding window are when the frosting conditions are less severe or the defrosting time error is more widespread. Based on the real-time weather data, the initial size and initial threshold of the sliding window are dynamically adjusted; wherein, when the real-time weather data indicates stable weather, the initial size and initial threshold of the sliding window are increased; when the real-time weather data indicates a sudden change in weather, the initial size and initial threshold of the sliding window are decreased. The predicted frost level is stored in a dynamically adjusted sliding window, and when the stored data reaches its maximum value, the predicted frost level in the dynamically adjusted sliding window is cumulatively analyzed.

2. The method according to claim 1, characterized in that, The frost severity prediction model includes a one-dimensional convolutional layer, a fully connected layer, an activation function, and a classifier; The fully connected layer includes a first fully connected layer, a second fully connected layer, and a third fully connected layer. The outputs of the one-dimensional convolutional layer, the first fully connected layer, and the second fully connected layer are introduced into the activation function, and the output of the third fully connected layer is connected to the classifier.

3. The method according to claim 1, characterized in that, The frost intensity prediction model was trained in the following manner: The runtime data in the dataset is normalized. The normalized running data is input into the neural network model to train the neural network model; The output of the neural network model is calibrated using the outdoor unit frost image labels, and the calibrated and trained neural network model is used as the frost degree prediction model.

4. The method according to claim 3, characterized in that, The normalization process for the runtime data in the dataset includes: Remove outliers and missing values ​​from the operational data; The processed running data is removed by using max-min normalization to linearly scale the running data to a fixed range.

5. The method according to claim 1, characterized in that, The step of determining the initial size and initial threshold of the sliding window based on the frosting conditions and defrosting time error includes: Based on the frosting conditions, determine the reference size and reference sliding window threshold of the sliding window; Calculate the maximum sliding window threshold based on the detection cycle of the defrosting time error and frost degree prediction model; Based on the maximum sliding window threshold, the baseline size and baseline sliding window threshold of the sliding window are adjusted, and the adjusted values ​​are used as the initial size and initial sliding window threshold of the sliding window.

6. The method according to any one of claims 1 to 5, characterized in that, The cumulative analysis of the predicted frost level results to determine the timing of defrosting includes: If the analysis results indicate that the number of severe frost formations in the dynamically adjusted sliding window data exceeds the detection threshold, a defrosting command is triggered. The predicted frosting level includes severe frosting and thin frosting.

7. An air conditioner frost detection device based on a neural network model, comprising a processor and a memory storing program instructions, characterized in that, The processor is configured to execute, when running the program instructions, the air conditioner frost detection method based on a neural network model as described in any one of claims 1 to 6.

8. An air conditioner, characterized in that, include: The outdoor unit of the air conditioner; The air conditioner frost detection device based on a neural network model as described in claim 7 is installed on the outdoor unit of the air conditioner.

Images