Non-intrusive load monitoring method, device, equipment, medium and program product

Abstract

The invention relates to a non-intrusive load monitoring method, device and equipment, a medium and a program product. The method comprises the steps that the total power sequence of a target building is acquired, the target building comprises a plurality of load devices, and the total power sequence comprises the total power of the plurality of load devices at different moments; inputting the total power sequence into a load monitoring model to determine the power and the working state of each load device at different moments; the load monitoring model is used for performing multi-target power estimation and multi-target state identification, the multi-target power estimation is used for determining a power sequence of each load device, and the multi-target state identification is used for determining a first working state sequence of each load device. By adopting the method, the load power and working state identification accuracy can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of power load monitoring, and in particular to a non-intrusive load monitoring method, device, equipment, medium and program product. Background Art

[0002] The non-intrusive load monitoring method only requires the installation of a measuring device at the power supply entrance of the target building. By analyzing the total power data, the power consumption information of the internal load equipment can be obtained and its power consumption can be estimated. This is a new type of power load monitoring method.

[0003] In the prior art, the total power data is analyzed through deep learning or pattern recognition technology to decompose and identify the working status of a single load device. However, when analyzing the working status of the load device in the existing method, there is a problem of low accuracy of load monitoring. Summary of the invention

[0004] Based on this, it is necessary to provide a non-invasive load monitoring method, device, equipment, medium and program product that can improve the accuracy of load power and working status identification in response to the above technical problems.

[0005] In a first aspect, the present application provides a non-intrusive load monitoring method, comprising:

[0006] Obtain a total power sequence of a target building, where the target building includes multiple load devices, and the total power sequence includes the total power of the multiple load devices at different times; input the total power sequence into a load monitoring model to determine the power and working status of each load device at different times; the load monitoring model is used to perform multi-objective power estimation and multi-objective state identification, where the multi-objective power estimation is used to determine the power sequence of each load device, and the multi-objective state identification is used to determine the first working state sequence of each load device.

[0007] In one embodiment, the load monitoring model includes an input layer, a feature extraction layer and an output layer. The feature extraction layer is used to extract multi-task shared features. The output layer includes a multi-target fractional regression task and a multi-classification task. The multi-target fractional regression task is used to perform multi-target power estimation based on the shared features, and the multi-classification task is used to perform multi-target state identification based on the shared features.

[0008] In one embodiment, the training process of the load monitoring model includes: obtaining historical electricity consumption data, the historical electricity consumption data includes total power, power and working status of each load device; determining training samples based on the historical electricity consumption data, using standard back propagation to optimize the objective function to train the initial load monitoring model to obtain the load monitoring model, and the objective function is determined based on a first objective function corresponding to multi-objective power estimation and a second objective function corresponding to multi-objective state identification.

[0009] In one embodiment, the method further comprises:

[0010] The second working state sequence of each load device is determined according to the power sequence of each load device; if at a certain moment, the first working state of the load device is inconsistent with the second working state corresponding to the load device, the target working state of the load device is determined according to a preset decision method.

[0011] In one of the embodiments, a second working state sequence of the load devices is determined based on the power sequence of each load device, including: obtaining a start-up power threshold of each load device, each start-up power threshold being determined based on a historical power curve of the load device; if the power at a certain moment in the power sequence of the load device is greater than or equal to the start-up power threshold corresponding to the load device, then determining that the second working state of the load device at that moment is the start-up state.

[0012] In one of the embodiments, the method further includes: if the power of the load device at a certain moment in the power sequence is less than the start-up power threshold corresponding to the load device, determining that the second working state of the load device at that moment is the off state.

[0013] In a second aspect, the present application also provides a non-intrusive load monitoring device, comprising:

[0014] An acquisition module is used to acquire a total power sequence of a target building, where the target building includes multiple load devices, and the total power sequence includes the total power of the multiple load devices at different times;

[0015] The determination module is used to input the total power sequence into the load monitoring model to determine the power and working status of each load device at different times; the load monitoring model is used to perform multi-objective power estimation and multi-objective state identification, the multi-objective power estimation is used to determine the power sequence of each load device, and the multi-objective state identification is used to determine the first working state sequence of each load device.

[0016] In one embodiment, the load monitoring model includes an input layer, a feature extraction layer and an output layer. The feature extraction layer is used to extract multi-task shared features. The output layer includes a multi-target fractional regression task and a multi-classification task. The multi-target fractional regression task is used to perform multi-target power estimation based on the shared features, and the multi-classification task is used to perform multi-target state identification based on the shared features.

[0017] In one embodiment, the non-intrusive load monitoring device also includes a training model, which is used to obtain historical power consumption data, the historical power consumption data including total power, power and working status of each load device; determine training samples based on the historical power consumption data, use standard back propagation optimization objective function to train the initial load monitoring model to obtain the load monitoring model, and the objective function is determined based on a first objective function corresponding to multi-objective power estimation and a second objective function corresponding to multi-objective state identification.

[0018] In one of the embodiments, the determination module is also used to determine the second working state sequence of each load device according to the power sequence of each load device; if at a certain moment, the first working state of the load device is inconsistent with the second working state corresponding to the load device, the target working state of the load device is determined according to a preset decision method.

[0019] In one of the embodiments, the determination module is also used to obtain the start-up power threshold of each load device, and each start-up power threshold is determined based on the historical power curve of the load device; if the power at a certain moment in the power sequence of the load device is greater than or equal to the start-up power threshold corresponding to the load device, then it is determined that the second working state of the load device at this moment is the start-up state.

[0020] In one embodiment, the determination module is further used to determine that the second working state of the load device at a certain moment in the power sequence of the load device is the off state if the power at that moment in the power sequence of the load device is less than the start-up power threshold corresponding to the load device.

[0021] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, any of the methods described in the first aspect is implemented.

[0022] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements any of the methods described in the first aspect above.

[0023] In a fifth aspect, the present application further provides a computer program product, including a computer program, which, when executed by a processor, implements any of the methods described in the first aspect above.

[0024] The above-mentioned non-intrusive load monitoring method, device, equipment, medium and program product obtain the total power sequence of the target building, the target building includes multiple load devices, and the total power sequence includes the total power of multiple load devices at different times; the total power sequence is input into the load monitoring model to determine the power and working state of each load device at different times; the load monitoring model is used to perform multi-objective power estimation and multi-objective state recognition, the multi-objective power estimation is used to determine the power sequence of each load device, and the multi-objective state recognition is used to determine the first working state sequence of each load device. In this way, by obtaining the total power sequence of the target building at multiple times, and determining the power and working state of each load device at different times based on multi-objective power estimation and multi-objective state recognition in a model, the model simultaneously learns the characteristics of multiple load devices, and learns multiple different related tasks of power estimation and state recognition, which can make the load monitoring model have better generalization ability, and based on the load monitoring model, the accuracy of load monitoring can be effectively improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0026] FIG1 is a schematic flow chart of a non-intrusive load monitoring method in one embodiment;

[0027] FIG2 is a schematic diagram of a model structure of a load monitoring model in one embodiment;

[0028] FIG3 is a flow chart of a load monitoring model training process in one embodiment;

[0029] FIG4 is a schematic flow chart of a non-intrusive load monitoring method in another embodiment;

[0030] FIG5 is a schematic flow chart of a non-intrusive load monitoring method in another embodiment;

[0031] FIG6 is a block diagram of a non-intrusive load monitoring device according to an embodiment;

[0032] FIG. 7 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0033] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0034] The non-intrusive load monitoring method only requires the installation of a measuring device at the power supply entrance of the target building. By analyzing the total power data, the power consumption information of the internal load equipment can be obtained and its power consumption can be estimated. This is a new type of power load monitoring method.

[0035] In the prior art, the total power data is analyzed through deep learning or pattern recognition technology to decompose and identify the working status of a single load device. However, when analyzing the working status of the load device in the existing method, there is a problem of low accuracy of load monitoring.

[0036] In view of this, the present application provides a non-intrusive load monitoring method that can improve the accuracy of load power and working status identification. The non-intrusive load monitoring method provided in the embodiment of the present application, its execution subject can be a non-intrusive load monitoring device, and the non-intrusive load monitoring device can be implemented by software, hardware, or a combination of software and hardware. It can be embedded in or independent of the processor in the computer device in the form of hardware, or it can be stored in the memory of the computer device in the form of software. The computer device can be a server. In the following method embodiments, the execution subject is taken as an example for explanation.

[0037] In an exemplary embodiment, as shown in FIG1 , a non-intrusive load monitoring method is provided, including the following steps 101 to 102. Among them:

[0038] S101, obtaining a total power sequence of a target building.

[0039] The target building includes multiple load devices, and the total power sequence includes the total power of the multiple load devices at different times.

[0040] Optionally, non-intrusive load monitoring (NILM) can install a measuring device at the power supply entrance of the target building to collect the total power of the target building in real time, send the collected total power to the server and store it in the storage unit of the server; or, through the smart grid, the total power of the target building can be obtained from the total electricity meter of the target building in real time, and the total power can be stored in the storage unit of the server in time series.

[0041] Optionally, the total power sequence of the target building may be obtained from a storage unit of the server.

[0042] Optionally, the total power of the target building at time t can be expressed by the following formula. Given a total power, the working state of the load equipment can be determined and the corresponding power of the load equipment can be estimated:

[0043]

[0044] in, is the total power at time t, M represents the number of load devices in the target building, is the power of the mth load device, is the working state of the mth load device, represents electrical and measurement noise that is not considered.

[0045] S102, inputting the total power sequence into the load monitoring model to determine the power and working status of each load device at different times.

[0046] The load monitoring model is used for multi-objective power estimation and multi-objective state identification. The multi-objective power estimation is used to determine the power sequence of each load device. The multi-objective state identification is used to determine the first working state sequence of each load device.

[0047] Optionally, the mapping relationship between the total power sequence and the power consumption of a single load device may be modeled to obtain a load monitoring model, and the load monitoring model obtains the power and working status of a single load device at different times through the total power sequence.

[0048] Optional, let represents the total input power sequence, which is derived from the total power of M load devices at the time 1-T, and Represents the power of each load device, where each load device can have k states, expressed as , so that .matrix Represents the related multi-tag status of M load devices. It can be given The goal is to learn a load monitoring model that predicts the working status of load equipment from the total input power sequence. and power .

[0049] Optionally, before inputting the acquired total power sequence into the load monitoring model, the acquired total power sequence needs to be preprocessed, including deleting singular values ​​and supplementing missing values.

[0050] In one possible implementation, the load monitoring model can be an optimization model that aims to find the best combination of all load devices in the target building so that the difference between the total power of the optimization result and the total power in the acquired total power sequence is less than a preset threshold.

[0051] In another possible implementation, the load monitoring model may be a hidden Markov model, which simulates the start and stop process of the load equipment inside the target building by processing the total power sequence.

[0052] In another possible implementation, the load monitoring model may be a deep neural network model, for example, a long short-term memory (LSTM) network model or a denoising autoencoder (LSTM) model may be used to perform power estimation and state recognition.

[0053] In the above method, the total power sequence of the target building is obtained, and the target building includes multiple load devices. The total power sequence includes the total power of multiple load devices at different times; the total power sequence is input into the load monitoring model to determine the power and working state of each load device at different times; the load monitoring model is used to perform multi-objective power estimation and multi-objective state recognition, and the multi-objective power estimation is used to determine the power sequence of each load device, and the multi-objective state recognition is used to determine the first working state sequence of each load device. In this way, by obtaining the total power sequence of the target building at multiple times, and determining the power and working state of each load device at different times based on multi-objective power estimation and multi-objective state recognition in a model, the model simultaneously learns the characteristics of multiple load devices, and learns multiple different related tasks of power estimation and state recognition, which can make the load monitoring model have better generalization ability, and the accuracy of load monitoring can be effectively improved based on the load monitoring model.

[0054] In an exemplary embodiment, optionally, the load monitoring model includes an input layer, a feature extraction layer and an output layer, the feature extraction layer is used to extract multi-task shared features, the output layer includes a multi-target fractional regression task and a multi-classification task, the multi-target fractional regression task is used to perform multi-target power estimation based on the shared features, and the multi-classification task is used to perform multi-target state identification based on the shared features.

[0055] Optionally, the load monitoring model can be implemented based on a fully connected neural network or a one-dimensional convolutional neural network.

[0056] Exemplarily, the structure of a possible load monitoring model is shown in FIG2 , which is implemented based on a one-dimensional convolutional neural network architecture. In this load monitoring model, the feature extraction layer includes multiple upsampling modules and multiple downsampling modules. The upsampling modules and the downsampling modules use the basic concept of U-Net to map the input total power sequence to a feature map sequence on a freely selected time scale. , which is achieved by computing increasingly high-level features at finer time scales using downsampling (1D convolution) blocks, and combining the high-level features with previously computed local high-resolution features using upsampling (1D transposed convolution) blocks, producing multi-scale features for prediction.

[0057] Optionally, the output layer may include a multi-level one-dimensional convolutional neural network that can capture the feature map sequence output by the feature extraction layer. non-temporal dependencies between tasks and generate latent feature maps shared among multiple tasks The output layer can also include multiple multilayer perceptron (MLP) layers, and the hidden layer size of each multilayer perceptron is related to the load device. Among them, the first MLP layer receives the potential feature map To produce a lower dimensional output feature map , the output feature map As a shared feature space, the remaining MLP layers are based on the output feature map Predict the multi-label working status and power of each load device.

[0058] Alternatively, sharing of feature space can reduce the risk of overfitting of the load monitoring model while introducing inductive bias.

[0059] Optionally, the multi-objective fractional regression task in the output layer can be implemented based on the quantile regression model, and the multi-objective regression can also be called multi-output regression, which can estimate the power of multiple load devices based on a set of shared feature spaces. The joint modeling of multiple load devices can be based on the statistical correlation between multiple load devices, which can effectively improve the breadth of the module compared to modeling each load device separately.

[0060] Optionally, quantile regression can be used to estimate the target variable conditioned on the observed features x (i.e., the features of the total power series) The conditional median or other quantile of (the power series of each load device), where , It can be the network parameters of the multi-target fractional regression task in the output layer.

[0061] In one possible implementation, the multi-classification task in the output layer can implicitly capture the relationship between multiple working states by using Softmax, and one or more working states can be predicted for each input instance.

[0062] In another possible implementation, the multi-classification task in the output layer can be obtained by using a deep neural network to determine the joint probability of multiple labels based on the input features and applying a Sigmoid activation function, and using an additional threshold mechanism to convert the Sigmoid probability into a multi-label output.

[0063] Optionally, the multi-classification task may be to output the working status of multiple load devices simultaneously, where: , It can be the network parameters of the multi-classification task in the output layer.

[0064] The above-mentioned load monitoring model includes an input layer, a feature extraction layer and an output layer. The feature extraction layer is used to extract multi-task shared features. The output layer includes a multi-target fractional regression task and a multi-classification task. The multi-target fractional regression task is used to perform multi-target power estimation based on shared features. The multi-classification task is used to perform multi-target state recognition based on shared features. The power and working state of each load device at different times can be determined based on the load monitoring module. The model simultaneously learns the characteristics of multiple load devices, and learns multiple different related tasks of power estimation and state recognition, which can make the load monitoring model have better generalization ability. Based on the load monitoring model, the accuracy of load monitoring can be effectively improved.

[0065] In an exemplary embodiment, as shown in FIG3 , optionally, the training process of the load monitoring model includes the following steps 301 to 302 . Among them:

[0066] S301, obtaining historical power consumption data, the historical power consumption data including total power, power and working status of each load device.

[0067] Optionally, after obtaining the historical electricity consumption data, the obtained data needs to be preprocessed, and the preprocessing includes deleting singular values, supplementing missing values, dividing the power series of each load device, and quantifying the working status label of each load device.

[0068] S302, determining training samples according to historical electricity consumption data, using standard back propagation to optimize the objective function to train the initial load monitoring model to obtain a load monitoring model.

[0069] The objective function is determined according to a first objective function corresponding to multi-objective power estimation and a second objective function corresponding to multi-objective state identification.

[0070] Optionally, the preprocessed historical electricity consumption data is used as training samples to perform supervised training on the initial load monitoring model.

[0071] Optionally, the objective function can be determined according to the following formula:

[0072]

[0073] in, It can be the second objective function corresponding to multi-objective state recognition, It can be the first objective function corresponding to the multi-objective power estimation, which can be determined by the following formulas:

[0074]

[0075]

[0076] Where T is the total time, is the working state of the mth load device at the tth moment in the training sample, The state of the mth load device at the tth moment output by the load monitoring model; is the quantile level, The values ​​of can include [2.5%, 10%, 50%, 90%, 97.5%], is a set of preset quantiles, is the power of the mth load device at the tth moment in the training sample, is the power of the mth load device at the tth moment output by the load monitoring model at the corresponding quantile level, is the loss function, which can be expressed according to the following formula:

[0077]

[0078] Among them, r can represent the deviation power, .

[0079] Optionally, in multi-objective quantile regression, the conditional distribution of y given X=x can be expressed as a cumulative distribution F(y|X=x):R → [0,1] For a continuous and strictly monotonically increasing cumulative distribution , the quantile function of the random variable y is just the inverse of the cumulative distribution: For all quantile levels Therefore, the entire conditional distribution of the target variable y is approximately Therefore, we can collect a set of Quantile estimates to obtain nonparametric probability density estimates , as shown below:

[0080]

[0081] Optionally, the multi-objective quantile regression provides uncertainty estimates when determining the power of each load device, and the prediction credibility can be determined based on the uncertainty estimates.

[0082] Optionally, you can use a deep learning optimization algorithm for multiple iterations and adjust the learning rate through learning rate scheduling measurements to avoid overfitting and accelerate convergence. For example, use the Adam optimizer for 50 iterations with an initial learning rate of , , the batch size is 128. After the ReduceLROnPlateu scheduler, once the learning stagnates for 5 consecutive iterations, the learning rate is reduced by a factor of 0.1. The Adam optimizer can improve the training effect by adaptively adjusting the learning rate of each parameter. ReduceLROnPlateu automatically adjusts the learning rate during training.

[0083] In an exemplary embodiment, when training the initial load monitoring model, a sliding window method can be used for training. can be the total power sequence with window size L, .

[0084] Exemplarily, through multiple experiments, a preset result can be obtained when the window size L is 100, and near real-time working state recognition and power estimation can be provided.

[0085] Optionally, the performance of the trained load monitoring model can be measured by performance evaluation indicators to reflect the power estimation capability and working state recognition capability of the load monitoring model.

[0086] Optionally, performance evaluation indicators may include , precision, recall and mean absolute error (MAE), among which, It can be determined by the following formula:

[0087]

[0088] in, , , TP (true positive) means that the load device is actually running and the recognition result is also running, FP (false positive) means that the load device is actually not running and the recognition result is running, FN (false negative) means that the load is actually running and the recognition result is not running, TN (true negative) means that the load is actually not running and the recognition result is also not running.

[0089] The mean absolute error (MAE) can be determined by the following formula:

[0090]

[0091] For example, taking the load data of a domestic user as an example, all data are divided into training, verification and test data sets in a ratio of 6:2:2. 10 target load devices are selected, namely: air conditioner, fan, refrigerator, hair dryer, heater, incandescent bulb, laptop, microwave oven, washing machine, vacuum cleaner, and the initial load monitoring model is obtained. The data in the training data set is input into the initial load monitoring model for training and the load monitoring model is obtained. The load monitoring module obtained by training is tested using the test data set, and the following results are obtained, as shown in Table 1:

[0092]

[0093] Optionally, the load monitoring model obtained by evaluating the index can be determined to accurately determine the power and working status of each load device.

[0094] In an exemplary embodiment, as shown in FIG4 , optionally, the method further includes the following steps 401 to 402. Among them:

[0095] S401, determining a second working state sequence of each load device according to a power sequence of each load device.

[0096] In a possible implementation, a start-up power threshold of each load device may be obtained, and the start-up power threshold may be compared with a power sequence of each load device to determine a second working state sequence of each load device.

[0097] Among them, each start-up power threshold is determined according to the historical power curve of the load equipment.

[0098] Optionally, the power threshold can be implemented based on the power sequence to quantile sliding window (seq2quantile) method, which can generate a sequence from the historical power curve of the load device. , and calculate its quantile ,in is the window size of each appliance, is the average on-time duration, is the sampling rate, For example, Take 50% as an example, The quantile means that the quantile is greater than or equal to 50% of the power values ​​in the power sequence of the load device, and the quantile is used as the start-up power threshold of the corresponding load device.

[0099] If the power at a certain moment in the power sequence of the load device is greater than or equal to the start-up power threshold corresponding to the load device, it is determined that the second working state of the load device at this moment is the start-up state.

[0100] If the power of the load device at a certain moment in the power sequence is less than the start-up power threshold corresponding to the load device, it is determined that the second working state of the load device at this moment is the off state.

[0101] Optionally, the quantiles can be used as filters to remove noise that may negatively affect the operating status of the load equipment.

[0102] S402: If at a certain moment, the first working state of the load device is inconsistent with the second working state corresponding to the load device, determine the target working state of the load device according to a preset decision method.

[0103] In one possible implementation, the target working state of the load device can be determined based on the confidence level corresponding to the first working state. For example, when the second working state is on and the first working state is off, and the confidence level of the first working state is greater than or equal to a preset threshold, the target working state of the load device is off, and when the confidence level of the first working state is less than the preset threshold, the target working state of the load device is on.

[0104] In another possible implementation, the first working state and the second working state can be input into a preset evaluation model to determine the target working state of the load equipment according to the output of the evaluation model. The evaluation model can be implemented based on the PCA principal component analysis method or the Topsis method, which is not limited in the embodiments of the present application.

[0105] As an optional implementation, as shown in FIG5 , the non-intrusive load monitoring method provided in the embodiment of the present application may include the following specific steps:

[0106] S501, obtaining a total power sequence of a target building.

[0107] The target building includes multiple load devices, and the total power sequence includes the total power of the multiple load devices at different times.

[0108] S502: Input the total power sequence into the load monitoring model to determine the power and working status of each load device at different times.

[0109] The load monitoring model is used for multi-objective power estimation and multi-objective state identification. The multi-objective power estimation is used to determine the power sequence of each load device, and the multi-objective state identification is used to determine the first working state sequence of each load device.

[0110] The load monitoring model includes an input layer, a feature extraction layer and an output layer. The feature extraction layer is used to extract multi-task shared features. The output layer includes a multi-target fractional regression task and a multi-classification task. The multi-target fractional regression task is used to perform multi-target power estimation based on shared features, and the multi-classification task is used to perform multi-target state recognition based on shared features.

[0111] S503: Obtain the start-up power threshold of each load device.

[0112] Each start-up power threshold is determined according to the historical power curve of the load equipment.

[0113] S504: If the power at a certain moment in the power sequence of the load device is greater than or equal to the start-up power threshold corresponding to the load device, determine that the second working state of the load device at this moment is the start-up state.

[0114] S505: If the power of the load device at a certain moment in the power sequence is less than the start-up power threshold corresponding to the load device, it is determined that the second working state of the load device at this moment is the off state.

[0115] S506: If at a certain moment, the first working state of the load device is inconsistent with the second working state corresponding to the load device, determine the target working state of the load device according to a preset decision method.

[0116] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0117] Based on the same inventive concept, the embodiment of the present application also provides a non-intrusive load monitoring device for implementing the non-intrusive load monitoring method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more non-intrusive load monitoring device embodiments provided below can refer to the limitations of the non-intrusive load monitoring method above, and will not be repeated here.

[0118] In an exemplary embodiment, as shown in FIG6 , a non-intrusive load monitoring device 600 is provided, including: an acquisition module 601 and a determination module 602, wherein:

[0119] An acquisition module 601 is used to acquire a total power sequence of a target building, where the target building includes multiple load devices, and the total power sequence includes the total power of the multiple load devices at different times;

[0120] Determination module 602 is used to input the total power sequence into the load monitoring model to determine the power and working status of each load device at different times; the load monitoring model is used to perform multi-objective power estimation and multi-objective state identification, the multi-objective power estimation is used to determine the power sequence of each load device, and the multi-objective state identification is used to determine the first working state sequence of each load device.

[0121] In an exemplary embodiment, the load monitoring model includes an input layer, a feature extraction layer and an output layer. The feature extraction layer is used to extract multi-task shared features. The output layer includes a multi-target fractional regression task and a multi-classification task. The multi-target fractional regression task is used to perform multi-target power estimation based on the shared features, and the multi-classification task is used to perform multi-target state identification based on the shared features.

[0122] In an exemplary embodiment, the non-intrusive load monitoring device also includes a training model, which is used to obtain historical power consumption data, the historical power consumption data including total power, power and working status of each load device; determine training samples based on the historical power consumption data, use standard back propagation to optimize the objective function to train the initial load monitoring model, and obtain the load monitoring model, wherein the objective function is determined based on a first objective function corresponding to multi-objective power estimation and a second objective function corresponding to multi-objective state identification.

[0123] In an exemplary embodiment, the determination module 602 is also used to determine the second working state sequence of each load device according to the power sequence of each load device; if at a certain moment, the first working state of the load device is inconsistent with the second working state corresponding to the load device, the target working state of the load device is determined according to a preset decision method.

[0124] In an exemplary embodiment, the determination module 602 is also used to obtain the start-up power threshold of each load device, and each start-up power threshold is determined according to the historical power curve of the load device; if the power at a certain moment in the power sequence of the load device is greater than or equal to the start-up power threshold corresponding to the load device, then it is determined that the second working state of the load device at this moment is the start state.

[0125] In an exemplary embodiment, the determination module 602 is further used to determine that the second working state of the load device at a certain moment in the power sequence of the load device is the off state if the power at that moment in the power sequence of the load device is less than the start-up power threshold corresponding to the load device.

[0126] Each module in the above non-intrusive load monitoring device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0127] In an exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be shown in FIG7. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. 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, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a non-intrusive load monitoring method is implemented.

[0128] Those skilled in the art will understand that the structure shown in FIG. 7 is merely a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0129] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0130] Obtain a total power sequence of a target building, where the target building includes multiple load devices, and the total power sequence includes the total power of the multiple load devices at different times; input the total power sequence into a load monitoring model to determine the power and working status of each load device at different times; the load monitoring model is used to perform multi-objective power estimation and multi-objective state identification, where the multi-objective power estimation is used to determine the power sequence of each load device, and the multi-objective state identification is used to determine the first working state sequence of each load device.

[0131] In one embodiment, the load monitoring model includes an input layer, a feature extraction layer and an output layer. The feature extraction layer is used to extract multi-task shared features. The output layer includes a multi-target fractional regression task and a multi-classification task. The multi-target fractional regression task is used to perform multi-target power estimation based on the shared features, and the multi-classification task is used to perform multi-target state identification based on the shared features.

[0132] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0133] Obtain historical electricity consumption data, which includes total power, power of each load device and working status; determine training samples based on the historical electricity consumption data, use standard back-propagation to optimize the objective function to train the initial load monitoring model, and obtain the load monitoring model. The objective function is determined based on the first objective function corresponding to multi-objective power estimation and the second objective function corresponding to multi-objective state identification.

[0134] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0135] The second working state sequence of each load device is determined according to the power sequence of each load device; if at a certain moment, the first working state of the load device is inconsistent with the second working state corresponding to the load device, the target working state of the load device is determined according to a preset decision method.

[0136] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0137] Obtain the start-up power threshold of each load device, and each start-up power threshold is determined according to the historical power curve of the load device; if the power at a certain moment in the power sequence of the load device is greater than or equal to the start-up power threshold corresponding to the load device, then determine that the second working state of the load device at that moment is the start state.

[0138] In one embodiment, when the processor executes the computer program, the following steps are also implemented: if the power of the load device at a certain moment in the power sequence is less than the start-up power threshold corresponding to the load device, it is determined that the second working state of the load device at that moment is the off state.

[0139] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0140] Obtain a total power sequence of a target building, where the target building includes multiple load devices, and the total power sequence includes the total power of the multiple load devices at different times; input the total power sequence into a load monitoring model to determine the power and working status of each load device at different times; the load monitoring model is used to perform multi-objective power estimation and multi-objective state identification, where the multi-objective power estimation is used to determine the power sequence of each load device, and the multi-objective state identification is used to determine the first working state sequence of each load device.

[0141] In one embodiment, the load monitoring model includes an input layer, a feature extraction layer and an output layer. The feature extraction layer is used to extract multi-task shared features. The output layer includes a multi-target fractional regression task and a multi-classification task. The multi-target fractional regression task is used to perform multi-target power estimation based on the shared features, and the multi-classification task is used to perform multi-target state identification based on the shared features.

[0142] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0143] Historical electricity consumption data, including total power, power of each load device and working status; determine training samples based on historical electricity consumption data, use standard back propagation to optimize the objective function to train the initial load monitoring model, and obtain the load monitoring model. The objective function is determined based on the first objective function corresponding to multi-objective power estimation and the second objective function corresponding to multi-objective state identification.

[0144] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0145] The second working state sequence of each load device is determined according to the power sequence of each load device; if at a certain moment, the first working state of the load device is inconsistent with the second working state corresponding to the load device, the target working state of the load device is determined according to a preset decision method.

[0146] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: determining the second working state sequence of the load devices according to the power sequence of each load device, including: obtaining the start-up power threshold of each load device, each start-up power threshold is determined according to the historical power curve of the load device; if the power at a certain moment in the power sequence of the load device is greater than or equal to the start-up power threshold corresponding to the load device, then determining that the second working state of the load device at that moment is the start-up state.

[0147] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0148] If the power of the load device at a certain moment in the power sequence is less than the start-up power threshold corresponding to the load device, it is determined that the second working state of the load device at this moment is the off state.

[0149] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0150] Obtain a total power sequence of a target building, where the target building includes multiple load devices, and the total power sequence includes the total power of the multiple load devices at different times; input the total power sequence into a load monitoring model to determine the power and working status of each load device at different times; the load monitoring model is used to perform multi-objective power estimation and multi-objective state identification, where the multi-objective power estimation is used to determine the power sequence of each load device, and the multi-objective state identification is used to determine the first working state sequence of each load device.

[0151] In one embodiment, the load monitoring model includes an input layer, a feature extraction layer and an output layer. The feature extraction layer is used to extract multi-task shared features. The output layer includes a multi-target fractional regression task and a multi-classification task. The multi-target fractional regression task is used to perform multi-target power estimation based on the shared features, and the multi-classification task is used to perform multi-target state identification based on the shared features.

[0152] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0153] Historical electricity consumption data, including total power, power of each load device and working status; determine training samples based on historical electricity consumption data, use standard back propagation to optimize the objective function to train the initial load monitoring model, and obtain the load monitoring model. The objective function is determined based on the first objective function corresponding to multi-objective power estimation and the second objective function corresponding to multi-objective state identification.

[0154] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0155] The second working state sequence of each load device is determined according to the power sequence of each load device; if at a certain moment, the first working state of the load device is inconsistent with the second working state corresponding to the load device, the target working state of the load device is determined according to a preset decision method.

[0156] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: determining the second working state sequence of the load devices according to the power sequence of each load device, including: obtaining the start-up power threshold of each load device, each start-up power threshold is determined according to the historical power curve of the load device; if the power at a certain moment in the power sequence of the load device is greater than or equal to the start-up power threshold corresponding to the load device, then determining that the second working state of the load device at that moment is the start-up state.

[0157] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0158] If the power of the load device at a certain moment in the power sequence is less than the start-up power threshold corresponding to the load device, it is determined that the second working state of the load device at this moment is the off state.

[0159] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0160] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0161] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A non-intrusive load monitoring method, characterized in that: The method comprises: Acquire a total power sequence of a target building, wherein the target building includes a plurality of load devices, and the total power sequence includes the total power of the plurality of load devices at different times; The total power sequence is input into the load monitoring model to determine the power and working status of each load device at different times; the load monitoring model is used to perform multi-objective power estimation and multi-objective state identification, the multi-objective power estimation is used to determine the power sequence of each load device, and the multi-objective state identification is used to determine the first working state sequence of each load device.

2. The method according to claim 1, characterized in that The load monitoring model includes an input layer, a feature extraction layer and an output layer. The feature extraction layer is used to extract multi-task shared features. The output layer includes a multi-target fractional regression task and a multi-classification task. The multi-target fractional regression task is used to perform multi-target power estimation based on the shared features, and the multi-classification task is used to perform multi-target state identification based on the shared features.

3. The method according to claim 1, characterized in that The training process of the load monitoring model includes: Acquire historical power consumption data, wherein the historical power consumption data includes total power, power of each load device and working status; The training samples are determined according to the historical electricity consumption data, and the initial load monitoring model is trained using the standard back-propagation optimization objective function to obtain the load monitoring model. The objective function is determined according to the first objective function corresponding to the multi-objective power estimation and the second objective function corresponding to the multi-objective state identification.

4. The method according to claim 1, characterized in that: The method further comprises: Determine a second working state sequence of each of the load devices according to the power sequence of each of the load devices; If at a certain moment, the first working state of the load device is inconsistent with the second working state corresponding to the load device, the target working state of the load device is determined according to a preset decision method.

5. The method according to claim 4, characterized in that The determining the second working state sequence of the load devices according to the power sequence of each of the load devices comprises: Obtaining a start-up power threshold of each of the load devices, wherein each of the start-up power thresholds is determined according to a historical power curve of the load device; If the power at a certain moment in the power sequence of the load device is greater than or equal to the start-up power threshold corresponding to the load device, it is determined that the second working state of the load device at this moment is the start-up state.

6. The method according to claim 5, characterized in that The method further comprises: If the power at a certain moment in the power sequence of the load device is less than the start-up power threshold corresponding to the load device, it is determined that the second working state of the load device at this moment is the off state.

7. A non-intrusive load monitoring device, characterized in that: The device comprises: An acquisition module, configured to acquire a total power sequence of a target building, wherein the target building includes a plurality of load devices, and the total power sequence includes the total power of the plurality of load devices at different times; A determination module is used to input the total power sequence into a load monitoring model to determine the power and working status of each load device at different times; the load monitoring model is used to perform multi-objective power estimation and multi-objective state identification, the multi-objective power estimation is used to determine the power sequence of each load device, and the multi-objective state identification is used to determine the first working state sequence of each load device.

8. 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 6 are implemented.

9. 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 6 are implemented.

10. A computer program product, comprising a computer program, 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 6 are implemented.

Images