Load identification method and system

Abstract

The invention discloses a load identification method and system. The method comprises the following steps: acquiring a first power sequence of a real-time to-be-identified electric appliance, wherein the first power sequence comprises power sequences of a plurality of electric appliances; pre-training a first load identification model, and obtaining a load identification result in combination with the first power sequence; and performing first pushing operation according to the load identification result. By monitoring and analyzing the power sequence of the electric appliance in real time, the use condition of the electric appliance in a family or office environment can be accurately identified, so that data support is provided for energy management. A pre-trained load identification model is adopted, different family environments and electric appliance types can be rapidly adapted, and the generalization ability and the identification accuracy of the model are improved. The instant pushing of the load identification result is realized, the user can know the use condition of the electric appliance in time, and the consciousness and habit of energy conservation and emission reduction of the user can be cultivated.

Description

Technical Field

[0001] The present invention relates to the technical field of non - intrusive load identification, and particularly to a load identification method and system. Background Art

[0002] With the increasing global attention to climate change, promoting low - carbon economy and achieving sustainable development have become the top priorities of governments around the world. In this context, improving energy use efficiency and reducing energy waste are important ways to achieve the goals of energy conservation and emission reduction. Household appliances account for a significant proportion of household energy consumption. Therefore, effective management and monitoring of their loads are of great significance. Non - intrusive load identification (NILM) technology, with its advantage of not requiring physical modification of electrical appliances, provides an efficient solution that can achieve real - time monitoring of the usage of electrical appliances without disturbing the daily life of users.

[0003] However, current NILM models face many challenges such as poor applicability and difficult deployment. The differences in different household environments, types of household appliances, and usage patterns make it difficult for a single model to meet the load identification requirements of multiple scenarios. In addition, existing models have high requirements for computing resources and real - time performance, often unable to meet the actual needs of ordinary household users. This not only limits their wide application in smart home and energy management systems but also affects users' acceptance of energy - saving technologies. Therefore, developing a lightweight neural network model that can improve the accuracy and adaptability of load identification has become an urgent task. By optimizing algorithms and model structures, this new model will help improve the intelligent management level of household appliances, promote energy conservation and emission reduction, and thus contribute to global sustainable development. Summary of the Invention

[0004] The purpose of this part is to outline some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this part, as well as in the abstract and title of the present application, to avoid obscuring the purpose of this part, the abstract, and the title. However, such simplifications or omissions cannot be used to limit the scope of the present invention.

[0005] In view of the above - mentioned existing problems, the present invention is proposed.

[0006] Therefore, the present invention provides a load identification method and system that can solve the problems mentioned in the background art.

[0007] To solve the above - mentioned technical problems, the present invention provides the following technical solutions:

[0008] In the first aspect, the present invention provides a load identification method, including:

[0009] Obtain the first power sequence of electrical appliances to be recognized in real time, where the first power sequence includes the power sequences of several types of electrical appliances;

[0010] Pre-train the first load recognition model, and obtain the load recognition result by combining the first power sequence;

[0011] Perform the first push operation according to the load recognition result.

[0012] As a preferred solution of the load recognition method described in the present invention, wherein: the first load recognition model includes:

[0013] The first load recognition model is any model with the first power sequence of historical or real-time electrical appliances to be recognized as the input and any parameter that can directly or indirectly obtain the power of the target electrical appliance as the output.

[0014] As a preferred solution of the load recognition method described in the present invention, wherein: the first push operation is any operation that can directly or indirectly clarify the specific load recognition result to the user.

[0015] As a preferred solution of the load recognition method described in the present invention, wherein: the performing the first push operation according to the load recognition result includes:

[0016] When the load recognition result of the first load recognition model is any parameter that can directly or indirectly obtain the power of the target electrical appliance, perform the first preprocessing on the load recognition result of the first load recognition model;

[0017] Perform the first push operation according to the first preprocessing result.

[0018] As a preferred solution of the load recognition method described in the present invention, wherein: the performing the first push operation according to the load recognition result further includes:

[0019] When the load recognition result of the first load recognition model is the power of the target electrical appliance, directly perform the first push operation.

[0020] As a preferred solution of the load recognition method described in the present invention, wherein: the first load recognition model further includes at least an input module, a shared dilated convolution module, a fully connected module, an output module, and an activation function;

[0021] The shared dilated convolution module is composed of a first dilated convolution layer, a max pooling layer, a second dilated convolution layer, a max pooling layer, a third dilated convolution layer, a fourth dilated convolution layer, a fifth dilated convolution layer, and a max pooling layer;

[0022] The fully connected module is composed of several groups of fully connected layers, and each group of fully connected layers is composed of a first fully connected layer and a second fully connected layer.

[0023] As a preferred solution of the load identification method described in the present invention, wherein: the load identification result includes:

[0024] If the target electrical appliance in the load identification result belongs to the type of working electrical appliances pre-existing in the network model, the classification result is the correct classification type;

[0025] If not, the classification result is the type of working electrical appliances with a similarity less than the first threshold among the types of working electrical appliances pre-existing in the network model.

[0026] In a second aspect, the present invention provides a load identification system, including:

[0027] A data acquisition module, configured to acquire a first power sequence of an electrical appliance to be identified in real time, where the first power sequence includes power sequences of several types of electrical appliances;

[0028] A model establishment module, configured to pre-train a first load identification model and obtain a load identification result in combination with the first power sequence;

[0029] A push module, configured to perform a first push operation according to the load identification result.

[0030] In a third aspect, the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, the steps of the method described above are implemented.

[0031] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method described above are implemented.

[0032] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention proposes a load identification method and system, which obtains a first power sequence of electrical appliances to be identified in real time, and the first power sequence includes power sequences of several types of electrical appliances; pre-trains a first load identification model, combines the first power sequence to obtain a load identification result; and performs a first push operation according to the load identification result. By monitoring and analyzing the power sequences of electrical appliances in real time, the usage of electrical appliances in a home or office environment can be accurately identified, thereby providing data support for energy management. By using a pre-trained load identification model, it can quickly adapt to different home environments and electrical appliance types, improving the generalization ability and identification accuracy of the model. The instant push of the load identification result is realized, and users can timely understand the usage of electrical appliances, which helps to cultivate users' awareness and habits of energy conservation and emission reduction. The system design of the present invention is flexible and easy to deploy, can meet the needs of different users for intelligent energy management, and has good market application prospects. Through optimized neural network models and algorithms, the present invention has significant advantages in terms of computing resources and real-time performance, is suitable for use in ordinary home environments, reduces the technical threshold, and improves user acceptance. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings. Among them:

[0034] Figure 1 It is a flowchart of a method for a load identification method and system provided by an embodiment of the present invention;

[0035] Figure 2 It is a structural diagram of a first load identification model for a load identification method and system provided by an embodiment of the present invention;

[0036] Figure 3 It is an internal structural diagram of a computer device for a load identification method and system provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0037] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following will describe the specific embodiments of the present invention in detail with reference to the drawings of the specification. Obviously, the described embodiments are some embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0038] Embodiment 1

[0039] Reference Figures 1 - 3 , which is the first embodiment of the present invention. This embodiment provides a load identification method and system, including:

[0040] In the existing related technologies, there are some problems. For example, the accuracy of load identification is not high, the model generalization ability is poor, and the real-time performance is insufficient, etc. These problems limit the wide application of load identification technology in smart home and energy management systems.

[0041] This application provides a method that can effectively solve the above-mentioned problems. Next, multiple embodiments will be combined to elaborate in detail how to implement this load identification method;

[0042] Figure 1 The method flow chart of a load identification method and system is shown, including:

[0043] S101, obtaining the first power sequence of the electrical appliances to be identified in real time, where the first power sequence includes the power sequences of several electrical appliances;

[0044] In an optional embodiment, the electrical appliances to be identified can be household electrical appliances such as refrigerators, air conditioners, washing machines, TVs, etc., or other types of electrical equipment, such as lighting equipment, water heaters or computers, etc. These electrical appliances will generate their respective power sequences when working, and this power sequence reflects the electricity consumption situation of the electrical appliances. Obtaining these power sequences in real time is the first step in realizing load identification and provides the basic data for subsequent analysis and identification.

[0045] In an optional embodiment, the power sequences of these electrical appliances can be monitored and collected in real time through devices such as smart meters and sensors, so as to provide a data basis for load identification.

[0046] In an optional embodiment, the first power sequence should be collected at a sampling frequency f. Considering the requirements of the device data storage space, the number of parameters and the calculation amount of the load identification model, the sampling frequency f should be small. It is recommended that the sampling frequency f be less than 10Hz.

[0047] In the embodiment of this application, the sampling frequency f is taken as 1Hz.

[0048] In the embodiment of this application, the power sequence of the working electrical appliances is obtained through a smart socket, the electrical signal is collected and subjected to analog-to-digital conversion, and is input into the computer device through UART serial port communication. The baud rate of transmitting data through the UART serial port should be greater than 115200 to ensure stable data transmission.

[0049] It should be noted that obtaining the first power sequence of electrical appliances to be recognized in real time, where the first power sequence includes the power sequences of several types of electrical appliances, can ensure that there is sufficient data input for the load recognition model in the subsequent steps, thereby improving the accuracy and reliability of load recognition. The model establishment module can utilize this power sequence data and combine it with the pre-trained first load recognition model to perform complex feature extraction and pattern recognition to distinguish the electricity consumption situations of different electrical appliances. In addition, the dataset containing the power sequences of multiple electrical appliances can also enhance the generalization ability of the model, enabling it to adapt to different household environments and electrical appliance types, and improving the applicability and practicality of load recognition.

[0050] S102, pre-train the first load recognition model and obtain the load recognition result by combining the first power sequence;

[0051] In the embodiments of the present application, the first load recognition model includes:

[0052] The first load recognition model is any model with the input being the first power sequence of historical or real-time electrical appliances to be recognized and the output being the power of the target electrical appliance or any parameter that can directly or indirectly obtain the power of the target electrical appliance.

[0053] In an optional embodiment, the first load recognition model can be constructed and trained through a deep learning framework, such as TensorFlow or PyTorch. These frameworks provide rich neural network layers, optimization algorithms, and tools, which are helpful for realizing efficient model development and optimization. During the training process, a large amount of historical data containing the power sequences of various electrical appliances can be used as training samples, and the model parameters can be continuously adjusted through the backpropagation algorithm to minimize the prediction error and improve the recognition accuracy.

[0054] In an optional embodiment, in order to further improve the generalization ability of the model, data augmentation techniques, such as adding noise, time shifting, frequency shifting, etc., can be adopted to increase the diversity and richness of the training data.

[0055] In an optional embodiment, a transfer learning strategy can also be adopted, using a model pre-trained in a related field or task as a starting point and adapting it to the load recognition task of the present invention through fine-tuning, thereby accelerating the training process and improving the model performance.

[0056] In an optional embodiment, the first load recognition model can also adopt an ensemble learning method, combining the prediction results of multiple base models, and through voting, averaging, or other combination strategies, to improve the accuracy and stability of the final load recognition. This method can make full use of the advantages of different models, reduce the bias and uncertainty that may be brought by a single model, and thus improve the overall recognition performance. In addition, ensemble learning can also increase the robustness of the model, making it more resistant to noise and abnormal data.

[0057] In the embodiment of the present application, the first load recognition model is selected to be improved based on the Seq2Point model. The Seq2Point model is a deep learning model based on sequence-to-point, which is particularly suitable for processing time series data, such as electrical appliance power sequences. This model can automatically learn features from the input power sequence and predict the power or related parameters of the target electrical appliance.

[0058] In the embodiment of the present invention, the Seq2Point model is improved and optimized to improve its accuracy and efficiency in the load recognition task. By introducing components such as dilated convolutional layers, fully connected layers, and activation functions, the improved model can better capture local and global features in the power sequence, thereby improving the accuracy of load recognition.

[0059] In an alternative embodiment, the training process of the model can also be optimized by adopting strategies such as adaptive learning rate adjustment and early stopping to avoid overfitting and accelerate the training process.

[0060] It should be noted that these improvement measures enable the first load recognition model to perform excellently in practical applications, being able to quickly and accurately identify the usage situations of various electrical appliances, providing strong support for energy management.

[0061] In the embodiment of the present application, the first load recognition model further includes at least an input module, a shared dilated convolutional module, a fully connected module, an output module, and an activation function;

[0062] The shared dilated convolutional module is composed of a first dilated convolutional layer, a max pooling layer, a second dilated convolutional layer, a max pooling layer, a third dilated convolutional layer, a fourth dilated convolutional layer, a fifth dilated convolutional layer, and a max pooling layer;

[0063] The fully connected module is composed of several groups of fully connected layers, and each group of fully connected layers is composed of a first fully connected layer and a second fully connected layer.

[0064] Specifically, in the embodiment of the present application, the working electrical appliance power sequence is input into the load recognition model combining dilated convolution and multi-task learning for classifying working electrical appliances;

[0065] In the embodiment of the present application, the load recognition model is the Seq2Point model, that is, the input parameter of the load recognition model is the bus power sequence Y t:t+W-1 , and the output parameter is x τ ; where the parameter t is the input time, W is the size of the sliding window of the bus power sequence, Y t:t+W-1 is the bus power sequence from time t to t + W - 1, τ is the output time, and x τ is the power of the target electrical appliance at time τ. To accurately obtain the power of the target electrical appliance, In this embodiment, in order to obtain the power sequence feature of the long-time scale working electrical appliance, the parameter W is taken as 600, and τ = t + 300.

[0066] Among them, the network structure of the load recognition model includes an input module, a shared dilated convolution module, a fully connected module, an output module, and an activation function. Among them, the input parameter of the input module is b bus power sequences Y t:t+W-1 , and through linking, a two-dimensional tensor T is formed W×b and output; the shared dilated convolution module is composed of dilated convolution layer 1, max pooling layer, dilated convolution layer 2, max pooling layer, dilated convolution layer 3, dilated convolution layer 4, dilated convolution layer 5, and max pooling layer; the fully connected module is composed of several groups of fully connected layers, and each group of fully connected layers is composed of fully connected layer 1 and fully connected layer 2; the output module is composed of an output layer.

[0067] In the embodiment of the present application, the shared dilated convolution module, by sharing parameters, realizes that when multi-task learning is performed for different scenarios and different users, the same set of feature extraction modules can be shared without the need to repeatedly train the feature extraction modules; the fully connected module is composed of several groups of fully connected layers, and when multi-task learning is performed for different scenarios and different users, different groups of fully connected layers are activated, so as to realize differential learning of multi-tasks.

[0068] In the embodiment of the present application, the difference between the dilated convolution layer and the standard convolution layer is that the receptive field of the convolution kernel of the standard convolution layer is expanded to capture information that cannot be captured by the convolution kernel of the standard convolution layer. The dilated convolution layer is realized by gradually increasing the gap of the convolution kernel. Specifically, the receptive field size of the convolution kernel of the standard convolution layer can be expressed as:

[0069]

[0070] Among them, the parameter ω is the receptive field size, n is the number of layers of the neural network, i is the serial number of the neural network layer, and k is the convolution kernel size.

[0071] In the embodiment of the present application, the receptive field size of the convolution kernel of the dilated convolution layer can be expressed as:

[0072]

[0073] Among them, the parameter b is the base number for gradually increasing the gap length of the convolution kernel of the dilated convolution layer, and the gap size of the dilated convolution layer in the i-th neural network layer is b i . In this embodiment, the parameter b is taken as 2.

[0074] In an alternative embodiment, the activation function includes using the sigmoid function, and the similarity distance metric formula for the input pair (x 1 , x 2 ) is:

[0075]

[0076] where σ(·) is the activation function, and a j is a parameter learned by the model during training and is used to measure the importance of the weighted component distance.

[0077] In an optional embodiment, the output module includes using the softmax function to transform the input pair (x 1 , x 2 ,..., x n ) with the transformation formula:

[0078]

[0079] In an optional embodiment, the load identification classification result includes that if the working electrical appliance to be classified belongs to the type of working electrical appliance pre - existing in the network model, the classification result is the correct classification type; if it does not belong to the type of working electrical appliance pre - existing in the network model, the classification result is the type of working electrical appliance that is the most similar among the pre - stored types, which is the type of working electrical appliance with the most similar electrical characteristics; where the types of working electrical appliances pre - existing in the network model include common working electrical appliances.

[0080] In the embodiment of the present application, the load identification result includes:

[0081] If the target electrical appliance in the load identification result belongs to the type of working electrical appliance pre - existing in the network model, the classification result is the correct classification type;

[0082] If not, the classification result is the type of working electrical appliance in the types of working electrical appliances pre - existing in the network model whose similarity is less than the first threshold.

[0083] Exemplarily, the first threshold can be set to 0.85, indicating that only when the similarity is greater than or equal to 0.85 is the classification result considered accurate. This setting can balance the classification accuracy and generalization ability to a certain extent and avoid misclassifying electrical appliance types that are too dissimilar as the same class.

[0084] In an optional embodiment, the first threshold can be obtained by analyzing a large amount of experimental data and validating the model. Specifically, a part of the power sequence data of known electrical appliance types can be selected as the test set, and the trained load identification model is used to classify and predict this data, and the similarity between the prediction result and the true label is calculated. Then, according to the classification accuracy and similarity distribution, a suitable threshold is selected as the first threshold.

[0085] It should be noted that in this way, it can be ensured that the load recognition model has good classification performance and generalization ability in practical applications, and this application does not make any restrictions.

[0086] It should also be noted that pre-training the first load recognition model and combining the first power sequence to obtain the load recognition result can make full use of the existing data and information, improving the efficiency and accuracy of load recognition. Specifically, through the pre-trained model, when facing a new power sequence of the electrical appliance to be recognized, the model can adapt faster and give accurate recognition results. In addition, the pre-trained model can also utilize the existing knowledge to more effectively extract and recognize the features in the power sequence of the electrical appliance to be recognized, thereby improving the accuracy and stability of load recognition. This is of great significance for realizing energy management and optimization in fields such as smart home and smart grid.

[0087] S103, perform a first push operation according to the load recognition result.

[0088] In the embodiment of this application, the first push operation is any operation that can directly or indirectly clarify the specific load recognition result to the user.

[0089] In an optional embodiment, the first push operation can specifically be to send a notification message containing the load recognition result to the user's smart device. For example, when the load recognition model recognizes that the currently used electrical appliance is an air conditioner, the system can send a notification to the user's smartphone, informing the user that the air conditioner is currently running, and may also include information such as the power and energy consumption of the air conditioner. In this way, the user can timely understand the usage situation of household electrical appliances and make corresponding adjustments or optimizations.

[0090] In an optional embodiment, the first push operation can also include automatically adjusting the household energy distribution according to the load recognition result. For example, when the system recognizes that a high-energy-consuming electrical appliance is running, it can automatically adjust the power of other electrical appliances to ensure that the total household energy consumption does not exceed a preset threshold. This can effectively avoid energy waste and reduce the household electricity cost.

[0091] In an optional embodiment, in order to enhance the user experience, the first push operation can also perform personalized push in combination with the user's usage habits and preferences. For example, for users who often forget to turn off electrical appliances, the system can send a reminder message when it recognizes that the relevant electrical appliance has been running for a long time, suggesting that the user turn off the electrical appliance to save energy. For users who are sensitive to energy consumption, the system can regularly send an energy consumption report to help the user understand the household electricity usage situation and give corresponding energy-saving suggestions.

[0092] In summary, through the first push operation, the load identification system of the present invention can provide users with more convenient and intelligent energy management services, helping users better understand and control household electricity consumption and achieve the goal of energy conservation and emission reduction.

[0093] In the embodiment of the present application, performing the first push operation according to the load identification result includes:

[0094] When any parameter for directly or indirectly obtaining the power of the target electrical appliance is obtained from the load identification result of the first load identification model, perform a first preprocessing on the load identification result of the first load identification model.

[0095] Perform the first push operation according to the first preprocessing result.

[0096] Performing the first push operation according to the load identification result further includes:

[0097] When the load identification result of the first load identification model is the power of the target electrical appliance, directly perform the first push operation.

[0098] In the embodiment of the present application, the specific push methods of the first push operation include but are not limited to displaying the load identification result through the user's smart device or informing the user of the load identification result by means of voice broadcast. Among them, the smart device can be a smart phone, a smart speaker, a smart TV, or other devices that can interact with the user.

[0099] It should be noted that by displaying the load identification result through the smart device, the user can intuitively see the usage situation of current household electrical appliances, including information such as the type, power, and energy consumption of the electrical appliances. The voice broadcast method is more convenient, and the user can obtain relevant information through hearing without having to view the device screen. Both of these methods can effectively improve the user's ability to understand and control household electricity consumption.

[0100] In an optional embodiment, in order to further improve the user experience, the first push operation can also perform intelligent push in combination with the user's usage scenario. For example, at night or during the user's rest period, the system can automatically reduce the push frequency or adopt a softer push method to avoid disturbing the user. During the user's electricity peak period or energy shortage period, the system can increase the push frequency to remind the user to pay attention to energy conservation and emission reduction and use electricity reasonably.

[0101] In an optional embodiment, a user feedback mechanism can also be designed to allow the user to give feedback on the pushed information, such as confirming the accuracy of the information and providing improvement suggestions. The system can continuously optimize the push strategy and content according to the user's feedback to improve the accuracy of the push and user satisfaction. In this way, the load identification system can better adapt to the changes in user needs and provide more personalized and intelligent energy management services.

[0102] In summary, the present invention proposes a load identification method, which obtains the first power sequence of the electrical appliances to be identified in real time. The first power sequence includes the power sequences of several types of electrical appliances; pre-trains the first load identification model, combines the first power sequence to obtain the load identification result; and performs the first push operation according to the load identification result. By monitoring and analyzing the power sequences of electrical appliances in real time, the usage of electrical appliances in a home or office environment can be accurately identified, thereby providing data support for energy management. By using a pre-trained load identification model, it can quickly adapt to different home environments and electrical appliance types, improving the generalization ability and identification accuracy of the model. The instant push of the load identification result is realized, and users can timely understand the usage of electrical appliances, which helps to cultivate users' awareness and habits of energy conservation and emission reduction. The system design of the present invention is flexible and easy to deploy, can meet the needs of different users for intelligent energy management, and has good market application prospects. Through the optimized neural network model and algorithm, the present invention has significant advantages in terms of computing resources and real-time performance, is suitable for use in ordinary home environments, reduces the technical threshold, and improves user acceptance.

[0103] Embodiment 2

[0104] In this embodiment, a load identification system is further provided, including:

[0105] A data acquisition module, configured to acquire the first power sequence of the electrical appliances to be identified in real time, where the first power sequence includes the power sequences of several types of electrical appliances;

[0106] A model establishment module, configured to pre-train the first load identification model and combine the first power sequence to obtain the load identification result;

[0107] A push module, configured to perform the first push operation according to the load identification result.

[0108] The above-mentioned unit modules can be embedded in the processor of the computer device in the form of hardware or be independent of the processor, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above-mentioned modules.

[0109] This embodiment also provides a computer device, which can be a terminal, and its internal structure diagram can be as Figure 3As shown in the figure. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner, and the wireless manner can be achieved through WIFI, carrier network, NFC (Near Field Communication), or other technologies. The computer program, when executed by the processor, implements a load recognition method. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a button, trackball, or touchpad set on the computer device housing, or an external keyboard, touchpad, or mouse, etc.

[0110] This embodiment also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0111] Obtain the first power sequence of the electrical appliances to be recognized in real time, and the first power sequence includes the power sequences of several electrical appliances;

[0112] Pre-train the first load recognition model, and obtain the load recognition result by combining the first power sequence;

[0113] Perform the first push operation according to the load recognition result.

[0114] Embodiment 3

[0115] Refer to Figure 2 , as an embodiment of the present invention, provides a load recognition method and system. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through experiments.

[0116] In order to verify and explain the technical effects adopted in this method, a classic deep learning model is deployed into a smart socket and compared with the method of the present invention through tests, and the experimental results are compared by scientific demonstration means to verify the real effects of this method.

[0117] A classic deep learning model is deployed into the same type of test equipment, and the algorithms adopted are LSTM and Seq2Seq. Table 1 is the test environment table, and Table 2 is the experimental result comparison table.

[0118] Test environment: As shown in Table 1, the classic deep learning model is deployed on Test Equipment 1, and the method of the present invention is deployed on Test Equipment 2. The recognition errors MAE and SAE are statistically calculated, and the calculation formulas are:

[0119]

[0120]

[0121] Among them, the parameter x t is the predicted power of the target working device at time t, and x t is the actual power of the target working device at time t. is the predicted bus power, and r is the actual bus power, that is

[0122] Taking the target working electrical appliance as a refrigerator as an example, the results are shown in Table 2.

[0123] Table 1 Comparative test diagram

[0124]

[0125] Table 2

[0126] Test group Test device 1 Test device 2 Test device 3 MAE 18.45 28.43 14.34 SAE 0.008 0.032 0.001

[0127] The experimental results show that compared with the test devices using classical deep learning models (such as LSTM, Seq2Seq), the test device using the method of the present invention shows lower MAE (Mean Absolute Error) and SAE (Sum of Absolute Squared Errors) in terms of recognition error. Specifically, the MAE obtained by using the LSTM algorithm in Test Device 1 is 18.45, and the SAE is 0.008; the MAE obtained by using the Seq2Seq algorithm in Test Device 2 is 28.43, and the SAE is 0.032; while the MAE obtained by using the method of the present invention in Test Device 3 is only 14.34, and the SAE is as low as 0.001. This result indicates that the method of the present invention has high accuracy and stability in load recognition, and is superior to traditional deep learning models.

[0128] In addition, through comparative tests, it can also be found that the method of the present invention shows good adaptability in identifying different types of working electrical appliances. Whether it is in identifying a specific working electrical appliance such as a refrigerator or in the identification of other working electrical appliances, the method of the present invention can achieve lower recognition errors, verifying its generalization ability and practicality.

[0129] In summary, the load recognition method and system proposed by the present invention realize the accurate, fast acquisition and push of load recognition results by using an optimized neural network model and algorithm. The experimental results show that this method is superior to traditional deep learning models in terms of accuracy and stability, and has high generalization ability and practicality. Therefore, the present invention has broad application prospects in the energy management and optimization of fields such as smart home and smart grid.

[0130] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

[0131] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present application can be implemented in various computer languages. For example, object-oriented programming languages such as Java and interpreted scripting languages such as JavaScript.

[0132] The present application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0133] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implements the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0134] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide means for implementing the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1Steps of the functions specified in one or more boxes.

[0135] Although the preferred embodiments of the present application have been described, those skilled in the art can make additional changes and modifications to these embodiments once they learn the basic creative concepts. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications falling within the scope of the present application.

[0136] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application is also intended to include these modifications and variations.

Claims

1. A load identification method, characterized in that: include: Acquire a first power sequence of electrical appliances to be identified in real time, wherein the first power sequence includes power sequences of several electrical appliances; Pre-training a first load identification model, and obtaining a load identification result in combination with the first power sequence; A first push operation is performed according to the load identification result.

2. The load identification method according to claim 1, characterized in that: The first load identification model includes: The first load identification model is any model whose input is a first power sequence of a historical or real-time electrical appliance to be identified, and whose output is a target electrical appliance power or any parameter that can directly or indirectly obtain the target electrical appliance power.

3. The load identification method according to claim 2, characterized in that: The first push operation is any operation that can directly or indirectly make clear the specific load identification result to the user.

4. The load identification method according to claim 3, characterized in that: The performing a first push operation according to the load identification result includes: If the load identification result of the first load identification model is any parameter that can directly or indirectly obtain the power of the target electrical appliance, performing a first preprocessing on the load identification result of the first load identification model; A first push operation is performed according to the first preprocessing result.

5. The load identification method according to claim 4, characterized in that: The performing a first push operation according to the load identification result further includes: If the load identification result of the first load identification model is the target electrical appliance power, the first push operation is directly performed.

6. The load identification method according to claim 5, characterized in that: The first load identification model also includes at least an input module, a shared hole convolution module, a fully connected module, an output module and an activation function; The shared dilated convolution module consists of a first dilated convolution layer, a maximum pooling layer, a second dilated convolution layer, a maximum pooling layer, a third dilated convolution layer, a fourth dilated convolution layer, a fifth dilated convolution layer, and a maximum pooling layer; The fully connected module is composed of several groups of fully connected layers, and each group of fully connected layers is composed of a first fully connected layer and a second fully connected layer.

7. The load identification method according to claim 6, characterized in that: The load identification result includes: If the target electrical appliance in the load identification result belongs to the working electrical appliance type pre-stored in the network model, the classification result is a correct classification type; If not, the classification result is a working electrical appliance type whose similarity is less than a first threshold among the working electrical appliance types pre-stored in the network model.

8. A load identification system, characterized in that: include: A data acquisition module, used to acquire a first power sequence of electrical appliances to be identified in real time, wherein the first power sequence includes power sequences of several electrical appliances; A model building module, used for pre-training a first load identification model, and obtaining a load identification result in combination with the first power sequence; A push module is used to perform a first push operation according to the load identification result.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images