A method and device for predicting equipment status
By using error-weighted model training data from similar devices, the method improves the accuracy of equipment status prediction models, addressing the issue of varying data distributions and scarce labeled data.
Patent Information
- Application Number
- CN202011010325.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-09-23
- Publication Date
- 2024-12-20
- Estimated Expiration
- 2040-09-23
AI Technical Summary
Existing equipment status prediction models have low accuracy due to differing data distributions among devices, especially when labeled data is scarce for the device to be predicted.
The method involves acquiring feature data and model training data from reference devices of the same type, determining error weights based on distribution probabilities, and using these to train a device state prediction model for the target device, ensuring accurate prediction by considering the error weights of the model training data.
This approach enhances the accuracy of device state prediction by leveraging data from similar devices, allowing for precise estimation of equipment status.
Smart Images
Figure CN114254790B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of energy technology, and in particular to a method and a device for predicting equipment status. Background Art
[0002] Automatic monitoring of equipment status has become an important technical means to ensure the normal operation of equipment. Equipment status prediction is to predict the probability of equipment failure or the remaining service life of equipment based on equipment operation data.
[0003] At present, for devices to be predicted that are difficult to obtain device status labels, a large amount of labeled data is obtained by obtaining the historical operation data of other devices and the device status label data corresponding to the historical operation data. Then, a mapping relationship between these historical operation data and the device status label data is established based on machine learning methods to obtain a device status prediction model. Based on the device status prediction model, the device status prediction of the device to be predicted can be achieved.
[0004] However, the data distribution of the historical operation data generated by different devices is different, resulting in relatively low accuracy of the equipment status prediction model obtained using the above technical solution. Summary of the invention
[0005] The present invention provides a device state prediction method, device, computer-readable storage medium and electronic device, which can improve the accuracy of the device state prediction model.
[0006] In a first aspect, the present invention provides a method for predicting a device state, comprising:
[0007] Acquire at least two feature data corresponding to a target device, at least two model training data corresponding to at least two reference devices, and device status labels corresponding to each of the model training data, wherein the target device and the reference device are of the same device type;
[0008] Determining the error weight corresponding to the model training data according to each of the feature data;
[0009] Model training is performed according to each of the model training data, the device state labels corresponding to each of the model training data, and the error weights corresponding to each of the model training data to determine a device state prediction model, which is used for device state prediction of the target device.
[0010] In a second aspect, the present invention provides a device state prediction apparatus, comprising:
[0011] an acquisition module, used to acquire at least two feature data corresponding to a target device, at least two model training data corresponding to at least two reference devices, and a device status label corresponding to each of the model training data, wherein the target device and the reference device are of the same device type;
[0012] A weight determination module, used to determine the error weight corresponding to the model training data according to each of the feature data;
[0013] The training module is used to perform model training according to each of the model training data, the device state labels corresponding to each of the model training data, and the error weights corresponding to each of the model training data, so as to determine a device state prediction model, wherein the device state prediction model is used for device state prediction of the target device.
[0014] In a third aspect, the present invention provides a computer-readable storage medium, comprising execution instructions. When a processor of an electronic device executes the execution instructions, the processor executes any method described in the first aspect.
[0015] In a fourth aspect, the present invention provides an electronic device, comprising a processor and a memory storing execution instructions, wherein when the processor executes the execution instructions stored in the memory, the processor executes any method described in the first aspect.
[0016] The present invention provides a device state prediction method, device, computer-readable storage medium and electronic device. The method obtains multiple feature data corresponding to a target device, multiple model training data corresponding to two or more reference devices, and a device state label corresponding to each model training data. The target device and the reference device have the same device type to ensure the value of information obtained by data mining based on the data of the reference device. Then, for each model training data, according to each feature data, the error weight corresponding to the model training data is determined. The obtained error weight indicates the importance of the error generated when the model training data is used to train the model corresponding to the target device, thereby realizing the migration of the model training data of the reference device to the target device. Afterwards, model training is performed according to each model training data, the device state label corresponding to each model training data, and the error weight of each model training data to determine the device state prediction model. The obtained device state prediction model takes into account the error weight of the model training data of the reference device relative to the target device, thereby ensuring the model accuracy. When the device state prediction model is used to predict the device state of the target device, the accuracy of the predicted device state can be ensured. In summary, the technical solution of the present invention can improve the accuracy of the device state prediction model, and when the device state of the target device is subsequently predicted based on the device state prediction model, the accuracy of the predicted device state can be ensured.
[0017] The further effects of the above-mentioned non-conventional preferred manner will be described below in conjunction with specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] In order to more clearly illustrate the embodiments of the present invention or the existing technical solutions, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0019] Figure 1 A schematic diagram of a flow chart of a device state prediction method provided by an embodiment of the present invention;
[0020] Figure 2 A schematic flow chart of another device state prediction method provided by an embodiment of the present invention;
[0021] Figure 3 A schematic diagram of the structure of a device state prediction apparatus provided by an embodiment of the present invention;
[0022] Figure 4 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0023] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0024] like Figure 1 The present invention provides a method for predicting device status, including the following steps:
[0025] Step 101, obtaining at least two feature data corresponding to a target device, at least two model training data corresponding to at least two reference devices, and device status labels corresponding to each of the model training data, wherein the target device and the reference device are of the same device type.
[0026] Specifically, the operating data of the reference device can be obtained based on the sensor installed on the reference device connected to the network, and multiple model training data corresponding to the reference device can be determined based on the operating data of the reference device. It should be noted that obtaining model training data based on data collected by sensors is a prior art, and this embodiment will not be described in detail.
[0027] Specifically, the operating data of the target device can be obtained based on the sensor installed on the target device connected to the network, and multiple feature data corresponding to the target device can be determined based on the operating data of the target device. It should be noted that obtaining feature data based on data collected by sensors is a prior art, and this embodiment will not be described in detail.
[0028] Specifically, the model training data includes several features and each feature corresponds to one or more feature values of the reference device, preferably one, wherein the several features are influencing factors that affect the device state. The several features corresponding to the model training data and the feature data are the same, and the feature values corresponding to the features may be different. Here, the model training data corresponds to a device state label respectively. It should be noted that the embodiments of the present invention do not intend to limit the number of reference devices, which needs to be determined in combination with actual scenarios.
[0029] In one embodiment, the device status label includes but is not limited to device failure information or the remaining service life of the device. It should be noted that the embodiment of the present invention does not intend to limit the method for obtaining the device status label corresponding to each model training data, which can also be manual labeling, rule labeling, or cluster labeling. The above labeling methods are all prior art and will not be described in detail here.
[0030] It should be noted that the reference device and the target device are of the same type, thereby ensuring the value of data association between the reference device and the target device, and further ensuring the reference value of the device state prediction model determined subsequently. For example, the reference device and the target device are both gas boilers.
[0031] It should also be noted that the characteristic data of the target device does not have a device status label and cannot use its own data to complete model training. Therefore, it is necessary to use the model training data of the reference device to obtain its corresponding device status prediction model.
[0032] Step 102: Determine the error weight corresponding to the model training data according to each of the feature data.
[0033] In one embodiment, the error weight corresponding to the model training data can be determined by the following method:
[0034] Determine the replacement data corresponding to the model training data from each feature data; determine the first distribution information of the model training data in each model training data corresponding to the reference device to which it belongs and the second distribution information of the replacement data in each feature data; determine the weight error corresponding to the model training data based on the first distribution information and the second distribution information.
[0035] In order to determine the error weight corresponding to the model training data of the reference device and understand the data relationship between the parameter device and the target device, it is necessary to find the replacement data corresponding to the model training data from the multiple feature data of the target device. Here, the replacement data is preferably the data that is the same as the model training data in the multiple feature data. If there is no data that is the same as the model training data in the multiple feature data, then the replacement data can be the data that is most similar to the model training data in the multiple feature data. As a possible situation, when there are 2 or more feature data that can replace the model training data in the multiple feature data, then the feature data of the replaceable model training data can be randomly selected as the replacement data.
[0036] Then, determine the first distribution information of the model training data in each model training data corresponding to the reference device to which it belongs. Here, the first distribution information indicates the distribution of the model training data in all model training data corresponding to the reference device to which it belongs. The first distribution information can be a distribution probability or an occurrence frequency, which is not specifically limited here.
[0037] Then, the second distribution information of the replacement data in each characteristic data is determined. Here, the second distribution information indicates the distribution of the replacement data in all characteristic data of the target device. The second distribution information can be a distribution probability or an occurrence frequency, which is not specifically limited here. It should be noted that the contents included in the first distribution information and the second distribution information should be unified. For example, the first distribution information and the second distribution information can both include a distribution probability or an occurrence frequency, which is not specifically limited here.
[0038] It should be noted that this embodiment is not intended to limit the method for determining distribution information. Any method for determining data distribution in the prior art may be used, for example, a parameter estimation method, a non-parameter estimation method, or a semi-parameter estimation method.
[0039] In one embodiment, the first distribution information is a first distribution probability, and the second distribution information is a second distribution probability. Specifically, the first distribution probability and the second distribution probability may be determined in the following manner:
[0040] Based on a first non-parametric estimation method corresponding to the reference device to which the model training data belongs, a first distribution probability of the model training data in each model training data corresponding to the reference device to which it belongs is determined; based on a second non-parametric estimation method, a second distribution probability of the feature data in each feature data is determined.
[0041] The non-parametric estimation method is not restricted by the overall distribution and does not assume the specific form of the overall distribution. It tries to obtain the required information from the data or sample itself and obtains the distribution structure through estimation. It plays an extremely important role in reducing bias, improving prediction accuracy and understanding the dynamic structure of data. Especially when the data distribution is very irregular, the non-parametric estimation method is more accurate than the parameter estimation. Therefore, the non-parametric estimation method is used to determine the distribution probability in this embodiment.
[0042] Considering that the data distributions corresponding to the reference device and the target device may be different, in order to ensure the accuracy of the obtained distribution probability, the probability density function of each model training data corresponding to the reference device is determined by a first non-parametric estimation method, and the first distribution probability of each model training data corresponding to the reference device is determined based on the probability density distribution function. The probability density function of each feature data corresponding to the target device is determined by a second non-parametric estimation method, and the second distribution probability of each feature data corresponding to the target device is determined based on the probability density distribution function.
[0043] Considering that the data volume and data distribution of devices may be different, if the same non-parametric estimation method is used, the accuracy of the estimated data distribution may be reduced. Therefore, it is necessary to select a suitable non-parametric estimation method to estimate the data distribution in combination with the data volume and data distribution of each reference device and the target device. When the difference between the data volume of each model training data corresponding to each reference device and the data volume of each feature data corresponding to the target device is small, the first non-parametric estimation method corresponding to each reference device and the second non-parametric estimation method corresponding to the target device are the same. As a possible situation, the first non-parametric estimation method and the second non-parametric estimation method are both kernel density estimation methods.
[0044] In one embodiment, when the first distribution information is a first distribution probability and the second distribution information is a second distribution probability, the error weight corresponding to the model training data may be determined in the following manner:
[0045] For each model training data, a distribution probability ratio of a second distribution probability of replacement data corresponding to the model training data and a first distribution probability of the model training data is determined, and the distribution probability ratio is determined as an error weight corresponding to the model training data.
[0046] In one embodiment, when the first distribution information is the first occurrence frequency and the second distribution information is the second occurrence frequency, the error weight corresponding to the model training data may be determined in the following manner:
[0047] For each model training data, determine the frequency ratio of the second occurrence frequency of the replacement data corresponding to the model training data to the first occurrence frequency of the model training data, and determine the frequency ratio as the error weight corresponding to the model training data.
[0048] Step 103, performing model training according to each of the model training data, the device state label corresponding to each of the model training data, and the error weight corresponding to each of the model training data, to determine a device state prediction model, wherein the device state prediction model is used for device state prediction of the target device.
[0049] In one embodiment, the device status prediction model can be a neural network model or a regression model, which is not limited here and needs to be determined in combination with actual conditions.
[0050] In one embodiment, the device state prediction method provided in this embodiment is applied to an edge server or a local server corresponding to the target device, thereby reducing the computing pressure of the cloud server.
[0051] Specifically, the device state prediction model uses multiple model training data and the device state label data corresponding to each model training data as training data, and adjusts the model parameters of the device state prediction model through the error weights corresponding to each model training data, so that the adjusted model parameters can reflect the relationship between the characteristic data and the device state of the target device, thereby ensuring the model accuracy of the device state prediction model.
[0052] In one embodiment, the device state prediction model may be determined by:
[0053] A1. Determine the first errors corresponding to the model training data respectively according to the prediction results of substituting the model training data into the model to be predicted and the device status labels corresponding to the model training data, and determine the second errors corresponding to the model training data according to the first errors corresponding to the model training data and the error weights;
[0054] A2, determining whether the number of iterations or the second errors corresponding to each model training data meet the preset conditions, if yes, determining the model to be predicted as the device state prediction model, if no, executing A3;
[0055] A3. According to the second errors corresponding to the training data of each model, the model parameters in the model to be predicted are adjusted to determine the adjusted model parameters, and the model parameters in the model to be predicted are replaced with the adjusted model parameters, and A1 is executed.
[0056] Here, whether the second errors corresponding to each model training data meet the preset conditions can be whether the mean of the second errors corresponding to each model training data is less than a threshold, or whether the standard deviation of the second errors corresponding to each model training data is less than a threshold, which is not specifically limited here.
[0057] Specifically, the prediction result can be understood as device status information, and the first error is the degree of closeness between the prediction result and the device status label. For example, the prediction result can be a fault or not a fault. When the prediction result is a fault and the device status label is also a fault, the first error can be 0, otherwise it is 1.
[0058] It can be seen from the above technical scheme that the beneficial effect of this embodiment is: by determining the error weights of multiple model training data corresponding to multiple reference devices respectively relative to the target device, the multiple model training data corresponding to the multiple reference devices are migrated to the target device. Subsequently, model training is performed using the multiple model training data corresponding to the multiple reference devices, the device state label corresponding to each model training data, and the weight error corresponding to each model training data to obtain a device state prediction model for predicting the device state of the target device. The obtained device state prediction model has relatively high accuracy because it comprehensively considers the error weights of the model training data for the target device. When the device state of the target device is subsequently predicted based on the device state prediction model, a relatively accurate device state of the target device can be obtained.
[0059] Figure 1 What is shown is only a basic embodiment of the method of the present invention. By performing certain optimization and expansion on this basis, other preferred embodiments of the method can be obtained.
[0060] like Figure 2 FIG. 1 is another specific embodiment of the device state prediction method of the present invention. Based on the above embodiments, this embodiment is described in more detail in combination with the application scenario. The method specifically includes the following steps:
[0061] Step 201: Obtain at least two feature data corresponding to a target device, at least two model training data corresponding to at least two reference devices, and device status labels corresponding to each of the model training data, wherein the target device and the reference device are of the same device type.
[0062] Assume that the three reference devices are boiler A, boiler B and boiler C, and the target device is boiler D. Boiler A, boiler B and boiler C correspond to N model training data respectively. Each model training data corresponds to a fault label, which can be faulty or normal. Boiler D corresponds to n feature data. The feature data and the model training data correspond to several features that are the same, but the feature values corresponding to each feature may be different.
[0063] Step 202: Determine a first distribution probability of the model training data in each of the model training data corresponding to the reference device to which the model training data belongs based on a first non-parametric estimation method corresponding to the reference device to which the model training data belongs.
[0064] Assuming that the first non-parametric estimation methods corresponding to boiler A, boiler B, and boiler C are all kernel density estimation methods, for boiler A, based on the kernel density estimation method, the probability density distribution function of N model training data corresponding to boiler A is determined, and for each model training data in boiler A, based on the probability density distribution function of N model training data corresponding to boiler A, the first distribution probability P1 of the model training data in all model training data in boiler A is determined. The calculation method of the first distribution probability P1 of each model training data corresponding to boiler B and boiler C is similar to that of boiler A, and will not be elaborated here.
[0065] Step 203: for each of the model training data, determine the replacement data corresponding to the model training data from each of the feature data; and determine the second distribution probability of the replacement data in each of the feature data based on a second non-parametric estimation method.
[0066] For each model training data corresponding to boiler A, boiler B and boiler C respectively, the replacement data corresponding to the model training data is determined from each feature data; based on the kernel density estimation method, the probability density distribution function of the n feature data corresponding to boiler D is determined; for each replacement data, based on the probability density distribution function of the n feature data corresponding to boiler D, the second distribution probability P2 of the replacement data in all feature data in boiler D is determined.
[0067] Step 204: for each of the model training data, determine a distribution probability ratio of a second distribution probability of the replacement data corresponding to the model training data and a first distribution probability of the model training data; and determine the distribution probability ratio as an error weight corresponding to the model training data.
[0068] For each model training data, the ratio P2 / P1 of the second distribution probability P2 of the replacement data corresponding to the model training data to the first distribution probability P1 of the model training data is calculated, and P2 / P1 is determined as the error weight W corresponding to the model training data.
[0069] Step 205: Perform model training according to each of the model training data, the device state label corresponding to each of the model training data, and the error weight corresponding to each of the model training data to determine a device state prediction model, wherein the device state prediction model is used for device state prediction of the target device.
[0070] Specifically, the equipment state prediction model is trained using N model training data corresponding to boiler A, boiler B, and boiler C, respectively, and the fault labels corresponding to each model training data, and the model parameters are adjusted based on the error weight W corresponding to each model training data, to finally obtain the equipment state prediction model.
[0071] Based on the same concept as the embodiment of the method of the present invention, please refer to Figure 3 The embodiment of the present invention further provides a device state prediction apparatus, comprising:
[0072] An acquisition module 301 is used to acquire at least two feature data corresponding to a target device, at least two model training data corresponding to at least two reference devices, and device status labels corresponding to each of the model training data, wherein the target device and the reference device are of the same device type;
[0073] A weight determination module 302 is used to determine the error weight corresponding to the model training data according to each of the feature data;
[0074] The training module 303 is used to perform model training according to each of the model training data, the device state labels corresponding to each of the model training data, and the error weights corresponding to each of the model training data, so as to determine a device state prediction model, wherein the device state prediction model is used for device state prediction of the target device.
[0075] In one embodiment, the weight determination module 302 includes: a replacement data determination unit, a distribution information determination unit and an importance determination unit; wherein:
[0076] The replacement data determination unit is used to determine the replacement data corresponding to the model training data from each of the feature data;
[0077] The distribution information determination unit is used to determine the first distribution information of the model training data in each of the model training data corresponding to the reference device to which the model training data belongs and the second distribution information of the replacement data in each of the feature data;
[0078] The importance determination unit is used to determine the error weight corresponding to the model training data according to the first distribution information of the model training data and the second distribution information of the replacement data corresponding to the model training data.
[0079] In one embodiment, the distribution information determination unit includes: a first information determination subunit and a second information determination subunit; wherein,
[0080] The first information determination subunit is used to determine, based on a first non-parametric estimation method corresponding to the reference device to which the model training data belongs, a first distribution probability of the model training data in each of the model training data corresponding to the reference device to which the model training data belongs, and determine the first distribution probability as the first distribution information;
[0081] The second information determination subunit is used to determine the second distribution probability of the replacement data in each of the feature data based on a second non-parametric estimation method, and determine the second distribution probability as the second distribution information.
[0082] In one embodiment, the first non-parametric estimation method and the second non-parametric estimation method are both kernel density estimation algorithms.
[0083] In one embodiment, the weight determination module 302 is used to determine, for each of the model training data, a distribution probability ratio of a second distribution probability of replacement data corresponding to the model training data and a first distribution probability of the model training data, and determine the distribution probability ratio as an error weight corresponding to the model training data.
[0084] In one embodiment, the method is applied to an edge server or a local server corresponding to the target device.
[0085] In one embodiment, the training module 303 includes: a training unit, a judgment unit and an adjustment unit; wherein:
[0086] The training unit is used to determine a first error corresponding to the model training data according to a prediction result of substituting the model training data into a model to be predicted and a device state label corresponding to the model training data, and to determine a second error corresponding to the model training data according to the first error corresponding to the model training data and an error weight;
[0087] The judging unit is used to judge whether the number of iterations or the second errors corresponding to each of the model training data meet the preset conditions, and if so, determine the model to be predicted as the device state prediction model; if not, trigger the adjustment unit;
[0088] The adjustment unit is used to adjust the model parameters in the model to be predicted according to the second errors corresponding to each of the model training data to determine the adjusted model parameters, and replace the model parameters in the model to be predicted with the adjusted model parameters to trigger the training unit.
[0089] Figure 4It is a structural diagram of an electronic device provided by an embodiment of the present invention. At the hardware level, the electronic device includes a processor 401 and a memory 402 storing execution instructions, and in one embodiment also includes an internal bus 403 and a network interface 404. Among them, the memory 402 may include a memory 4021, such as a high-speed random access memory (Random-Access Memory, RAM), and may also include a non-volatile memory 4022 (non-volatile memory), such as at least one disk storage, etc.; the processor 401, the network interface 404 and the memory 402 can be interconnected through an internal bus 403, and the internal bus 403 can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc.; the internal bus 403 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 Only one bidirectional arrow is used in the figure, but it does not mean that there is only one bus or one type of bus. Of course, the electronic device may also include hardware required for other services. When the processor 401 executes the execution instruction stored in the memory 402, the processor 401 executes the method in any embodiment of the present invention, and is at least used to execute the following steps: Figure 1 or Figure 2 The method shown.
[0090] In a possible implementation, the processor reads the corresponding execution instruction from the non-volatile memory into the memory and then runs it, and can also obtain the corresponding execution instruction from other devices to form a device state prediction device at the logical level. The processor executes the execution instruction stored in the memory to implement a device state prediction method provided in any embodiment of the present invention through the executed execution instruction.
[0091] The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor or instructions in the form of software. The above processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The disclosed methods, steps and logic block diagrams in the embodiments of the present invention can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.
[0092] The embodiment of the present invention further provides a computer-readable storage medium, including an execution instruction. When a processor of an electronic device executes the execution instruction, the processor executes the method provided in any embodiment of the present invention. The electronic device may be specifically as follows: Figure 4 The electronic device shown; the execution instruction is a computer program corresponding to the device state prediction device.
[0093] Those skilled in the art will appreciate that the embodiments of the present invention may be provided as methods or computer program products. Therefore, the present invention may be implemented in the form of a complete hardware embodiment, a complete software embodiment, or a combination of software and hardware.
[0094] The various embodiments of the present invention are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.
[0095] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.
[0096] The above description is only an embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.
Claims
1. A device status prediction method, It is characterized in that include: Acquire at least two feature data corresponding to a target device, at least two model training data corresponding to at least two reference devices, and device status labels corresponding to each of the model training data, wherein the target device and the reference device are of the same device type; Determining the error weight corresponding to the model training data according to each of the feature data; Performing model training according to each of the model training data, the device state label corresponding to each of the model training data, and the error weight corresponding to each of the model training data, so as to determine a device state prediction model, wherein the device state prediction model is used for device state prediction of the target device; Determining the error weight corresponding to the model training data according to each of the feature data includes: Determine replacement data corresponding to the model training data from each of the feature data; the replacement data is data identical to the model training data among the multiple feature data, and if the multiple feature data do not contain data identical to the model training data, the replacement data is data most similar to the model training data among the multiple feature data; Determine first distribution information of the model training data in each of the model training data corresponding to the reference device to which the model training data belongs and second distribution information of the replacement data in each of the feature data; Determining an error weight corresponding to the model training data according to first distribution information of the model training data and second distribution information of replacement data corresponding to the model training data; The performing model training according to each of the model training data, the device state label corresponding to each of the model training data, and the error weight corresponding to each of the model training data to determine the device state prediction model includes: A1. Determine a first error corresponding to the model training data according to a prediction result of substituting the model training data into a model to be predicted and a device status label corresponding to the model training data, and determine a second error corresponding to the model training data according to the first error corresponding to the model training data and an error weight; A2, judging whether the number of iterations or the second errors corresponding to each of the model training data meet the preset conditions, if yes, determining the model to be predicted as the equipment state prediction model, if no, executing A3; A3. According to the second errors corresponding to the model training data, the model parameters in the model to be predicted are adjusted to determine the adjusted model parameters, and the model parameters in the model to be predicted are replaced with the adjusted model parameters, and A1 is executed.
2. The method according to claim 1, It is characterized in that The determining of first distribution information of the model training data in each of the model training data corresponding to the reference device to which the model training data belongs and second distribution information of the replacement data in each of the feature data includes: Determine, based on a first non-parametric estimation method corresponding to the reference device to which the model training data belongs, a first distribution probability of the model training data in each of the model training data corresponding to the reference device to which the model training data belongs, and determine the first distribution probability as first distribution information; Based on a second non-parametric estimation method, a second distribution probability of the replacement data in each of the feature data is determined, and the second distribution probability is determined as second distribution information.
3. The method according to claim 2, It is characterized in that The first non-parametric estimation method and the second non-parametric estimation method are both kernel density estimation algorithms.
4. The method according to claim 2, It is characterized in that The determining, according to the first distribution information of the model training data and the second distribution information of the replacement data corresponding to the model training data, the error weight corresponding to the model training data comprises: For each of the model training data, determine the distribution probability ratio of the second distribution probability of the replacement data corresponding to the model training data and the first distribution probability of the model training data, and determine the distribution probability ratio to determine the error weight corresponding to the model training data.
5. The method according to claim 1, It is characterized in that The method is applied to an edge server or a local server corresponding to the target device.
6. A device for predicting equipment status, It is characterized in that include: an acquisition module, used to acquire at least two feature data corresponding to a target device, at least two model training data corresponding to at least two reference devices, and a device status label corresponding to each of the model training data, wherein the target device and the reference device are of the same device type; A weight determination module, used to determine the error weight corresponding to the model training data according to each of the feature data; A training module, used for performing model training according to each of the model training data, the device state label corresponding to each of the model training data, and the error weight corresponding to each of the model training data, so as to determine a device state prediction model, wherein the device state prediction model is used for device state prediction of the target device; The weight determination module is specifically used to: determine the replacement data corresponding to the model training data from each of the feature data; the replacement data is the data that is the same as the model training data in the multiple feature data, and if there is no data that is the same as the model training data in the multiple feature data, then the replacement data is the data that is most similar to the model training data in the multiple feature data; determine the first distribution information of the model training data in each of the model training data corresponding to the reference device to which it belongs and the second distribution information of the replacement data in each of the feature data; Determining an error weight corresponding to the model training data according to first distribution information of the model training data and second distribution information of replacement data corresponding to the model training data; The training module is specifically used to: A1, determine the first error corresponding to the model training data according to the prediction result of substituting the model training data into the model to be predicted and the device state label corresponding to the model training data, and determine the second error corresponding to the model training data according to the first error corresponding to the model training data and the error weight; A2, determine whether the number of iterations is met or whether the second errors corresponding to each of the model training data meet the preset conditions, if yes, determine the model to be predicted as the device state prediction model, if not, execute A3; A3. According to the second errors corresponding to the model training data, the model parameters in the model to be predicted are adjusted to determine the adjusted model parameters, and the model parameters in the model to be predicted are replaced with the adjusted model parameters, and A1 is executed.
7. A computer-readable storage medium, comprising an execution instruction. When a processor of an electronic device executes the execution instruction, the processor executes the method according to any one of claims 1 to 5.
8. An electronic device, comprising a processor and a memory storing execution instructions, wherein when the processor executes the execution instructions stored in the memory, the processor executes the method according to any one of claims 1 to 5.