Non-intrusive load monitoring method and system

Abstract

The invention discloses a non-intrusive load monitoring method and system. The method comprises the following steps: collecting real-time first electrical signal data, and carrying out first preprocessing on the real-time first electrical signal data; pre-training a first load monitoring model, wherein the first load monitoring model comprises a first improved loss function; and taking the real-time first electrical signal data as the input of the first load monitoring model, and outputting a load monitoring result according to the first load monitoring model. The implementation cost is reduced, the application scene of the load monitoring technology is widened, and a more convenient and efficient electricity utilization information acquisition mode is provided for family and industrial users.

Description

Technical Field

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

[0002] With the development of society and economy, the global demand for energy has been growing rapidly every year. The increasingly serious energy crisis is also deeply hindering the global implementation of sustainable development. Through the technical method of load monitoring, high - timeliness electricity consumption information can be provided for household or industrial users, helping users to improve their electricity consumption plans and habits in a timely manner, thereby effectively reducing energy waste.

[0003] Traditional intrusive load monitoring methods (ILM) require the installation of devices that can collect electrical information on each electrical device of household or industrial users. The implementation difficulty is relatively high, and it is accompanied by a certain amount of device cost and labor cost. Therefore, the non - intrusive load monitoring method (NILM) proposed by Hart has become the focus of research by researchers in this field in recent years. Compared with intrusive load monitoring, non - intrusive load monitoring greatly reduces the implementation cost and also broadens the application scenarios of load monitoring technology to a certain extent. Currently, the mainstream non - intrusive load monitoring methods mainly include those based on optimization problems and those based on machine learning. The latter also shows good model performance due to the rapid development of artificial intelligence algorithms such as deep learning in recent years. The electrical bus data collected by non - intrusive load monitoring technology is often time - series data. Traditional machine - learning - based non - intrusive load monitoring models fail to effectively extract the interdependent characteristics of data at previous and subsequent time nodes from the sequence data and also lack in - depth exploration of the coupling between multi - dimensional features of load data. 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 non - intrusive load monitoring method and system, which 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 non - intrusive load monitoring method, including:

[0009] Collect real-time first electrical signal data and perform first preprocessing on the real-time first electrical signal data;

[0010] Pre-train a first load monitoring model, where the first load monitoring model includes a first improved loss function;

[0011] Use the real-time first electrical signal data as the input of the first load monitoring model and obtain a load monitoring result according to the output of the first load monitoring model.

[0012] As a preferred solution of the non-intrusive load monitoring method described in the present invention, where: the first load monitoring model includes:

[0013] The first load monitoring model is any model with the first electrical signal data as the input and the load monitoring result or relevant parameters that can directly or indirectly obtain the load monitoring result as the output.

[0014] As a preferred solution of the non-intrusive load monitoring method described in the present invention, where: the first load monitoring model further includes at least a plurality of forward transfer channels, reverse transfer channels, and a channel attention processing module.

[0015] As a preferred solution of the non-intrusive load monitoring method described in the present invention, where: the first load monitoring model further includes:

[0016] Construct a plurality of forward transfer channels;

[0017] Construct a plurality of reverse transfer channels;

[0018] Perform a first combination on the output results of the forward transfer channels and the output results of the reverse transfer channels to form the output of the first load monitoring model.

[0019] As a preferred solution of the non-intrusive load monitoring method described in the present invention, where: the channel attention processing module includes at least a squeezing sub-module and an excitation sub-module;

[0020] The squeezing sub-module is used to globally embed the global electrical information, and use global average pooling in the channel manner to compress the global multi-dimensional feature map into a feature vector;

[0021] The excitation sub-module is used to perform a non-linear transformation on the output of the squeezing sub-module to obtain an adaptive and re-calibrated weight.

[0022] As a preferred solution of the non-intrusive load monitoring method described in the present invention, where: the pre-training of the first load monitoring model includes:

[0023] Obtain a training sample set containing historical first electrical signal data under different scenarios;

[0024] Use the historical first electrical signal data as the input of the first load monitoring model;

[0025] In the training sample set, except for the historical first electrical signal data, it is used as the output of the first load monitoring model.

[0026] As a preferred solution of the non-intrusive load monitoring method of the present invention, wherein: the first preprocessing of the real-time first electrical signal data includes:

[0027] Process the missing values and invalid values in the real-time first electrical signal data, and replace and supplement them with the average value of the bus data at adjacent moments;

[0028] Perform imbalance processing on the real-time first electrical signal data after processing the missing values and invalid values;

[0029] Perform normalization processing on the real-time first electrical signal data after performing imbalance processing.

[0030] In a second aspect, the present invention provides a non-intrusive load monitoring method and system, including:

[0031] A data acquisition module for collecting real-time first electrical signal data and performing first preprocessing on the real-time first electrical signal data;

[0032] A model establishment module for pre-training a first load monitoring model, and the first load monitoring model includes a first improved loss function;

[0033] A monitoring module for using the real-time first electrical signal data as the input of the first load monitoring model and monitoring the load according to the output of the first load monitoring model.

[0034] In a third aspect, the present invention provides a computer device, including a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the steps of the method described above are implemented.

[0035] 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.

[0036] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention provides a non-intrusive load monitoring method and system, which collects real-time first electrical signal data and performs first preprocessing on the real-time first electrical signal data; pre-trains a first load monitoring model, and the first load monitoring model includes a first improved loss function; uses the real-time first electrical signal data as the input of the first load monitoring model, and obtains a load monitoring result according to the output of the first load monitoring model. By using the improved loss function, the sensitivity of the model to load changes and the prediction accuracy are improved, thereby enhancing the reliability of the monitoring result. By using the channel attention processing module, the ability of the model to extract key information in electrical signals is enhanced, enabling the model to better identify and distinguish different load states. By constructing a forward transmission channel and a reverse transmission channel, the model can simultaneously learn the forward and reverse features of electrical signals, further improving the accuracy of load monitoring. By adopting global average pooling and non-linear transformation, the effective integration of global features of electrical signals and the dynamic adjustment of adaptive weights are realized, enhancing the generalization ability of the model. The preprocessing steps of real-time electrical signal data are implemented, including the processing of missing values and invalid values, imbalance processing and normalization processing, ensuring the data quality and providing a more stable and accurate input for model training. It not only reduces the implementation cost, but also broadens the application scenarios of load monitoring technology, providing a more convenient and efficient way for household and industrial users to obtain electricity consumption information. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings without creative efforts. Among them:

[0038] Figure 1 It is a flowchart of a method for a non-intrusive load monitoring method and system provided by an embodiment of the present invention;

[0039] Figure 2 It is a flowchart of a channel attention processing module for a non-intrusive load monitoring method and system provided by an embodiment of the present invention;

[0040] Figure 3 It is an internal structure diagram of a computer device for a non-intrusive load monitoring method and system provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0041] To make the above objects, features, and advantages of the present invention more apparent and understandable, the following provides a detailed description of the specific embodiments of the present invention in conjunction with the accompanying drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present invention.

[0042] Embodiment 1

[0043] Referring to Figures 1 - 3 , which is the first embodiment of the present invention. This embodiment provides a non-intrusive load monitoring method and system, including:

[0044] In the existing related technologies, there are some problems. For example, the accuracy of load monitoring is not high, and it is difficult to accurately capture the load changes in complex power usage environments, resulting in users being unable to obtain accurate power usage information in a timely manner, thereby affecting the improvement of power usage plans and the improvement of power usage habits.

[0045] This application provides a method that can effectively solve the above-mentioned problems. Next, multiple embodiments will be combined to elaborate in detail on how to implement this non-intrusive load monitoring method;

[0046] Figure 1 shows a method flow chart of a non-intrusive load monitoring method and system, including:

[0047] S101, collect real-time first electrical signal data and perform first preprocessing on the real-time first electrical signal data;

[0048] In an optional embodiment, the first electrical signal data can be parameter data such as current, voltage, and power directly obtained from electrical equipment or a power bus. These data reflect the power usage of users at different time points and are the basis for load monitoring.

[0049] In an optional embodiment, process the missing values and invalid values in the time-series electrical data sequence of the first electrical signal data.

[0050] In the embodiments of this application, performing first preprocessing on the real-time first electrical signal data includes:

[0051] Process the missing values and invalid values in the real-time first electrical signal data and replace and supplement them with the mean value of the bus data at adjacent times;

[0052] Perform imbalance processing on the real-time first electrical signal data after processing the missing values and invalid values;

[0053] Perform normalization processing on the real-time first electrical signal data after performing imbalance processing.

[0054] Exemplarily, oversampling or undersampling techniques can be used for imbalance processing to balance signals of different categories or magnitudes in the real-time first electrical signal data, so as to reduce the bias and variance in the model training process and improve the generalization ability of the model. Normalization processing can scale the real-time first electrical signal data to a unified range, enabling data of different dimensions to have the same scale, which helps to accelerate the convergence speed of the model and improve the accuracy of the model.

[0055] It should be noted that collecting the real-time first electrical signal data and performing the first preprocessing on the real-time first electrical signal data can ensure the data quality input into the load monitoring model, reduce the model error caused by data missing, invalidity or imbalance, and improve the accuracy and reliability of the model. By preprocessing the real-time first electrical signal data, the data can be made more stable, which is conducive to the model's sensitive capture of load changes, thereby providing users with more accurate electricity consumption information and helping users better improve their electricity consumption plans and habits. In addition, the preprocessing steps can also reduce the complexity and time cost of subsequent model training, and improve the efficiency and performance of the entire load monitoring system.

[0056] S102, pre-train the first load monitoring model, where the first load monitoring model includes a first improved loss function;

[0057] In the embodiment of the present application, the first load monitoring model includes:

[0058] The first load monitoring model is any model with the first electrical signal data as the input and the load monitoring result or relevant parameters that can directly or indirectly obtain the load monitoring result as the output.

[0059] In an optional embodiment, the first load monitoring model can use a neural network model based on deep learning, such as a convolutional neural network (CNN), a recurrent neural network (RNN), or their variants, such as a long short-term memory network (LSTM) or a gated recurrent unit (GRU). These models can automatically learn and extract features from the input electrical signal data for load monitoring.

[0060] In another optional embodiment, the first load monitoring model can also use an ensemble learning model, such as a random forest, a gradient boosting decision tree (GBDT), or XGBoost, etc. These models can combine the prediction results of multiple weak learners and improve the overall prediction accuracy through voting or weighted averaging, etc.

[0061] In another alternative embodiment, the first load monitoring model may also use a bidirectional temporal convolutional network non-intrusive load monitoring model, which combines a temporal convolutional network and a bidirectional transmission mechanism and can capture the context information of electrical signals simultaneously, further improving the accuracy of load monitoring. The bidirectional temporal convolutional network performs convolutional operations in the forward and backward directions respectively and fuses the results, so as to more comprehensively understand the temporal characteristics of electrical signals and achieve accurate judgment of the load state.

[0062] In the embodiment of the present application, the first load monitoring model uses a bidirectional temporal convolutional network non-intrusive load monitoring model.

[0063] In an alternative embodiment, the relevant parameters that can directly or indirectly obtain the load monitoring result may include the type, power, operating state, etc. of the electrical equipment. These parameters can provide users with detailed electricity consumption information, help users understand the electricity consumption of each device, and thus perform more refined electricity consumption management.

[0064] In the embodiment of the present application, the first load monitoring model further includes at least a plurality of forward transmission channels, reverse transmission channels, and a channel attention processing module.

[0065] In the embodiment of the present application, the first load monitoring model further includes:

[0066] Construct a plurality of forward transmission channels;

[0067] Construct a plurality of reverse transmission channels;

[0068] Perform a first combination on the output results of the forward transmission channels and the output results of the reverse transmission channels to form the output of the first load monitoring model.

[0069] In the embodiment of the present application, the channel attention processing module at least includes a squeezing sub-module and an excitation sub-module;

[0070] The squeezing sub-module is used to globally embed the global electrical information, use global average pooling in the channel manner, and compress the global multi-dimensional feature map into a feature vector;

[0071] The excitation sub-module is used to perform a non-linear transformation on the output of the squeezing sub-module, and then obtain an adaptive and re-calibrated weight.

[0072] In the embodiment of the present application, pre-training the first load monitoring model includes:

[0073] Obtain a training sample set containing historical first electrical signal data under different scenarios;

[0074] Use the historical first electrical signal data as the input of the first load monitoring model;

[0075] In the training sample set, except for the historical first electrical signal data, it is used as the output of the first load monitoring model.

[0076] In an alternative embodiment, a forward propagation channel of a bidirectional temporal convolutional network is constructed. The output expression of the i-th convolutional sublayer in forward propagation is:

[0077]

[0078] In the formula, is the output of the i-th convolutional sublayer at position j in forward propagation, w i,k is the k-th filter of the i-th convolutional sublayer, is the output of the (i - 1)-th convolutional sublayer at position j + k·d in forward propagation i d i is the dilation factor of the i-th convolutional sublayer, and δ is the non-linear activation function.

[0079] In an alternative embodiment, a backward propagation channel of the bidirectional temporal convolutional network is constructed. The output expression of the i-th convolutional sublayer in backward propagation is:

[0080]

[0081] In the formula, is the output of the i-th convolutional sublayer at position j in backward propagation, w i,k is the k-th filter of the i-th convolutional sublayer, is the output of the (i - 1)-th convolutional sublayer at position j + k·d in forward propagation i d i is the dilation factor of the i-th convolutional sublayer, and δ is the non-linear activation function.

[0082] In the embodiment of the present application, the total output of the bidirectional temporal convolutional network is jointly composed of the forward propagation result and the backward propagation result, and the formula is expressed as:

[0083]

[0084] In the formula, and are respectively the outputs of the last convolutional sublayers of forward propagation and backward propagation, and L is the number of convolutional sublayers in the bidirectional temporal convolutional network.

[0085] It should be noted that through the construction of the bidirectional temporal convolutional network, the non-intrusive load monitoring method and system of the present invention can make full use of the temporal characteristics of electrical signals, improving the accuracy and efficiency of load monitoring. The forward propagation channel and the backward propagation channel of the bidirectional temporal convolutional network can capture the forward and backward features of electrical signals respectively, enabling the model to better understand the temporal changes of electrical signals, and thus more accurately judge the load status. At the same time, through the cooperation of the squeezing sub-module and the excitation sub-module, the channel attention processing module realizes the effective extraction of global electrical information and the dynamic adjustment of adaptive weights, further enhancing the generalization ability and robustness of the model.

[0086] In an optional embodiment, a squeezing sub-module is constructed. The squeezing sub-module embeds the overall global electrical information and uses global average pooling in the channel manner to compress the global multi-dimensional feature map into a feature vector, and the formula is expressed as:

[0087]

[0088] where z i is the squeezing result of the i-th channel, x i,j is the input value of the global feature map at the position j of the i-th channel, and H is the weight of the input feature map.

[0089] In an optional embodiment, an excitation sub-module is constructed. The excitation sub-module performs a non-linear transformation on the output of the squeezing sub-module to obtain an adaptive and re-calibrated weight, and the formula is expressed as:

[0090] s i =σ(W 2 δ(W 1 z i ))

[0091] where s i is the channel weighting factor of the i-th channel, σ and δ are non-linear activation functions, W 1 and W 2 are learnable weight matrices, and z i is the output value of the squeezing sub-module of the i-th channel.

[0092] In the embodiment of the present application, the output expression of the channel attention processing module is:

[0093] y i,j =s i ×x i,j

[0094] where y i,j is the output of the channel attention processing module at the position j of the i-th channel. The construction process of the channel attention processing module is as Figure 2 shown.

[0095] In an alternative embodiment, in addition to the squeezing sub-module and the excitation sub-module, an adaptive feature fusion module can also be introduced. This module can dynamically adjust the feature weights of each channel according to the feature importance of different channels, so as to enhance the key features and suppress the redundant features. By introducing the attention mechanism, the adaptive feature fusion module enables the model to pay more attention to the features that have an important impact on the load monitoring results, further improving the accuracy and robustness of the load monitoring.

[0096] In the embodiment of the present application, the two-way temporal convolutional non-intrusive load monitoring model of the constructed embedded channel attention module is trained. When training, the total loss function calculated is a combination of the Huber loss function and the quantile loss function, including the following steps:

[0097] Calculate the value of the Huber loss function, and the formula is expressed as:

[0098]

[0099] In the formula, y is the true value of the data label, f(x) is the predicted value of the data label, and δ is the outlier tolerance control parameter.

[0100] In the embodiment of the present application, calculate the value of the quantile loss function, and the formula is expressed as:

[0101]

[0102] In the formula, y is the true value of the data label, f(x) is the predicted value of the data label, and q is the selected quantile. In the embodiment of the present application, calculate the value of the total loss function, and the formula is expressed as:

[0103] L(y,f(x))=(1-α)L δ (y,f(x))+αL q (y,f(x))

[0104] In the formula, L δ (y,f(x)) is the Huber loss function, L q (y,f(x)) is the quantile loss function, α is the parameter that controls the trade-off between the two loss functions. When α = 0, the total loss function is the Huber loss function, and when α = 1, the total loss function is the quantile loss function.

[0105] It should be noted that for the pre-trained first load monitoring model, the first load monitoring model includes a first improved loss function that can comprehensively consider the overall distribution and local features of the data, improving the adaptability of the model to different types of electrical signals. By introducing the Huber loss function, the robustness to outliers can be maintained when the error is large, avoiding overfitting of the model to noisy data; while when the error is small, the model parameters can be finely adjusted to improve the prediction accuracy. At the same time, the introduction of the quantile loss function enables the model to focus on different quantiles of the data distribution, thereby more comprehensively capturing the changing characteristics of electrical signals and further improving the accuracy of load monitoring.

[0106] In addition, by adjusting the parameter that controls the trade-off between the two loss functions, the robustness and accuracy of the model can be flexibly balanced to meet the requirements of different application scenarios. Therefore, training the bidirectional time convolutional non-intrusive load monitoring model with this combined loss function can significantly improve the performance and practicality of the model.

[0107] S103, Use the real-time first electrical signal data as the input of the first load monitoring model, and obtain the load monitoring result according to the output of the first load monitoring model.

[0108] In an optional embodiment, the load monitoring result can be displayed through a preset visualization interface, so that users can intuitively understand the power consumption of each electrical device. The visualization interface can include the real-time display of information such as the type, power, and operating status of the electrical device, as well as the comparative analysis of historical data, helping users to more comprehensively master the power consumption situation and make more reasonable power consumption management decisions.

[0109] In an optional embodiment, the load monitoring result can also be reminded through a preset alarm mechanism. When the power consumption of the electrical device is abnormal, such as too high power or unstable operating status, the system can automatically trigger an alarm to remind the user to process it in time to avoid potential safety hazards.

[0110] In a specific example, if it is detected that the power of a certain air conditioner device suddenly surges and exceeds the preset safety threshold, the system can immediately notify the user or management personnel through a preset alarm mechanism, such as sending text messages, emails, or popping up a system prompt box. At the same time, the system can also provide a detailed abnormal data report, including information such as the time of the abnormality, the power change curve, and possible fault causes, for the user or management personnel to refer to, so as to quickly locate the problem and take corresponding solutions. Through such a load monitoring system and method, not only can the efficiency of power consumption management be improved, but also safety accidents caused by abnormal power consumption can be effectively prevented, ensuring power consumption safety.

[0111] It should be noted that in this way, the non-intrusive load monitoring method and system of the present invention can provide users with more comprehensive, accurate and real-time power consumption information, help users achieve more refined power consumption management, and improve power consumption efficiency and safety.

[0112] In summary, the present invention proposes a non-intrusive load monitoring method, which collects real-time first electrical signal data and performs first preprocessing on the real-time first electrical signal data; pre-trains a first load monitoring model, and the first load monitoring model includes a first improved loss function; uses the real-time first electrical signal data as the input of the first load monitoring model, and obtains the load monitoring result according to the output of the first load monitoring model. By using the improved loss function, the sensitivity of the model to load changes and the prediction accuracy are improved, thereby enhancing the reliability of the monitoring result. By using the channel attention processing module, the ability of the model to extract key information in the electrical signal is enhanced, enabling the model to better identify and distinguish different load states. By constructing a forward transmission channel and a reverse transmission channel, the model can simultaneously learn the forward and reverse features of the electrical signal, further improving the accuracy of load monitoring. By adopting global average pooling and non-linear transformation, the effective integration of the global features of the electrical signal and the dynamic adjustment of the adaptive weights are realized, enhancing the generalization ability of the model. The preprocessing steps of the real-time electrical signal data are implemented, including the processing of missing values and invalid values, imbalance processing and normalization processing, ensuring the data quality and providing a more stable and accurate input for model training. It not only reduces the implementation cost, but also broadens the application scenarios of load monitoring technology, providing a more convenient and efficient way for household and industrial users to obtain power consumption information.

[0113] Embodiment 2

[0114] In this embodiment, a non-intrusive load monitoring method and system are further provided, including:

[0115] A data acquisition module, configured to collect real-time first electrical signal data and perform first preprocessing on the real-time first electrical signal data;

[0116] A model establishment module, configured to pre-train a first load monitoring model, and the first load monitoring model includes a first improved loss function;

[0117] A monitoring module, configured to use the real-time first electrical signal data as the input of the first load monitoring model and obtain the load monitoring result according to the output of the first load monitoring model.

[0118] The above-mentioned unit modules can be embedded in or independent of the processor in the computer device in the form of hardware, or stored in the memory in the computer device in the form of software, so as to facilitate the processor to call and execute the operations corresponding to the above-mentioned modules.

[0119] This embodiment also provides a computer device, which can be a terminal, and its internal structure diagram can be as shown in Figure 3 shown. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through 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 a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a non-intrusive load monitoring 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, a trackball, or a touchpad provided on the housing of the computer device, or an external keyboard, a touchpad, or a mouse, etc.

[0120] 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:

[0121] Collect real-time first electrical signal data, and perform first preprocessing on the real-time first electrical signal data;

[0122] Pre-train a first load monitoring model, and the first load monitoring model includes a first improved loss function;

[0123] Use the real-time first electrical signal data as the input of the first load monitoring model, and obtain the load monitoring result according to the output of the first load monitoring model.

[0124] Embodiment 3

[0125] Referring to Figure 2 , as an embodiment of the present invention, a non-intrusive load monitoring method and system are provided. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through experiments.

[0126] The following experiments are all carried out using Pytorch in a Python 3 environment.

[0127] A. Dataset

[0128] 1) REDD Dataset: Since its public release in 2011, the REDD dataset has been widely adopted by researchers in the field of NILM. This public dataset collected data on the houses and electrical appliances used by six US households, with a sampling period of 1 second. In addition, high-frequency voltage and current data with a sampling frequency of 15 kHz was also provided.

[0129] 2) UK-DALE Dataset: The UK-DALE dataset was first published in 2014. The public dataset collected power data from five UK households, with a sampling period of 1 second for the main bus data and a sampling period of 6 seconds for each load.

[0130] B. Control Group

[0131] To evaluate the performance of the non-intrusive load monitoring model (CABiTCN) based on bidirectional temporal convolution and channel attention proposed in the present invention, several current state-of-the-art NILM models were used for comparative experiments, including the sequence-to-point convolutional neural network (Seq2Point), the sequence-to-sequence convolutional neural network (Seq2Seq), and the temporal convolutional network (TCN).

[0132] The three models of Seq2Point, Seq2Seq, and TCN were all modified to have the same hyperparameters as the CABiTCN model proposed in the present invention, including: a learning rate of 0.001, a maximum number of training epochs of 100, and a sequence length of 129.

[0133] 2. Experimental Results

[0134] According to the above experimental settings, the four groups of models were trained until convergence, and the recognition accuracy, F1-score, mean absolute error (MAE), and sum of absolute errors (SAE) were selected as the main experimental result comparison metrics.

[0135] A. For the REDD dataset, the data of Buildings 2 and 3 in this dataset were used for the training of each model, and the data of the microwave oven, refrigerator, dishwasher, and washing machine in Building 1 were used for verification. The experimental results of the four models on the REDD dataset are shown in Table 1.

[0136] Table 1. Comparison results of training and testing different models and the model (CABiTCN) proposed in the present invention on REDD

[0137]

[0138]

[0139] B. For the UK-DALE dataset, the data of Building 1 in the dataset is used for model training and validation, which includes data of five household appliances: electric kettle, microwave oven, refrigerator, dishwasher, and washing machine. The experimental results of the four models on the UK-DALE dataset are shown in Table 2.

[0140] Table 2. Comparison results of different models and the model (CABiTCN) proposed in the present invention when training and testing on UK-DALE

[0141]

[0142] As can be seen from Table 1 and Table 2, the non-intrusive load monitoring model based on bidirectional temporal convolution and channel attention proposed in the present invention has improved load recognition accuracy compared with the Seq2Point, Seq2Seq, and TCN models on different datasets, and also has excellent performance in terms of indicators such as F1-score, MAE, and SAE, which proves the effectiveness of the model proposed in the present invention.

[0143] 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 within the scope of the claims of the present invention.

[0144] 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 complete hardware embodiment, a complete 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.

[0145] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, and the combination of processes and / or blocks in the flowchart and / or block diagram, can be realized 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 for realizing in the processFigure 1 means for the functions specified in one or more processes and / or blocks Figure 1 or blocks.

[0146] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction means that implements the functions specified in one process Figure 1 or more processes and / or blocks Figure 1 or blocks.

[0147] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus, such that a series of operational steps are performed on the computer or other programmable apparatus to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one process Figure 1 or more processes and / or blocks Figure 1 or blocks.

[0148] Although the preferred embodiments of the present application have been described, additional changes and modifications can be made by those skilled in the art once they learn the basic creative concept. Therefore, the appended claims are intended to be construed to cover the preferred embodiments as well as all changes and modifications that fall within the scope of the present application.

[0149] Obviously, those skilled in the art can make various changes and variations 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 non-intrusive load monitoring method, characterized in that: include: Collecting real-time first electrical signal data, and performing first preprocessing on the real-time first electrical signal data; pre-training a first load monitoring model, wherein the first load monitoring model includes a first improved loss function; The real-time first electrical signal data is used as an input of a first load monitoring model, and a load monitoring result is output according to the first load monitoring model.

2. The non-intrusive load monitoring method according to claim 1, characterized in that: The first load monitoring model includes: The first load monitoring model is any model whose input is the first electrical signal data and whose output is the load monitoring result or the relevant parameters of the load monitoring result that can be obtained directly or indirectly.

3. The non-intrusive load monitoring method according to claim 2, characterized in that: The first load monitoring model also includes at least a plurality of forward transmission channels, a reverse transmission channel, and a channel attention processing module.

4. The non-intrusive load monitoring method according to claim 3, characterized in that: The first load monitoring model also includes: Construct several forward transmission channels; Construct several reverse transmission channels; The output result of the forward transmission channel and the output result of the reverse transmission channel are first combined to form the output of the first load monitoring model.

5. The non-intrusive load monitoring method according to claim 4, characterized in that: The channel attention processing module at least includes a squeeze submodule and an excitation submodule; The extrusion submodule is used to embed the global electrical information as a whole, and compress the global multi-dimensional feature map into a feature vector using channel-wise global average pooling; The excitation submodule is used to perform nonlinear changes on the output of the squeezing submodule, thereby obtaining adaptive and recalibrated weights.

6. The non-intrusive load monitoring method according to claim 5, characterized in that: The pre-trained first load monitoring model includes: Obtaining a training sample set containing historical first electrical signal data in different scenarios; Using the historical first electrical signal data as input to a first load monitoring model; The historical first electrical signal data is removed from the training sample set as the output of the first load monitoring model.

7. The non-intrusive load monitoring method according to claim 6, characterized in that: The first preprocessing of the real-time first electrical signal data comprises: Processing missing values ​​and invalid values ​​in the real-time first electrical signal data, and replacing and supplementing them with the average of the bus data at adjacent moments; Performing unbalance processing on the real-time first electrical signal data after the missing value and the invalid value are processed; The real-time first electrical signal data after the unbalance processing is normalized.

8. A non-intrusive load monitoring method and system, characterized in that: include: A data acquisition module, used for acquiring real-time first electrical signal data and performing a first preprocessing on the real-time first electrical signal data; A model building module, used for pre-training a first load monitoring model, wherein the first load monitoring model includes a first improved loss function; The monitoring module is used to use the real-time first electrical signal data as an input of a first load monitoring model and output a load monitoring result according to the output of the first load monitoring model.

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