Electrical load measurement method and apparatus for electric bicycle, and device and storage medium

Abstract

Disclosed in the present application are an electrical load measurement method and apparatus for an electric bicycle, and a device and a storage medium. The method comprises: collecting original current data, in a power distribution network, of electricity consumers that belong to residential household electricity consumption; according to electricity consumption behavior data, clustering the electricity consumers to obtain user categories to which the electricity consumers belong; on the basis of the user categories, extracting part of the original current data to serve as target current data; loading a charging detection network on the basis of the user categories; encoding the target current data in an encoder to obtain first target current features configured for distinguishing whether there is an electric bicycle charging behavior; decoding the first target current features in a decoder, and on the basis of a decoding result, performing reconstruction to obtain second target current features, wherein the second target current features represent current features of a plurality of levels of when the user categories have the electric bicycle charging behavior; and in a classifier, under the enhancement of the second target current features of the plurality of levels and on the basis of the first target current features, detecting an electrical load for indicating whether the electricity consumers have the electric bicycle charging behavior.

Description

Electric bicycle power load detection method, device, equipment and storage medium

[0001] This application claims priority to the Chinese patent application filed with the China Patent Office on August 17, 2023, with application number 202311034988.9. The entire contents of the above application are incorporated by reference into this application. Technical Field

[0002] The present application relates to the technical field of power grids, for example, to a method, device, equipment and storage medium for detecting the power load of an electric bicycle. Background Art

[0003] An electric bicycle (also known as an electric car) is a vehicle that uses a battery as an auxiliary energy source and is based on an ordinary bicycle. It is equipped with a motor, controller, battery, handlebar, brake handle and other operating components and a display instrument system.

[0004] Electric bicycles generally include motors, batteries, controllers, chargers, instruments, sensors, frames, decorative parts, etc. The downstream end consumers of the electric bicycle industry are mainly individual residents or families.

[0005] With the widespread popularity of electric bicycles, the number of electric bicycles in use is growing, reaching tens of millions. Some users will push the electric bicycles back to their homes for charging or unplug the batteries and carry the batteries back to their homes for charging alone, which will increase the risk of fire in the homes and pose a hidden danger to the safety of users' lives and property.

[0006] Usually, based on deep learning neural networks, non-invasive electrical signal loads can be analyzed to identify whether there are characteristics of electric bicycle charging in the user's home, thereby identifying the behavior of electric bicycle charging in the home.

[0007] However, there are complex power usage signals in residential electricity consumption, which can cause a lot of interference to the charging of electric bicycles, resulting in low accuracy in identifying the charging of electric bicycles.

[0008] Summary of the Invention

[0009] The present application provides a method, device, equipment and storage medium for detecting the electrical load of an electric bicycle, so as to solve the problem of how to improve the accuracy of detecting the charging of an electric bicycle.

[0010] This application provides a method for detecting the electrical load of an electric bicycle, comprising:

[0011] Collecting raw current data from residential electricity users in the distribution network;

[0012] Clustering the electricity users according to their electricity usage behavior data to obtain user categories to which the electricity users belong;

[0013] intercepting a portion of the original current data according to the user category as target current data;

[0014] loading a charging detection network according to the user category, the charging detection network comprising an encoder, a decoder trained for the user category, and a classifier;

[0015] encoding the target current data in the encoder to obtain a first target current feature for distinguishing whether there is an electric bicycle charging behavior;

[0016] The first target current feature is decoded in the decoder, and a second target current feature is reconstructed based on the decoding result, wherein the second target current feature represents a current feature of multiple levels when the user category has an electric bicycle charging behavior;

[0017] In the classifier, under the enhanced effect of the second target current characteristics of multiple levels, it is detected based on the first target current characteristics whether the electricity user has an electricity load for charging electric bicycles.

[0018] The present application also provides an electric bicycle power load detection device, comprising:

[0019] The original current data acquisition module is configured to collect original current data of electricity users belonging to residential electricity consumption in the distribution network;

[0020] An electricity user clustering module is configured to cluster the electricity users according to their electricity usage behavior data to obtain user categories to which the electricity users belong;

[0021] a target current data interception module, configured to intercept part of the original current data according to the user category as target current data;

[0022] a charging detection network loading module configured to load a charging detection network according to the user category, the charging detection network comprising an encoder, a decoder trained for the user category, and a classifier;

[0023] an encoding module configured to encode the target current data in the encoder to obtain a first target current feature configured to distinguish whether there is an electric bicycle charging behavior;

[0024] a decoding module configured to decode the first target current feature in the decoder and reconstruct a second target current feature based on a decoding result, wherein the second target current feature represents a current feature at multiple levels when the user category has an electric bicycle charging behavior;

[0025] The classification module is configured to detect whether the electricity user has an electric bicycle charging load based on the first target current feature in the classifier under the enhanced effect of the second target current feature at multiple levels.

[0026] The present application also provides an electronic device, comprising:

[0027] at least one processor; and

[0028] a memory communicatively connected to the at least one processor; wherein,

[0029] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the electric bicycle power load detection method described in any embodiment of the present application.

[0030] The present application also provides a computer-readable storage medium, which stores a computer program. The computer program is used to enable a processor to implement the electric bicycle power load detection method described in any embodiment of the present application when executed. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] FIG1 is a flow chart of a method for detecting the electrical load of an electric bicycle according to an embodiment of the present application;

[0032] FIG2 is a schematic diagram of the structure of a charging detection network provided according to an embodiment of the present application;

[0033] FIG3 is a flow chart of a method for detecting the electrical load of an electric bicycle according to an embodiment of the present application;

[0034] FIG4 is a schematic structural diagram of an electric bicycle power load detection device provided according to an embodiment of the present application;

[0035] FIG5 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0036] The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in a sequence other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, in addition to the process, method, system, product or equipment comprising a series of steps or units shown in the embodiments of the present application, other processes, methods, systems, products and equipment of this series of steps or units not clearly listed may also be included, or other steps or units inherent to these processes, methods, systems, products or equipment.

[0037] FIG1 is a flow chart of a method for detecting the electrical load of an electric bicycle provided in an embodiment of the present application. The method can be performed by an electric bicycle electrical load detection device, which can be implemented in the form of hardware and / or software and configured in an electronic device. As shown in FIG1 , the method includes steps 101-107.

[0038] In 101 , original current data of residential electricity users in the distribution network are collected.

[0039] In actual applications, the electricity consumption of the distribution network can be classified into residential electricity consumption, general industrial and commercial electricity consumption, large industrial electricity consumption, and agricultural production electricity consumption according to the nature of electricity consumption in each industry.

[0040] Among them, residential electricity consumption includes electricity consumption in urban and rural residential buildings, electricity consumption for public ancillary facilities in urban and rural residential communities, electricity consumption for school teaching and student life, electricity consumption for social welfare facilities, etc.

[0041] In this embodiment, current data of electricity users belonging to residential electricity consumption in the distribution network is collected and recorded as raw current data.

[0042] For the raw current data, preprocessing can be performed, such as removing redundant data, removing outliers, filling missing values, etc., to improve the quality of the raw current data.

[0043] In 102 , electricity users are clustered according to their electricity usage behavior data to obtain user categories to which the electricity users belong.

[0044] In this embodiment, data representing the electricity usage behavior of the current electricity user can be collected and recorded as electricity usage behavior data, such as the time point of the switching event (i.e., the event of turning on or off the electrical equipment), the electricity power, the electricity usage duration, etc. Based on the dimension of the electricity usage behavior data, the current electricity users are clustered to obtain the user category to which the current electricity user belongs.

[0045] In a specific implementation, a clustering algorithm such as K-means clustering (k-means) can be used in advance to cluster electricity users based on the dimension of electricity consumption behavior data to obtain multiple clusters representing user categories, each cluster having a center point.

[0046] When clustering current electricity users, multiple pre-trained clusters representing user categories can be loaded.

[0047] The electricity usage behavior data of current electricity users is collected, and the electricity usage behavior data of current electricity users is divided into first electricity usage behavior data unrelated to electric bicycle charging and second electricity usage behavior data related to electric bicycle charging according to thresholds, types, etc. The divided second electricity usage behavior data related to electric bicycle charging can enhance the proportion of electric bicycle charging characteristics, which is convenient for identifying potential groups charging electric bicycles during clustering.

[0048] The first electricity usage behavior data is converted into a first vector, and the second electricity usage behavior data is converted into a second vector using an encoding algorithm such as one-hot encoding.

[0049] The first vector and the second vector are concatenated into a third vector, and the distance between the third vector and the center point of each cluster is calculated using the Euclidean distance algorithm.

[0050] The multiple distances are compared to determine the cluster corresponding to the smallest distance, the current electricity user is divided into the cluster with the smallest distance, and the user category represented by the cluster is set as the user category to which the current electricity user belongs.

[0051] In step 103 , a portion of the original current data is intercepted according to the user category and used as target current data.

[0052] In this embodiment, the characteristics of electricity consumption behaviors of electricity users of the same user category may be analyzed, and part of the original current data that may be used for charging electric vehicles may be intercepted and used as the target current data.

[0053] Electricity users in different user categories have different corresponding electricity usage behaviors, and accordingly, their electric bicycle charging behaviors are also different. For example, some electricity users are couriers or deliverymen, and ride electric bicycles for a long time. Due to the occupational restrictions of electricity users in this category and the high battery loss, these electricity users often replace batteries and charge the replaced batteries in their spare time, and the remaining battery power may be higher. For example, some electricity users often ride electric bicycles to and from get off work, pick up children, or buy groceries in the morning and evening. The battery loss of electricity users in this category is less, and they usually charge the battery when the remaining power is low, and so on.

[0054] The charging time of electric bicycles (especially batteries) is not consistent in different situations. Therefore, a corresponding window can be set in advance for electricity users in each user category. The length of the window can cover the first inflection point and the second inflection point when electricity users in this user category charge electric bicycles.

[0055] Most electric bicycle batteries are charged in three stages:

[0056] 1. Constant current stage

[0057] Use constant current charging to avoid low voltage and excessive current at the beginning of charging, which may cause battery heating, fire or other irreversible damage.

[0058] The charging mechanism is that the charging current of the charger remains constant, the charged amount increases steadily, and the battery voltage rises gradually.

[0059] 2. Constant pressure stage

[0060] Use constant voltage charging to ensure that the battery eventually reaches the rated voltage.

[0061] The charging mechanism is that the charging voltage of the charger is kept constant and the battery voltage rises slowly until it reaches the rated voltage of the battery.

[0062] 3. Floating charge stage (trickle stage)

[0063] The charging current is small, mainly for battery maintenance.

[0064] The charging mechanism is that the battery is fully charged, the charging voltage of the charger is basically the same as the voltage of the battery, and the charging current is very small, which has a certain effect on maintaining the capacity of the battery.

[0065] The first inflection point is the intersection of the constant current stage and the constant voltage stage, and the second inflection point is the intersection of the constant voltage stage and the trickle current stage. The corresponding window of each user category covers the first inflection point and the second inflection point, which helps to identify the charging waveform.

[0066] Exemplarily, the window and downsampling factor configured for the user category may be loaded to obtain target current data of uniform length, wherein the window and downsampling factor are matching parameters.

[0067] The windows are loaded in sequence in the original current data, and part of the data within the window is intercepted as candidate current data. The candidate current data includes part of the data of the first inflection point and the second inflection point, thereby ensuring the integrity of the charging curve and improving the accuracy of the detection.

[0068] Since the length of the window is not consistent, the length of the candidate current data is also inconsistent. However, the charging detection network based on deep learning requires the input data to have consistent length. Therefore, the candidate current data can be downsampled according to the downsampling factor to obtain target current data of uniform length.

[0069] Exemplarily, downsampling can be expressed as y(n)=x(n*L), where y(n) is the target current data, x is the candidate current data, n is the time domain subscript, and L is the downsampling factor.

[0070] In 104 , a charging detection network is loaded according to the user category.

[0071] In this embodiment, a charging detection network may be constructed and trained in advance for different user categories, and the charging detection network is used to detect whether electricity users in the user category have any behavior of charging electric bicycles.

[0072] The charging detection network is not limited to artificially designed neural networks, but can also be a neural network optimized by a model quantization method, or a neural network searched for the characteristics of electric bicycle charging by a neural network structure search (NAS) method. This embodiment does not impose any restrictions on this.

[0073] The charging detection network includes an encoder, a decoder trained on user categories, and a classifier.

[0074] The encoder is universal for different user categories, the decoder is applicable to a specific user category, and the classifier is universal for different user categories.

[0075] During training, the target current data of electricity users without distinguishing user categories and the current data of electric bicycles during charging separated from the target current data can be used as samples. The samples are labeled with the label of whether the electric bicycle is charging. The encoder, decoder and classifier are trained respectively using the labeled samples. The loss function during training can be relative entropy.

[0076] When the training is completed, the encoder and classifier are maintained without updating parameters. The current data of the electric bicycle when charging in the target current data corresponding to the specified user category is used as a sample, and the sample is labeled with the label whether the electric bicycle is charging. The labeled sample is used to fine-tune the decoder. The loss function during training can be the mean squared error loss (MSE Loss).

[0077] In 105 , the target current data is encoded in an encoder to obtain a first target current feature for distinguishing whether there is an electric bicycle charging behavior.

[0078] In this embodiment, the target current data is input into the encoder, and the encoder can encode the target current data and extract high-level features from the target current data, which are recorded as first target current features.

[0079] The boundaries between different user categories are relatively fuzzy. As old appliances age or are even eliminated, new appliances are purchased, and the weather changes, electricity users' electricity usage behavior will change to a certain extent. If current data labeled with different user categories is used as samples to train the charging detection network, the charging detection network will have poor performance and lack generalization. However, charging an electric bicycle is a long-term operation. The current data within a longer window can more clearly reflect the characteristics of electric bicycle charging and can, to a certain extent, distinguish whether electricity users in this user category charge electric bicycles. Therefore, when training the encoder, the encoder can be trained so that the first target current feature it encodes can be used to distinguish whether electricity users in this user category have electric bicycle charging behavior. The encoder serves as a feature extractor shared by the decoder and classifier, and can provide basic features for the decoder and classifier, avoiding the defect of low accuracy in detecting electric bicycle charging behavior due to fuzzy boundaries between different user categories.

[0080] In one embodiment of the present application, as shown in FIG2 , the encoder includes a long short-term memory (LSTM) network, a first convolutional layer Conv_1, a second convolutional layer Conv_2, a third convolutional layer Conv_3, a first pooling layer Pooling_1, and a second pooling layer Pooling_2.

[0081] For LSTM, the following key variables can be divided into:

[0082] Input variables, which include ht-1 (hidden layer at time t-1) and x t (eigenvector at time t)

[0083] Output variables, which include h t (h t Adding activation functions such as softmax can be used as the real output, otherwise it is used as a hidden layer); and

[0084] Mainline / memory variables, which include c t-1 and c t

[0085] There are three main stages within the long short-term memory network: forgetting stage, selective memory stage and output stage.

[0086] 1. Forgetting stage

[0087] The forgetting stage is mainly to selectively forget the input from the previous node, specifically by calculating the z f (f means forget, where z f =σ(W f *[h t-1 ,x t ]), W f Represents the weight matrix used to calculate the gating signal of the forget gate) as the forget gating to control the c of the previous state t-1 What to keep and what to forget.

[0088] 2. Select the memory stage

[0089] The selective memory stage selectively "memorizes" the input, mainly for multiple input variables x t Select memory, record the important ones and remember less of the unimportant ones. The current input content is z(z=tanh(W*[h t-1 ,x t ]), W represents the weight matrix used to calculate the current input content. The gate signal selected is z i (i stands for information, where z i =σ(W i *[h t-1 ,x t ]), W i represents the weight matrix used to calculate the gating signal of the input gate) for control.

[0090] will forget the z in the stage f with c t-1 The result of the multiplication and the selection memory stage z and z iAdd the multiplication results to get the c transmitted to the next state t .

[0091] 3. Output stage

[0092] The output stage will determine what will be considered as the output of the current state. Mainly through z o (z o =σ(W o *[h t-1 ,x t ]), where W o represents the weight matrix used to calculate the gating signal of the output gate) for control. And also the z obtained in the previous stage o The image is scaled (via a hyperbolic tangent function (tanh) activation function).

[0093] The first convolutional layer Conv_1, the second convolutional layer Conv_2 and the third convolutional layer Conv_3 are all convolutional layers. A convolutional layer is usually composed of several layers of convolutional units. The parameters of each convolutional unit are optimized by the back-propagation algorithm during training. The lower convolutional layers can extract some low-level features, such as edges, lines, and corners. The higher-level convolutional layers can iteratively extract higher-level and more complex features from the lower-level features.

[0094] The first pooling layer, Pooling_1, and the second pooling layer, Pooling_2, are both pooling layers. Pooling layers provide sampling operations and compress input data in a certain way to speed up the computation of the charging detection network. The "something" mentioned here refers to a pooling algorithm, such as max pooling, average pooling, or min pooling.

[0095] In this embodiment, the pooling layer provides a minimum pooling operation, that is, a region is represented by a minimum value of a local region.

[0096] Generally speaking, after preprocessing, the raw current data will not have abnormal values ​​of 0. Therefore, the impact of the minimum pooling operation on the gradient is relatively limited. For electric bicycle charging, other electrical signals are noise, and taking the minimum value of the raw current data helps suppress noise.

[0097] In this embodiment, the target current data is input into the long short-term memory network LSTM, and time series processing is performed on the target current data in the long short-term memory network LSTM to obtain a first reference current feature. The target current data belongs to a time series sequence, and the LSTM can extract features from the time series sequence.

[0098] The first reference current feature is input into the first convolution layer Conv_1, and a convolution operation is performed on the first reference current feature in the first convolution layer Conv_1 to obtain a second reference current feature.

[0099] The second reference current feature is input into the first pooling layer Pooling_1 , and a minimum pooling operation is performed on the second reference current feature in the first pooling layer Pooling_1 to obtain a third reference current feature.

[0100] The third reference current feature is input into the second convolutional layer Conv_2, and a convolution operation is performed on the third reference current feature in the second convolutional layer Conv_2 to obtain a fourth reference current feature.

[0101] The fourth reference current feature is input into the third convolutional layer Conv_3, and a convolution operation is performed on the fourth reference current feature in the third convolutional layer Conv_3 to obtain a fifth reference current feature.

[0102] The fifth reference current feature is input into the second pooling layer Pooling_2, and a minimum pooling operation is performed on the fifth reference current feature in the second pooling layer Pooling_2 to obtain a first target current feature for distinguishing whether there is an electric bicycle charging behavior.

[0103] In 106 , the first target current feature is decoded in a decoder, and a second target current feature is reconstructed based on the decoding result, wherein the second target current feature represents multiple levels of current features when the user category has an electric bicycle charging behavior.

[0104] The first target current feature output by the encoder is processed in two ways. One of the ways is input into the decoder, which decodes the first target current feature of that way. During the decoding process, the decoder sequentially reconstructs multiple levels of current features that represent the electric bicycle charging behavior of electricity users in the user category corresponding to the first target current feature.

[0105] Electricity users of different user categories differ in their choice of electric bicycles, their behavior in using them, and their charging behaviors. Therefore, the current characteristics of electric bicycles charged by different user categories vary. Based on this, a decoder can be trained for each user category, allowing the decoder to reconstruct a second target current characteristic based on the first target current characteristic of the corresponding user category. This reconstruction process further enhances the characteristics of electric bicycle charging in the first target current characteristic and reduces noise.

[0106] Typically, the target current data is dominated by noise, so the encoder typically performs downsampling, which reduces the noise and highlights the characteristics of the e-bike charging signal in terms of timing. The decoder typically performs upsampling, and no residual connection (skip connection) is established between the encoder and decoder.

[0107] In one embodiment of the present application, as shown in FIG2 , the encoder Decoder includes a first deconvolution layer Deconv_1 , a second deconvolution layer Deconv_2 , and a third deconvolution layer Deconv_3 .

[0108] The first deconvolution layer, Deconv_1, the second deconvolution layer, Deconv_2, and the third deconvolution layer, Deconv_3, are all deconvolution layers, also known as transposed convolution. Deconvolution layers provide deconvolution operations, upsampling the feature map of the previous layer by interpolating neighborhood points. The interpolation process uses a convolution kernel to interpolate the neighborhood of each feature point. For effective interpolation, the convolution kernel should be large enough to cover the output.

[0109] The interpolation process is to multiply the value of the convolution kernel by each input and add overlapping responses in the output.

[0110] In this embodiment, the first target current feature is input into the first deconvolution layer Deconv_1, and a deconvolution operation is performed on the first target current feature in the first deconvolution layer Deconv_1 to obtain a first-level second target current feature that characterizes the user category having electric bicycle charging behavior.

[0111] The second target current feature of the first level is input into the second deconvolution layer Deconv_2, and a deconvolution operation is performed on the second target current feature of the first level in the second deconvolution layer Deconv_2 to obtain the second target current feature of the second level when the user category has electric bicycle charging behavior.

[0112] The second target current feature of the second level is input into the third deconvolution layer Deconv_3, and a deconvolution operation is performed on the second target current feature of the second level in the third deconvolution layer Deconv_3 to obtain the second target current feature of the third level that characterizes the user category having electric bicycle charging behavior.

[0113] In 107 , in the classifier, under the enhanced effect of the second target current characteristics of multiple levels, it is detected based on the first target current characteristics whether the electricity user has an electricity load for charging electric bicycles.

[0114] The first target current features output by the encoder are processed in two ways, and the other way is input into the classifier. In addition, the multiple levels of second target current features output by the decoder are also input into the classifier.

[0115] When reconstructing the second target current feature, the decoder strengthens the electric bicycle charging feature in the first target current feature and weakens the noise. Compared with the target current data, the second target current feature has a certain degree of distortion. Therefore, in the classifier, classification is performed based on the first target current feature as the main feature and the second target current features of multiple levels as the auxiliary features. At this time, the second target current features of multiple levels can enhance the first target current feature. The classifier classifies the first target current feature and outputs the detection result of whether the electricity user has the electricity load of electric bicycle charging.

[0116] In one embodiment of the present application, as shown in Figure 2, the classifier includes a first fully connected layer FC_1, a second fully connected layer FC_2, a third fully connected layer FC_3, a fourth fully connected layer FC_4, a fifth fully connected layer FC_5, a first separable convolutional layer SCL_1, a second separable convolutional layer SCL_2 and a third separable convolutional layer SCL_3.

[0117] The first fully connected layer FC_1, the second fully connected layer FC_2, the third fully connected layer FC_3, the fourth fully connected layer FC_4 and the fifth fully connected layer FC_5 are all fully connected layers (FC). The neurons in a fully connected layer are fully connected to the neurons in the previous layer. The fully connected layer can map the learned "distributed feature representation" to a specified multidimensional space, for example, the sample label space.

[0118] The first separable convolution layer SCL_1, the second separable convolution layer SCL_2, and the third separable convolution layer SCL_3 are all separable convolution layers, especially spatial separable convolution layers. Spatial separable convolution is a technology that splits the standard convolution operation into multiple small convolution kernels in the spatial dimension to perform convolution operations.

[0119] In this embodiment, the second target current feature of the first level is input into the first fully connected layer FC_1, and the second target current feature of the first level is mapped to the first candidate current feature in the first fully connected layer FC_1, wherein the dimension of the first candidate current feature is less than or equal to the dimension of the first target current feature.

[0120] The first target current feature and the first candidate current feature are fused into a first fused current feature through functions such as Concat.

[0121] The first fused current feature is input into the first separable convolution layer SCL_1 , and a separable convolution operation is performed on the first fused current feature in the first separable convolution layer SCL_1 to obtain a first separated current feature.

[0122] The second target current feature of the second level is input into the second fully connected layer FC_2, and the second target current feature of the second level is mapped into a second candidate current feature in the second fully connected layer FC_2, wherein the dimension of the second candidate current feature is less than or equal to the dimension of the first separated current feature.

[0123] The first separated current feature and the second candidate current feature are fused into a second fused current feature through a function such as Concat.

[0124] The second fused current feature is input into the second separable convolution layer SCL_2, and a separable convolution operation is performed on the second fused current feature in the second separable convolution layer SCL_2 to obtain a second separated current feature.

[0125] The second target current feature of the third level is input into the third fully connected layer FC_3, and the second target current feature of the third level is mapped into a third candidate current feature in the third fully connected layer FC_3, wherein the dimension of the third candidate current feature is less than or equal to the dimension of the second separated current feature.

[0126] The second separated current feature and the third candidate current feature are fused into a third fused current feature through functions such as Concat.

[0127] The third fused current feature is input into the third separable convolution layer SCL_3, and a separable convolution operation is performed on the third fused current feature in the third separable convolution layer SCL_3 to obtain a third separated current feature.

[0128] The third separated current feature is input into the fourth fully connected layer FC_4, and the third separated current feature is mapped into the fourth separated current feature in the fourth fully connected layer FC_4.

[0129] Input the fourth separated current feature into the fifth fully connected layer FC_5, and map the fourth separated current feature into the probability that the electricity user has an electric bicycle charging behavior in the fifth fully connected layer FC_5;

[0130] If the probability is greater than a preset first threshold, it is determined that the electricity user has an electricity load for charging electric bicycles.

[0131] In this embodiment, original current data of electricity users belonging to residential electricity consumption in the distribution network are collected; electricity users are clustered according to electricity consumption behavior data to obtain the user category to which the electricity users belong; part of the original current data is intercepted according to the user category as target current data; a charging detection network is loaded according to the user category, and the charging detection network includes an encoder, a decoder trained for the user category, and a classifier; the target current data is encoded in the encoder to obtain a first target current feature for distinguishing whether there is electric bicycle charging behavior; the first target current feature is decoded in the decoder, and reconstructed according to the decoding result to obtain a second target current feature, wherein the second target current feature represents multiple levels of current features when the user category has electric bicycle charging behavior; in the classifier, under the enhanced effect of the second target current feature, it is detected based on the first target current feature whether the electricity user has an electric bicycle charging load. The encoder provides a common, basic first target current feature for the decoder and classifier, which can ensure the accuracy of the first target current feature for binary classification and reduce the impact of the blurred boundaries between different user categories on the detection of electric bicycle charging. The decoder reconstructs multiple levels of current features corresponding to different user categories when electric bicycle charging behavior occurs. The second target current features at different levels enhance the first target current feature in turn, enriching the information content of the first target current feature and improving the accuracy of detecting electric bicycle charging. Moreover, the decoder and classifier are part of the entire charging detection network and do not care about the underlying features. The complexity of their structure is relatively small. They can not only distinguish whether electricity users under different user categories have electric bicycle charging behavior, improve electricity safety, but also reduce the amount of detection calculations and occupied resources.

[0132] FIG3 is a flow chart of a method for detecting the electrical load of an electric bicycle provided in an embodiment of the present application. This embodiment adds an alarm operation based on the above embodiment. As shown in FIG3 , the method includes steps 301 to 309 .

[0133] In 301 , original current data of residential electricity users in the distribution network are collected.

[0134] In 302 , electricity users are clustered according to electricity usage behavior data to obtain user categories to which the electricity users belong.

[0135] In 303 , part of the original current data is intercepted according to the user category as target current data.

[0136] In 304 , a charging detection network is loaded according to the user category.

[0137] The charging detection network includes an encoder, a decoder trained on user categories, and a classifier.

[0138] In 305 , the target current data is encoded in an encoder to obtain a first target current feature for distinguishing whether an electric bicycle charging behavior occurs.

[0139] In 306 , the first target current feature is decoded in a decoder, and a second target current feature is reconstructed based on the decoding result, wherein the second target current feature represents multiple levels of current features when the user category has an electric bicycle charging behavior.

[0140] In 307 , in the classifier, under the enhanced effect of the second target current features of multiple levels, it is detected based on the first target current features whether the electricity user has an electricity load for charging electric bicycles.

[0141] In 308 , the proportion of electricity loads of electric bicycles charged by electricity users within a preset time period is counted.

[0142] A single detection of whether an electricity user has an electricity load for charging electric bicycles may have a certain false detection rate. Therefore, in this embodiment, the same electricity user can be detected multiple times within a preset time period to determine whether there is an electricity load for charging electric bicycles. For example, when the electricity user is detected for the first time to have an electricity load for charging electric bicycles, the timing of the start time period (such as 60 minutes) is started. During the timing period, the target current data is intercepted from the original current data at a certain interval (such as 1 minute) to detect whether the electricity user has an electricity load for charging electric bicycles.

[0143] During this time period, the statistical detection result is the proportion of the power load with electric bicycle charging in all detection results.

[0144] In 309 , if the ratio is greater than a preset second threshold, an alarm message is generated for the electricity user.

[0145] The proportion of the power load for electric bicycle charging in all detection results is compared with a preset second threshold value. If the proportion of the power load for electric bicycle charging in all detection results is greater than the preset second threshold value, it means that the probability of the power user having a power load for electric bicycle charging is relatively high. It can be confirmed that the power user has electric bicycle charging behavior, which reduces the false detection rate. At this time, an alarm message can be generated for the power user and handed over to the designated management personnel for investigation.

[0146] FIG4 is a schematic diagram of the structure of an electric bicycle power load detection device provided in an embodiment of the present application. As shown in FIG4 , the device includes:

[0147] The original current data collection module 401 is configured to collect original current data of electricity users belonging to residential electricity consumption in the distribution network;

[0148] The electricity user clustering module 402 is configured to cluster the electricity users according to the electricity usage behavior data to obtain the user category to which the electricity users belong;

[0149] a target current data interception module 403 configured to intercept part of the original current data according to the user category as target current data;

[0150] a charging detection network loading module 404 configured to load a charging detection network according to the user category, the charging detection network including an encoder, a decoder trained for the user category, and a classifier;

[0151] An encoding module 405 is configured to encode the target current data in the encoder to obtain a first target current feature for distinguishing whether the electric bicycle is being charged;

[0152] a decoding module 406 configured to decode the first target current feature in the decoder and reconstruct a second target current feature based on the decoding result, wherein the second target current feature represents a plurality of levels of current features when the user category has an electric bicycle charging behavior; and

[0153] The classification module 407 is configured to detect whether the electricity user has an electric bicycle charging load based on the first target current feature in the classifier under the enhanced effect of the second target current feature at multiple levels.

[0154] In one embodiment of the present application, the electricity user clustering module 402 is further configured to:

[0155] Loading a plurality of pre-trained clusters representing user categories, each of the clusters having a center point;

[0156] collecting first electricity usage behavior data unrelated to electric bicycle charging and second electricity usage behavior data related to electric bicycle charging of the electricity user;

[0157] converting the first electricity usage behavior data into a first vector and converting the second electricity usage behavior data into a second vector;

[0158] concatenate the first vector and the second vector into a third vector;

[0159] calculating the distance between the third vector and each of the center points; and

[0160] The electricity users are divided into the cluster with the smallest distance, and the user category represented by the cluster is set as the user category to which the electricity users belong.

[0161] In one embodiment of the present application, the target current data interception module 403 is further configured to:

[0162] Loading a window and downsampling factor configured for the user category, where the length of the window covers the first inflection point and the second inflection point when the electricity user in the user category charges the electric bicycle, where the first inflection point is the intersection of the constant current stage and the constant voltage stage, and the second inflection point is the intersection of the constant voltage stage and the trickle stage;

[0163] Loading the window into the original current data, and intercepting a portion of the data within the window as candidate current data; and

[0164] A downsampling operation is performed on the candidate current data according to the downsampling factor to obtain target current data of uniform length.

[0165] In one embodiment of the present application, the encoder includes a long short-term memory network, a first convolutional layer, a second convolutional layer, a third convolutional layer, a first pooling layer, and a second pooling layer;

[0166] The encoding module 405 is further configured to:

[0167] performing time series processing on the target current data in the long short-term memory network to obtain a first reference current feature;

[0168] performing a convolution operation on the first reference current feature in the first convolution layer to obtain a second reference current feature;

[0169] performing a minimum pooling operation on the second reference current feature in the first pooling layer to obtain a third reference current feature;

[0170] performing a convolution operation on the third reference current feature in the second convolution layer to obtain a fourth reference current feature;

[0171] performing a convolution operation on the fourth reference current feature in the third convolution layer to obtain a fifth reference current feature; and

[0172] A minimum pooling operation is performed on the fifth reference current feature in the second pooling layer to obtain a first target current feature configured to distinguish whether there is an electric bicycle charging behavior.

[0173] In one embodiment of the present application, the encoder further includes a first deconvolution layer, a second deconvolution layer, and a third deconvolution layer;

[0174] The decoding module 406 is further configured to:

[0175] Performing a deconvolution operation on the first target current feature in the first deconvolution layer to obtain a second target current feature of a first level;

[0176] performing a deconvolution operation on the second target current feature of the first level in the second deconvolution layer to obtain a second target current feature of the second level; and

[0177] A deconvolution operation is performed on the second target current feature of the second level in the third deconvolution layer to obtain the second target current feature of the third level.

[0178] In one embodiment of the present application, the classifier includes a first fully connected layer, a second fully connected layer, a third fully connected layer, a fourth fully connected layer, a fifth fully connected layer, a first separable convolutional layer, a second separable convolutional layer, and a third separable convolutional layer;

[0179] The classification module 407 is further configured to:

[0180] Mapping the second target current feature of the first level to a first candidate current feature in the first fully connected layer;

[0181] fusing the first target current feature and the first candidate current feature into a first fused current feature;

[0182] performing a separable convolution operation on the first fused current feature in the first separable convolution layer to obtain a first separated current feature;

[0183] Mapping the second target current feature of the second level to a second candidate current feature in the second fully connected layer;

[0184] fusing the first separated current feature and the second candidate current feature into a second fused current feature;

[0185] performing a separable convolution operation on the second fused current feature in the second separable convolution layer to obtain a second separated current feature;

[0186] Mapping the second target current feature of the third level to a third candidate current feature in the third fully connected layer;

[0187] fusing the second separated current feature and the third candidate current feature into a third fused current feature;

[0188] performing a separable convolution operation on the third fused current feature in the third separable convolution layer to obtain a third separated current feature;

[0189] Mapping the third separation current feature into a fourth separation current feature in the fourth fully connected layer;

[0190] Mapping the fourth separated current feature into a probability that the electricity user has an electric bicycle charging behavior in the fifth fully connected layer; and

[0191] If the probability is greater than a preset first threshold, it is determined that the electricity user has an electricity load for charging electric bicycles.

[0192] In one embodiment of the present application, it further includes:

[0193] A proportion statistics module is used to count the proportion of electricity loads used by electric bicycles to charge electricity among the electricity users within a preset time period;

[0194] The alarm information generating module is configured to generate an alarm message to the electricity user if the ratio is greater than a preset second threshold.

[0195] The electric bicycle power load detection device provided in the embodiments of the present application can execute the electric bicycle power load detection method provided in any embodiment of the present application, and has the corresponding functional modules and beneficial effects for executing the electric bicycle power load detection method.

[0196] FIG5 shows a block diagram of an electronic device 10 that can be used to implement an embodiment of the present application. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or required herein.

[0197] As shown in FIG5 , the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 and a random access memory (RAM) 13, that is communicatively connected to the at least one processor 11. The memory stores a computer program that can be executed by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from the storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are connected to each other via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0198] Multiple components in the electronic device 10 are connected to the I / O interface 15, including an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a magnetic disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0199] The processor 11 can be a variety of general-purpose and / or specialized processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors that run machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as the electric bicycle power load detection method.

[0200] In some embodiments, the electric bicycle electrical load detection method can be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the electric bicycle electrical load detection method described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to execute the electric bicycle electrical load detection method in any other appropriate manner (for example, by means of firmware).

[0201] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard parts (ASSPs), system on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0202] Computer programs for implementing the methods of the present application may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the computer program is executed by the processor, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The computer program may be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0203] In the context of the present application, a computer-readable storage medium can be a tangible medium that can contain or store a computer program for use by an instruction execution system, device or equipment or used in combination with an instruction execution system, device or equipment. A computer-readable storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium can be a machine-readable signal medium. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0204] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a cathode ray tube (CRT) or a liquid crystal display (LCD) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0205] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0206] A computing system may include a client and a server. The client and server are generally remote from each other and typically interact via a communication network. The client-server relationship arises through computer programs running on the respective computers and establishing a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, a host product within a cloud computing service ecosystem that addresses the management difficulties and limited scalability of traditional physical hosts and virtual private server (VPS) services.

[0207] An embodiment of the present application also provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the electric bicycle power load detection method provided in any embodiment of the present application.

[0208] The computer program product may be implemented in a computer program code that performs the operations of the present application written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0209] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the multiple steps described in this application can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of this application can be achieved. This is not limited herein.

Claims

1. A method for detecting the electrical load of an electric bicycle, comprising: Collect the original current data of electricity users in the distribution network for residential electricity consumption; Clustering the electricity users according to the electricity consumption behavior data of the electricity users to obtain the user categories to which the electricity users belong; intercepting part of the original current data according to the user category as target current data; loading a charging detection network according to the user category, the charging detection network comprising an encoder, a decoder and a classifier trained for the user category; Encoding the target current data in the encoder to obtain a first target current feature for distinguishing whether there is a charging behavior of the electric bicycle; The first target current feature is decoded in the decoder, and a second target current feature is obtained by reconstructing the feature according to the decoding result, wherein the second target current feature represents a current feature of multiple levels when the user category has an electric bicycle charging behavior; as well as In the classifier, under the enhanced effect of the second target current characteristics at multiple levels, it is detected based on the first target current characteristics whether the electricity user has an electricity load for charging electric bicycles.

2. The method according to claim 1, wherein: The electricity users are clustered according to the electricity consumption behavior data to obtain the user categories to which the electricity users belong, including: Loading a plurality of pre-trained clusters representing user categories, each of the clusters having a center point; Collecting first electricity usage behavior data unrelated to electric bicycle charging and second electricity usage behavior data related to electric bicycle charging of the electricity user; respectively converting the first electricity usage behavior data into a first vector and converting the second electricity usage behavior data into a second vector; concatenate the first vector and the second vector into a third vector; calculating the distance between the third vector and each of the center points; and The electricity users are divided into the cluster with the smallest distance, and the user category represented by the cluster is set as the user category to which the electricity users belong.

3. The method according to claim 1, wherein: The intercepting part of the original current data according to the user category as the target current data includes: Loading a window and a downsampling factor configured for the user category, wherein the length of the window covers a first inflection point and a second inflection point when an electricity user in the user category charges an electric bicycle, wherein the first inflection point is an intersection point between a constant current stage and a constant voltage stage, and the second inflection point is an intersection point between a constant voltage stage and a trickle current stage; Loading the window on the original current data, and intercepting a portion of the data within the window as candidate current data; and A downsampling operation is performed on the candidate current data according to the downsampling factor to obtain target current data of uniform length.

4. The method according to any one of claims 1 to 3, wherein: The encoder includes a long short-term memory network, a first convolutional layer, a second convolutional layer, a third convolutional layer, a first pooling layer and a second pooling layer; The encoding of the target current data in the encoder to obtain a first target current feature for distinguishing whether there is an electric bicycle charging behavior includes: performing time series processing on the target current data in the long short-term memory network to obtain a first reference current feature; Performing a convolution operation on the first reference current feature in the first convolution layer to obtain a second reference current feature; Performing a minimum pooling operation on the second reference current feature in the first pooling layer to obtain a third reference current feature; Performing a convolution operation on the third reference current feature in the second convolution layer to obtain a fourth reference current feature; performing a convolution operation on the fourth reference current feature in the third convolution layer to obtain a fifth reference current feature; and A minimum pooling operation is performed on the fifth reference current feature in the second pooling layer to obtain a first target current feature for distinguishing whether there is an electric bicycle charging behavior.

5. The method according to claim 4, wherein: The encoder further comprises a first deconvolution layer, a second deconvolution layer and a third deconvolution layer; The first target current feature is decoded in the decoder, and a second target current feature is reconstructed according to the decoding result, wherein the second target current feature represents a plurality of levels of current features when the user category has an electric bicycle charging behavior, including: Performing a deconvolution operation on the first target current feature in the first deconvolution layer to obtain a second target current feature of a first level; performing a deconvolution operation on the second target current feature of the first level in the second deconvolution layer to obtain a second target current feature of the second level; and A deconvolution operation is performed on the second target current feature of the second level in the third deconvolution layer to obtain the second target current feature of the third level.

6. The method according to claim 5, wherein: The classifier includes a first fully connected layer, a second fully connected layer, a third fully connected layer, a fourth fully connected layer, a fifth fully connected layer, a first separable convolutional layer, a second separable convolutional layer and a third separable convolutional layer; In the classifier, under the enhanced effect of the second target current characteristics at multiple levels, detecting whether the electricity user has an electric bicycle charging power load according to the first target current characteristics includes: Mapping the second target current feature of the first level to a first candidate current feature in the first fully connected layer; fusing the first target current feature and the first candidate current feature into a first fused current feature; Performing a separable convolution operation on the first fused current feature in the first separable convolution layer to obtain a first separated current feature; Mapping the second target current feature of the second level to a second candidate current feature in the second fully connected layer; fusing the first separated current feature and the second candidate current feature into a second fused current feature; performing a separable convolution operation on the second fused current feature in the second separable convolution layer to obtain a second separated current feature; Mapping the second target current feature of the third level to a third candidate current feature in the third fully connected layer; fusing the second separated current feature and the third candidate current feature into a third fused current feature; performing a separable convolution operation on the third fused current feature in the third separable convolution layer to obtain a third separated current feature; Mapping the third separation current feature into a fourth separation current feature in the fourth fully connected layer; Mapping the fourth separated current feature into a probability that the electricity user has an electric bicycle charging behavior in the fifth fully connected layer; and If the probability is greater than a preset first threshold, it is determined that the electricity user has an electricity load for charging electric bicycles.

7. The method according to any one of claims 1 to 3, further comprising: Counting the proportion of the electricity load of the electricity users that are charged with electric bicycles within a preset time period; If the ratio is greater than a preset second threshold, an alarm message is generated for the electricity user.

8. An electric bicycle power load detection device, comprising: The original current data acquisition module is configured to acquire the original current data of the electricity users belonging to the residential electricity users in the distribution network; An electricity user clustering module, configured to cluster the electricity users according to the electricity usage behavior data to obtain the user category to which the electricity users belong; a target current data interception module, configured to intercept part of the original current data according to the user category as target current data; a charging detection network loading module, configured to load a charging detection network according to the user category, the charging detection network comprising an encoder, a decoder and a classifier trained for the user category; an encoding module, configured to encode the target current data in the encoder to obtain a first target current feature for distinguishing whether there is a charging behavior of the electric bicycle; A decoding module, configured to decode the first target current feature in the decoder, and reconstruct the second target current feature according to the decoding result, wherein the second target current feature represents a plurality of levels of current features when the user category has an electric bicycle charging behavior; as well as The classification module is configured to detect whether the electricity user has an electricity load for charging electric bicycles based on the first target current characteristics in the classifier under the enhanced effect of the second target current characteristics at multiple levels.

9. An electronic device, comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the electric bicycle electrical load detection method according to any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, wherein the computer program is configured to implement the electric bicycle power load detection method according to any one of claims 1 to 7 when executed by a processor.

Images