Method, device and equipment for predicting electric quantity of electric appliance and medium

Abstract

The invention discloses an electric appliance electric quantity prediction method and device, equipment and a medium, and relates to the technical field of computers, and the method comprises the steps: obtaining a first data set of an electric appliance, inputting the first data set into a target encoder, and carrying out the feature extraction, so as to obtain a corresponding target feature; the first data set comprises electricity consumption cycle electricity data of the electric appliance and total electricity consumption data of a user; obtaining a second data set of the electric appliance, and training an encoder and a decoder based on the target feature and the second data set to obtain an optimized encoder and an optimized decoder; the second data set comprises the user total electric quantity data and the electric appliance sub-electric quantity data; and obtaining user total electric quantity data of a target electric appliance, and predicting the power consumption of the target electric appliance based on the user total electric quantity data, the optimized encoder and the optimized decoder to obtain a corresponding electric quantity prediction result. The power consumption of different types of electric appliances can be quickly and accurately predicted.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method, device, equipment and medium for predicting the power consumption of an electrical appliance. Background Art

[0002] Existing methods for energy separation primarily employ supervised learning strategies, using centralized, end-to-end model training based on large amounts of labeled data. Specifically, the model input is the user's total energy consumption data, and the output is the target appliance's energy consumption data for the corresponding time period. These separation models are mostly based on deep learning models, such as recurrent neural networks, convolutional neural networks, Transformers, and their variants. For the energy separation task, existing small-shot learning approaches typically employ meta-learning to construct a set of appliance classification tasks. A classification model is first trained using a large amount of labeled data, and then transferred to a target domain with limited labeled data using transfer learning. For the energy separation task, existing appliance type query encoding methods often employ one-hot encoding. This method assigns a fixed binary vector to each known appliance type, where only one bit is set to 1 to identify the specific appliance category, and all other bits are set to 0. This allows different appliance types to be represented distinctly and used as query input to the model.

[0003] Existing deep learning-based power separation technologies typically train specialized separation models for specific appliances. This approach has the disadvantage that, when new appliances need to be separated, a large, paired dataset containing the appliance's usage information (total user power usage and appliance power usage by item) must be collected before targeted model training can be performed.

[0004] To address the problem of a single model being limited to a single appliance, a study proposed query-based electricity separation based on one-hot encoding. This method inputs the one-hot encoding of the appliance type into the model along with the total electricity consumption, enabling a single model to separate multiple known appliances. However, this query-based model also has flaws. First, the use of fixed-length one-hot encoding limits the system's compatibility with future appliance types, and its scalability remains poor. Second, for new appliance types not included in training, the model still requires sufficient labeled data for joint training or adjustment, and does not completely break away from its reliance on large amounts of new data and retraining.

[0005] When applying few-shot learning to the task of power separation, most current research does not directly address the energy consumption estimation problem, but rather reduces it to the task of appliance classification. These methods typically rely on appliance on / off event detection to identify appliance start / stop times, thereby capturing snippets of power consumption data during the appliance's on-time. This extracted data is then used to train a few-shot classification model capable of identifying appliance types. However, this strategy has several key limitations: First, it introduces a reliance on an additional event detection module, increasing system complexity and potential sources of error. Second, its core objective is classification (determining the appliance), not separation (estimating its energy consumption), which fails to meet the need for quantifying specific power consumption. Third, this architecture does not support "query-based" operation, meaning it cannot dynamically separate the energy consumption of specific appliances based on user-provided examples. Summary of the Invention

[0006] In view of this, the present invention aims to provide a method, device, equipment, and medium for predicting electrical appliance power consumption, which significantly reduces reliance on massive amounts of labeled data and greatly simplifies the model's response to new scenarios. A single trained model can support separate tasks for multiple different appliances, improving the model's reuse rate and application flexibility. The specific solution is as follows:

[0007] In a first aspect, the present application discloses a method for predicting electrical power consumption of an electrical appliance, comprising:

[0008] Obtaining a first data set of the electrical appliance, and inputting the first data set into a target encoder for feature extraction to obtain corresponding target features; the first data set includes power consumption data of the electrical appliance during a power consumption cycle and total power consumption data of a user;

[0009] Obtaining a second data set of the appliance, and training an encoder and a decoder based on the target feature and the second data set to obtain an optimized encoder and an optimized decoder; the second data set includes the user's total power consumption data and the appliance's power consumption data;

[0010] The total power consumption data of the target appliance is obtained, and the power consumption of the target appliance is predicted based on the total power consumption data, the optimized encoder and the optimized decoder to obtain a corresponding power prediction result.

[0011] Optionally, inputting the first data set into a target encoder for feature extraction to obtain corresponding target features includes:

[0012] Inputting the first data sets of the electrical appliances into target encoders respectively; the target encoders include a total power feature encoder, an electrical appliance power feature encoder, and a power consumption differential feature encoder;

[0013] Encoding the user's total power consumption data using the total power characteristic encoder to generate a corresponding first characteristic;

[0014] Encoding the power consumption cycle data using the appliance power consumption feature encoder based on the first feature to obtain a second feature;

[0015] Determining the difference between the total power consumption data of the user and the power consumption cycle data;

[0016] Encoding the differential data using the power consumption differential feature encoder based on the first feature and the second feature to obtain a third feature;

[0017] A corresponding target feature is obtained through the first feature, the second feature, and the third feature.

[0018] Optionally, obtaining a corresponding target feature through the first feature, the second feature, and the third feature includes:

[0019] The first feature, the second feature, and the third feature are spliced ​​together to obtain the target feature.

[0020] Optionally, after inputting the first data set into a target encoder for feature extraction to obtain corresponding target features, the method further includes:

[0021] Determine the cumulative power consumption of the electrical appliance within a target time period, and determine a test sample based on the target time period and the cumulative power consumption;

[0022] Inputting the test sample and the second feature into an appliance code quality detector, so that the appliance code quality detector determines whether the accuracy of the second feature extracted by the appliance power consumption feature encoder meets a preset condition by performing a question-and-answer prediction on the cumulative power consumption of the appliance in any time period;

[0023] If not, the electrical appliance power consumption feature encoder is retrained until the accuracy of the second feature meets the preset condition.

[0024] Optionally, before training the encoder and the decoder based on the target feature and the second data set, the method further includes:

[0025] A masking operation is performed on the user's total power consumption data to obtain masked data.

[0026] Optionally, the training the encoder and the decoder based on the target feature and the second data set to obtain an optimized encoder and an optimized decoder includes:

[0027] An encoder and a decoder are trained based on the target feature, the second data set, the masked data, and the detection sample to obtain an optimized encoder and an optimized decoder.

[0028] In a second aspect, the present application discloses an electrical appliance power prediction device, comprising:

[0029] a target feature acquisition module, configured to acquire a first data set of the electrical appliance, and input the first data set into a target encoder for feature extraction to obtain corresponding target features; the first data set includes power consumption data of the electrical appliance during a power consumption cycle and total power consumption data of the user;

[0030] a training module, configured to obtain a second data set of the appliance, and train an encoder and a decoder based on the target feature and the second data set to obtain an optimized encoder and an optimized decoder; the second data set includes the user's total power consumption data and the appliance's specific power consumption data;

[0031] The power consumption prediction module is used to obtain the user's total power consumption data of the target appliance, and predict the power consumption of the target appliance based on the user's total power consumption data, the optimized encoder and the optimized decoder to obtain a corresponding power consumption prediction result.

[0032] Optionally, the target feature acquisition module includes:

[0033] A data set input unit, configured to input the first data set of the electrical appliances into a target encoder respectively; the target encoder includes a total power feature encoder, an electrical appliance power feature encoder, and a power consumption differential feature encoder;

[0034] a first encoding unit, configured to encode the user's total power consumption data using the total power characteristic encoder to generate a corresponding first characteristic;

[0035] a second encoding unit, configured to encode the power consumption cycle data using the electrical appliance power consumption feature encoder based on the first feature to obtain a second feature;

[0036] a differential data determining unit, configured to determine differential data between the user's total power consumption data and the power consumption cycle power data;

[0037] a third encoding unit, configured to encode the differential data using the power consumption differential feature encoder based on the first feature and the second feature to obtain a third feature;

[0038] A target feature determination unit is configured to obtain a corresponding target feature through the first feature, the second feature, and the third feature.

[0039] In a third aspect, the present application discloses an electronic device, comprising:

[0040] Memory for storing computer programs;

[0041] The processor is used to execute a computer program to implement the aforementioned electrical appliance power prediction method.

[0042] In a fourth aspect, the present application discloses a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the aforementioned electrical appliance power prediction method is implemented.

[0043] The present application first obtains a first data set of an electrical appliance and inputs the first data set into a target encoder for feature extraction to obtain corresponding target features. The first data set includes the electrical appliance's power cycle data and the user's total power consumption data. A second data set of the electrical appliance is obtained, and an encoder and decoder are trained based on the target features and the second data set to obtain an optimized encoder and an optimized decoder. The second data set includes the user's total power consumption data and the appliance's sub-power consumption data. The user's total power consumption data of the target electrical appliance is obtained, and the target electrical appliance's power consumption is predicted based on the user's total power consumption data, the optimized encoder, and the optimized decoder to obtain a corresponding power prediction result. It can be seen that the present application uses an encoder to extract and encode unique power consumption features that can characterize an electrical appliance. These encoded comprehensive features are used as "query conditions" and input into the decoder together with the user's total power consumption time data to be analyzed. Based on this query condition, the decoder predicts and outputs the target electrical appliance's sub-item power consumption time from the input total power consumption. This allows a single trained model to support the separation of multiple known appliances. When a new, untrained appliance type needs to be processed, only a small set of sample data corresponding to the new appliance is required, and query-based separation can be performed directly using the existing model without any model retraining. This significantly improves the flexibility of the power separation system and its adaptability to future appliance diversity; it significantly reduces reliance on massive amounts of labeled data, and increases model reuse and application flexibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0045] Figure 1 This is a flow chart of a method for predicting the power consumption of an electrical appliance disclosed in this application;

[0046] Figure 2 This is a schematic diagram of a specific method for predicting the power consumption of an electrical appliance disclosed in this application;

[0047] Figure 3 This is a schematic diagram of the structure of an electrical appliance power prediction device disclosed in this application;

[0048] Figure 4 This is a structural diagram of an electronic device disclosed in this application. DETAILED DESCRIPTION

[0049] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying 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 making creative efforts are within the scope of protection of the present invention.

[0050] The typical process of current methods is as follows: First, they rely on appliance switch event detection technology to identify the start and stop times of appliances, thereby obtaining a snippet of their power usage data during the on-time period. This extracted data is then used to train a small-sample classification model capable of identifying appliance types. However, this strategy has several key limitations: First, it introduces a reliance on an additional event detection module, increasing system complexity and potential sources of error. Second, its core goal is classification (determining the appliance), not separation (estimating its energy consumption), which fails to meet the need for quantifying specific power usage. Third, this architecture does not support a "query-based" operation mode, meaning it cannot dynamically separate the energy consumption of specific appliances based on user-provided sample examples. To address the above technical issues, this application discloses a method, apparatus, device, and medium for predicting appliance power consumption. These methods significantly reduce the reliance on massive amounts of labeled data and greatly simplify the model's response to new scenarios. A single trained model can support separation tasks for multiple different appliances, improving model reuse and application flexibility.

[0051] See also Figure 1 As shown, an embodiment of the present invention discloses a method for predicting the power consumption of an electrical appliance, comprising:

[0052] Step S11, obtaining a first data set of the electrical appliance, inputting the first data set into a target encoder for feature extraction to obtain corresponding target features; the first data set includes the power consumption cycle data of the electrical appliance and the total power consumption data of the user.

[0053] like Figure 2As shown, the small sample appliance query coding module in this application: through the coupling of multi-dimensional feature coding, a group of small sample complete power consumption cycle data of different appliances and the corresponding user total power consumption data are deeply integrated. The total power consumption, target appliance power consumption and background power consumption (the differential signal of the total power consumption and the target appliance power consumption) feature extraction modules are integrated through cascade and jump connections to construct a power consumption joint coding that integrates multi-dimensional information, providing high-quality power consumption feature information for the subsequent query-type power separation module. Specifically, for a given appliance k, define The sample set for its example query. This set contains M pairs of data (where M is a small integer, such as 20), each pair is a complete power cycle sequence of appliance k collected by the same user in the same time period. And the corresponding user total electricity consumption sequence Where R is a real number, The query samples of different electrical appliances are fed into three specialized encoders. Feature extraction is performed. That is, the first data set of the appliance is input into a target encoder; the target encoder includes a total power feature encoder, an appliance power feature encoder, and a power consumption differential feature encoder; the user's total power consumption data is encoded using the total power feature encoder to generate a corresponding first feature; based on the first feature, the appliance power feature encoder is used to encode the power consumption cycle data to obtain a second feature; differential data between the user's total power consumption data and the power consumption cycle data is determined; based on the first and second features, the power consumption differential feature encoder is used to encode the differential data to obtain a third feature; and the corresponding target feature is obtained using the first, second, and third features.

[0054] In a specific embodiment, the total power characteristic encoder :Sequence of total electricity consumption of users Encode and generate features , which focuses on capturing the macro patterns of total electricity usage. Target appliance feature encoder :exist Based on the complete power cycle sequence of appliance k Encode and generate features This encoder is dedicated to extracting the unique power consumption characteristics of the target appliance while considering the overall power consumption background. :exist and Based on the difference between the total power consumption of the user and the power consumption of appliance k, Encode and generate features The encoder is designed to accurately capture the characteristics of the remaining background (non-target appliances) electricity consumption after removing the target appliance usage from the total electricity consumption, so as to more accurately understand the non-target load pattern in a complex power consumption environment.

[0055] These three encoders Work together to generate corresponding features for their respective coding targets (total power, target appliance, power consumption difference) . It can be implemented through neural network architectures such as recurrent neural networks, convolutional neural networks, and Transformer. Finally, the generated features will be spliced ​​to obtain In other words, the first, second, and third features are concatenated to obtain the target feature. Together, they form the comprehensive query condition for appliance k. This comprehensive query condition serves as input to assist the power separation module in performing efficient query-based separation, thereby accurately identifying and extracting the actual power usage of the target appliance.

[0056] Furthermore, to ensure the encoding module outputs valid and high-quality appliance power usage signatures, an auxiliary verification mechanism is incorporated within the module. Specifically, the series of encoding signatures generated by the appliance power usage signature encoding submodule are fed into an appliance code quality detector simultaneously with the query-based power separation module. This detector operates by receiving an "appliance code detection question" (i.e., a query regarding the appliance's cumulative power usage within a specific time period) and generating a corresponding "appliance code detection answer" (question-and-answer prediction). This question-and-answer verification method ensures that the features extracted by the encoding module accurately reflect the target appliance's actual power usage pattern. Specifically, after inputting the first data set into the target encoder for feature extraction to obtain the corresponding target features, the present application determines the appliance's cumulative power usage within a target time period. A test sample is then generated based on the target time period and the cumulative power usage. This test sample and the second feature are then input into the appliance code quality detector. The appliance code quality detector then determines whether the accuracy of the second feature extracted by the appliance power usage signature encoder meets a preset condition by performing question-and-answer prediction on the appliance's cumulative power usage within any time period. If not, the appliance power usage signature encoder is retrained until the accuracy of the second feature meets the preset condition. Specifically, for a given electrical appliance k, define This is a series of electrical appliance coding quality test samples, including Indicates the start and end point of a specific time period. Indicates the time period The cumulative power consumption of the appliance k. The encoding quality detection samples of different appliances and the corresponding target appliance feature encoder Output features , sent to the electrical appliance encoding quality detector g, thereby assisting in training the target electrical appliance feature encoder The detector g can also be implemented through neural network architectures such as recurrent neural networks, convolutional neural networks, and Transformer. It works by performing question-answering predictions on the cumulative power consumption of the target appliance within a specific time period to ensure The extracted features can accurately reflect the actual power consumption pattern of the target appliance.

[0057] Step S12: Obtain a second data set of the appliance, and train an encoder and a decoder based on the target features and the second data set to obtain an optimized encoder and an optimized decoder; the second data set includes the user's total power consumption data and the appliance's power consumption data.

[0058] In this embodiment, the query type power separation module: This module will use the output of the small sample appliance query coding module as the query condition to predict the power distribution of the application appliance k for a given user's total power sequence. The specific description is as follows: Given an appliance k, let Represents a set of synchronously collected user total power and appliance power distribution sequences, where The total power consumption sequence of the user and the power consumption sequence of the appliance k collected from the same user in the same time period are called training samples of the separation model. . Thus, the present application obtains a second data set of electrical appliances. Afterwards, the encoder and decoder can be trained based on the target features and the second data set. Prior to this, a mask operation is performed on the user's total power consumption data to obtain the masked data. In order to allow the separation module to extract the target appliance power consumption features from the small sample appliance query encoding module as much as possible, before the separation model training, a mask operation is performed on several consecutive segments of the user's total power consumption data, that is, the power consumption subsequence is randomly masked: After obtaining the masked user's total power data, query the code and At the same time, the query code is sent to the anti-noise decoder f to predict the corresponding electrical power distribution: The query code-guided noise-resistant decoder f needs to effectively integrate the comprehensive query code output by the small sample appliance query code module while dealing with missing values ​​in the total power sequence, so as to obtain the corresponding appliance power distribution. f can be implemented through recurrent neural networks, convolutional neural networks, Transformers, etc. It represents the prediction of the electricity consumption of the appliance k.

[0059] Specifically, a set of query-based power separation model training data for hybrid appliances can be constructed , and use these training data to train the encoder and decoder to optimize the following objective function:

[0060] ;

[0061] in, and The encoders are , parameters of detector g and decoder f; is the loss function, such as mean square error. Based on the above process, the optimized encoder and decoder are finally obtained.

[0062] Step S13: Obtain the user's total power consumption data of the target appliance, and predict the power consumption of the target appliance based on the user's total power consumption data, the optimized encoder, and the optimized decoder to obtain a corresponding power consumption prediction result.

[0063] In this embodiment, it is assumed that there is a new type of electrical appliance The training of the query-based power separation model is not involved. First, the total power consumption data of the target appliance is obtained. The first data set Utilizing the trained encoder and decoder, the electrical appliances can be directly Separate the electricity consumption:

[0064] ;

[0065] Where x is a given sequence of total power consumption of a user, Represents the electrical appliances predicted by the query-based electricity separation model The distribution of electricity.

[0066] Specifically, the model receives a small set of paired "total power consumption" samples containing the target appliance's power usage information as conditional input. By analyzing these samples, the model not only learns the target appliance's own power usage characteristics but also captures the dynamic relationship between these characteristics and the user's total power usage. This enables the model to identify the target appliance's power usage characteristics based on a small number of provided examples and effectively separate its power usage, even for "unknown" appliances not encountered during training. This method enables efficient and flexible small-sample query-based power separation. This allows the model to learn and separate power consumption based on only a small set of paired "power cycle-total power consumption" samples of the target appliance, significantly reducing its reliance on massive amounts of labeled data. When dealing with new, untrained appliance types, the model does not need to be retrained; only a small number of examples of the new appliance are required to achieve separation, greatly simplifying the model's response to new scenarios. A single trained model can support separation tasks for a variety of different appliances, improving model reuse and application flexibility, and contributing to a more lightweight and easily deployable system.

[0067] In summary, the present application first obtains a first data set for an appliance and inputs the first data set into a target encoder for feature extraction to obtain corresponding target features. The first data set includes the appliance's power cycle data and the user's total power consumption data. A second data set for the appliance is obtained, and an encoder and decoder are trained based on the target features and the second data set to obtain an optimized encoder and decoder. The second data set includes the user's total power consumption data and appliance's sub-power consumption data. The user's total power consumption data for the target appliance is obtained, and the target appliance's power consumption is predicted based on the user's total power consumption data, the optimized encoder, and the optimized decoder to obtain a corresponding power prediction result. It can be seen that the present application uses an encoder to extract and encode unique power consumption features that can characterize an appliance. These encoded comprehensive features are used as "query conditions" and input into a decoder along with the user's total power consumption time data to be analyzed. Based on this query condition, the decoder predicts and outputs the target appliance's sub-power consumption time from the input total power consumption. This allows a single trained model to support the separation of multiple known appliances. When a new, untrained appliance type needs to be processed, only a small set of sample data corresponding to the new appliance is required, and query-based separation can be performed directly using the existing model without any model retraining. This significantly improves the flexibility of the power separation system and its adaptability to future appliance diversity; it significantly reduces reliance on massive amounts of labeled data, and increases model reuse and application flexibility.

[0068] See also Figure 3 As shown, an embodiment of the present invention discloses an electrical appliance power prediction device, comprising:

[0069] The target feature acquisition module 11 is used to acquire a first data set of the electrical appliance and input the first data set into a target encoder for feature extraction to obtain corresponding target features; the first data set includes the power consumption data of the electrical appliance during the power consumption cycle and the total power consumption data of the user;

[0070] a training module 12 configured to obtain a second data set of the appliance, and train an encoder and a decoder based on the target feature and the second data set to obtain an optimized encoder and an optimized decoder; the second data set includes the user's total power consumption data and the appliance's power consumption data by type;

[0071] The power consumption prediction module 13 is used to obtain the user's total power consumption data of the target appliance, and predict the power consumption of the target appliance based on the user's total power consumption data, the optimized encoder and the optimized decoder to obtain a corresponding power consumption prediction result.

[0072] The present application first obtains a first data set of an electrical appliance and inputs the first data set into a target encoder for feature extraction to obtain corresponding target features. The first data set includes the electrical appliance's power cycle data and the user's total power consumption data. A second data set of the electrical appliance is obtained, and an encoder and decoder are trained based on the target features and the second data set to obtain an optimized encoder and an optimized decoder. The second data set includes the user's total power consumption data and the appliance's sub-power consumption data. The user's total power consumption data of the target electrical appliance is obtained, and the target electrical appliance's power consumption is predicted based on the user's total power consumption data, the optimized encoder, and the optimized decoder to obtain a corresponding power prediction result. It can be seen that the present application uses an encoder to extract and encode unique power consumption features that can characterize an electrical appliance. These encoded comprehensive features are used as "query conditions" and input into the decoder together with the user's total power consumption time data to be analyzed. Based on this query condition, the decoder predicts and outputs the target electrical appliance's sub-item power consumption time from the input total power consumption. This allows a single trained model to support the separation of multiple known appliances. When a new, untrained appliance type needs to be processed, only a small set of sample data corresponding to the new appliance is required, and query-based separation can be performed directly using the existing model without any model retraining. This significantly improves the flexibility of the power separation system and its adaptability to future appliance diversity; it significantly reduces reliance on massive amounts of labeled data, and increases model reuse and application flexibility.

[0073] In some specific embodiments, the target feature acquisition module 11 may specifically include:

[0074] A data set input unit, configured to input the first data set of the electrical appliances into a target encoder respectively; the target encoder includes a total power feature encoder, an electrical appliance power feature encoder, and a power consumption differential feature encoder;

[0075] a first encoding unit, configured to encode the user's total power consumption data using the total power characteristic encoder to generate a corresponding first characteristic;

[0076] a second encoding unit, configured to encode the power consumption cycle data using the electrical appliance power consumption feature encoder based on the first feature to obtain a second feature;

[0077] a differential data determining unit, configured to determine differential data between the user's total power consumption data and the power consumption cycle power data;

[0078] a third encoding unit, configured to encode the differential data using the power consumption differential feature encoder based on the first feature and the second feature to obtain a third feature;

[0079] A target feature determination unit is configured to obtain a corresponding target feature through the first feature, the second feature, and the third feature.

[0080] In some specific embodiments, the target feature determination unit may specifically include:

[0081] The target feature acquisition subunit is used to splice the first feature, the second feature and the third feature to obtain the target feature.

[0082] In some specific embodiments, the device may further include:

[0083] A detection sample determination module is used to determine the cumulative power consumption of the electrical appliance within a target time period, and determine a detection sample based on the target time period and the cumulative power consumption;

[0084] a determination module, configured to input the test sample and the second feature into an appliance code quality detector, so that the appliance code quality detector determines whether the accuracy of the second feature extracted by the appliance power consumption feature encoder meets a preset condition by performing a question-and-answer prediction on the cumulative power consumption of the appliance in any time period;

[0085] The first judgment result execution module is used to retrain the electrical appliance power consumption feature encoder if the condition is not met until the accuracy of the second feature meets the preset condition.

[0086] In some specific embodiments, the device may further include:

[0087] The masking module is used to perform a masking operation on the user's total power consumption data to obtain masked data.

[0088] In some specific embodiments, the training module 12 may specifically include:

[0089] An encoder and a decoder are trained based on the target feature, the second data set, the masked data, and the detection sample to obtain an optimized encoder and an optimized decoder.

[0090] Furthermore, the embodiment of the present application also discloses an electronic device, Figure 4 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content in the diagram should not be considered as any limitation to the scope of application of the present application.

[0091] Figure 4This is a schematic diagram of the structure of an electronic device 20 provided in an embodiment of the present application. The electronic device 20 may include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps of the appliance power prediction method disclosed in any of the aforementioned embodiments. Furthermore, the electronic device 20 in this embodiment may be a computer.

[0092] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device. The communication protocol it follows is any communication protocol that can be applied to the technical solution of this application and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world. Its specific interface type can be selected according to specific application needs and is not specifically limited here.

[0093] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or CD, etc. The resources stored thereon can include an operating system 221, a computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0094] The operating system 221 is used to manage and control the hardware devices on the electronic device 20 and the computer program 222. The operating system 221 can be Windows Server, NetWare, Unix, Linux, etc. In addition to including a computer program capable of implementing the appliance power prediction method performed by the electronic device 20 as disclosed in any of the aforementioned embodiments, the computer program 222 can further include computer programs capable of performing other specific tasks.

[0095] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when executed by a processor, the computer program implements the aforementioned method for predicting the power consumption of an electrical appliance. The specific steps of this method can be found in the corresponding contents disclosed in the aforementioned embodiments and will not be further described here.

[0096] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from the other embodiments. Reference can be made to the descriptions of the identical or similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple, and the relevant parts can be referred to the descriptions of the methods.

[0097] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0098] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0099] Finally, it should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or device comprising the element.

[0100] The above is a detailed introduction to the technical solution provided by the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for those skilled in the art, according to the ideas of the present application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for predicting electrical power consumption of an electrical appliance, characterized in that: include: Acquire a first data set of the electrical appliance, and input the first data set into a target encoder for feature extraction to obtain corresponding target features; The first data set includes the power consumption data of the electrical appliance during the power consumption period and the total power consumption data of the user; Obtaining a second data set of the appliance, and training an encoder and a decoder based on the target feature and the second data set to obtain an optimized encoder and an optimized decoder; the second data set includes the user's total power consumption data and the appliance's power consumption data; The total power consumption data of the target appliance is obtained, and the power consumption of the target appliance is predicted based on the total power consumption data, the optimized encoder and the optimized decoder to obtain a corresponding power prediction result.

2. The electrical appliance power prediction method according to claim 1, characterized in that: Inputting the first data set into a target encoder for feature extraction to obtain corresponding target features includes: Inputting the first data sets of the electrical appliances into target encoders respectively; the target encoders include a total power feature encoder, an electrical appliance power feature encoder, and a power consumption differential feature encoder; Encoding the user's total power consumption data using the total power characteristic encoder to generate a corresponding first characteristic; Encoding the power consumption cycle data using the appliance power consumption feature encoder based on the first feature to obtain a second feature; Determining the difference between the total power consumption data of the user and the power consumption cycle data; Encoding the differential data using the power consumption differential feature encoder based on the first feature and the second feature to obtain a third feature; A corresponding target feature is obtained through the first feature, the second feature, and the third feature.

3. The electrical appliance power prediction method according to claim 2, characterized in that: The acquiring the corresponding target feature through the first feature, the second feature, and the third feature includes: The first feature, the second feature, and the third feature are spliced ​​together to obtain the target feature.

4. The method for predicting electrical power consumption according to claim 2, characterized in that: After inputting the first data set into the target encoder for feature extraction to obtain corresponding target features, the method further includes: Determine the cumulative power consumption of the electrical appliance within a target time period, and determine a test sample based on the target time period and the cumulative power consumption; Inputting the test sample and the second feature into an appliance code quality detector, so that the appliance code quality detector determines whether the accuracy of the second feature extracted by the appliance power consumption feature encoder meets a preset condition by performing a question-and-answer prediction on the cumulative power consumption of the appliance in any time period; If not, the electrical appliance power consumption feature encoder is retrained until the accuracy of the second feature meets the preset condition.

5. The electrical appliance power prediction method according to claim 4, characterized in that: Before training the encoder and the decoder based on the target feature and the second data set, the method further includes: A masking operation is performed on the user's total power consumption data to obtain masked data.

6. The method for predicting electrical power consumption of an electrical appliance according to claim 5, characterized in that: The step of training the encoder and the decoder based on the target feature and the second data set to obtain an optimized encoder and an optimized decoder includes: An encoder and a decoder are trained based on the target feature, the second data set, the masked data, and the detection sample to obtain an optimized encoder and an optimized decoder.

7. An electrical appliance power prediction device, characterized in that: include: a target feature acquisition module, configured to acquire a first data set of the electrical appliance, and input the first data set into a target encoder for feature extraction to obtain corresponding target features; The first data set includes the power consumption data of the electrical appliance during the power consumption period and the total power consumption data of the user; a training module, configured to obtain a second data set of the appliance, and train an encoder and a decoder based on the target feature and the second data set to obtain an optimized encoder and an optimized decoder; the second data set includes the user's total power consumption data and the appliance's specific power consumption data; The power consumption prediction module is used to obtain the user's total power consumption data of the target appliance, and predict the power consumption of the target appliance based on the user's total power consumption data, the optimized encoder and the optimized decoder to obtain a corresponding power consumption prediction result.

8. The electrical appliance power prediction device according to claim 7, characterized in that: The target feature acquisition module includes: A data set input unit, configured to input the first data set of the electrical appliances into a target encoder respectively; the target encoder includes a total power feature encoder, an electrical appliance power feature encoder, and a power consumption differential feature encoder; a first encoding unit, configured to encode the user's total power consumption data using the total power characteristic encoder to generate a corresponding first characteristic; a second encoding unit, configured to encode the power consumption cycle data using the electrical appliance power consumption feature encoder based on the first feature to obtain a second feature; a differential data determining unit, configured to determine differential data between the user's total power consumption data and the power consumption cycle power data; a third encoding unit, configured to encode the differential data using the power consumption differential feature encoder based on the first feature and the second feature to obtain a third feature; A target feature determination unit is configured to obtain a corresponding target feature through the first feature, the second feature, and the third feature.

9. An electronic device, characterized in that: include: Memory for storing computer programs; A processor, configured to execute a computer program to implement the steps of the method for predicting electrical power consumption of an electrical appliance as claimed in any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that A computer program is stored on a computer-readable storage medium, and when the computer program is executed by a processor, the steps of the method for predicting the power consumption of an electrical appliance as claimed in any one of claims 1 to 6 are implemented.

Images