Electric bicycle home-entry charging identification method, medium and terminal

Abstract

The invention is suitable for the technical field of load identification, and relates to an electric bicycle home-entry charging identification method, a medium and a terminal, and the method comprises the steps: S10, carrying out the real-time data collection of a voltage u (t) and a current i (t) of a wire inlet of a main ammeter, and calculating the time sequence of an active power P and a reactive power Q; s20, through an inverted Z-shaped window sliding event detection method, the maximum active power difference value of the left window and the right window is calculated through the sliding window, whether the difference value exceeds a set threshold value or not is judged, and therefore the starting event of the electric equipment is determined; s30, identifying a maximum active power difference value, a maximum reactive power difference value and a ratio of the maximum active power difference value to the maximum reactive power difference value by adopting a characteristic increment threshold value, and if the three values meet conditions at the same time, executing the step S40; and S40, analyzing the output data by using a one-dimensional convolutional neural network model, and identifying a charging starting event of the electric bicycle. According to the invention, the electric bicycle battery charging identification accuracy is improved, and the false alarm rate and the missing report rate are reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of load identification, and in particular relates to a method, medium and terminal for identifying home charging of an electric bicycle. Background Art

[0002] As the concept of green, energy-saving and environmentally friendly life has become popular, electric bicycles have gradually become the main means of transportation for short-distance travel for urban residents due to their convenience, speed, economy and environmental protection. When an electric bicycle starts charging, its starting power and waveform characteristics will be different due to factors such as battery capacity, remaining battery power and charger model. When multiple electrical appliances are used at the same time, the power signal of the electric bicycle starting charging is easily interfered by other electrical appliances. In the case of aliasing, the charging start event cannot be accurately detected, and the electric bicycle battery charging identification effect is poor, with high false alarm and missed alarm rates.

[0003] The patent application with publication number CN117370777A provides a method and system for selecting the identification features of the charging load of an electric bicycle. The method includes collecting data of the electric bicycle when it is charging based on the electric bicycle load data collection platform; analyzing and comparing the charging load features of the electric bicycle based on the collected data; considering the discrimination and redundancy, an improved Fisher scoring method is proposed to measure the feature discrimination and the feature set redundancy is measured based on the maximum information coefficient as feature selection; a greedy algorithm is used to sort all the features in the feature set according to the feature selection target, and the feature subset with the best identification effect is selected. The method in this patent application relies on the two-point difference feature, which is easily affected by noise and other electrical aliasing interference, resulting in inaccurate manual solution of steady-state and transient features. When these features are used as identification conditions, misidentification and missed identification may occur, which has the same disadvantages as the prior art.

[0004] Therefore, how to provide a highly accurate electric bicycle battery charging identification method is a problem that needs to be solved urgently by personnel in the technical field. Summary of the invention

[0005] In view of the shortcomings of the prior art, the purpose of the present invention is to provide an electric bicycle home charging identification method to solve the problems of poor electric bicycle battery charging identification effect, high false alarm rate and missed alarm rate in the prior art; in addition, the present invention also provides an electric bicycle home charging identification medium and terminal.

[0006] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0007] In a first aspect, the present invention provides a method for identifying an electric bicycle for home charging, comprising the following steps:

[0008] S10. Collect the voltage u(t) and current i(t) of the main meter inlet in real time, and calculate the time series of active power P and reactive power Q; S20. Calculate the maximum active power difference between the left and right windows through the inverted Z-shaped window event detection method, and judge whether the difference exceeds the set threshold, so as to determine the start-up event of the electrical equipment; S30. Use the characteristic increment threshold to identify the maximum active power difference, the maximum reactive power difference, and the ratio of the maximum active power difference to the maximum reactive power difference. If all three conditions are met at the same time, execute step S40; S40. Use a one-dimensional convolutional neural network model to analyze the output data and identify the event of starting charging of the electric bicycle.

[0009] Furthermore, in step S10, the calculation formula of active power P is as follows:

[0010]

[0011] Effective voltage U rms and current I rms The calculation formula is as follows:

[0012]

[0013]

[0014] The calculation formulas for apparent power S and reactive power Q are as follows:

[0015] S=U rms ×I rms ;

[0016]

[0017] Furthermore, the specific steps in step S20 are as follows:

[0018] S201, calculation of the power average value of the left window: select the first M period points and calculate the power average value of the left window;

[0019] S202, calculating the power average value of the first window on the right: selecting M periodic points, and calculating the power average value starting from the first window on the right;

[0020] S203, movement of the right window: the right window slides with a step size of M, and the average power values ​​of the second to Nth windows are calculated in sequence;

[0021] S204, maximum active power difference calculation: subtract the average power value of the left window from the average power value of the right window to obtain multiple differences, and take the maximum value as the maximum active power difference;

[0022] S205, threshold determination: determine whether the maximum active power difference is greater than a preset threshold, if so, trigger the first identification filter, otherwise repeat steps S201 to S204.

[0023] Furthermore, in step S30, the characteristic increment threshold is used to identify the maximum active power difference as follows:

[0024] P diff_max >40&&P diff_max <600;

[0025] The characteristic increment threshold is used to identify the maximum reactive power difference as follows:

[0026] Q diff_max >40&&Q diff_max <600;

[0027] The ratio of the maximum active power difference to the maximum reactive power difference is identified using the characteristic increment threshold as follows:

[0028]

[0029] Furthermore, the threshold setting principles include relaxing the floating range and eliminating obvious non-electric bicycle charging events, allowing power changes within a certain range to adapt to the differences in characteristics of different models of electric bicycles and chargers when starting charging, and setting thresholds to filter out other power change events that are not related to electric bicycle charging.

[0030] Furthermore, each input data of the one-dimensional convolutional neural network model in step S40 contains 200 frequency data points, and the input data has two channels, namely active power and reactive power; the category of the output data is a binary classification result: 0 indicates that the electric bicycle is detected to have started charging, and 1 indicates that the electric bicycle is not detected to have started charging, which may be a power change caused by the operation of other electrical appliances.

[0031] Furthermore, the one-dimensional convolutional neural network model structure includes an input layer, a convolutional layer 1, a convolutional layer 2, a convolutional layer 3, a flattening layer, a fully connected layer, a Dropout layer and an output layer; wherein the input layer accepts input data of shape (200,2), that is, each input has 200 time points, and each time point contains 2 features, namely active power and reactive power; convolutional layer 1 has 20 filters, a convolution kernel size of 7, and a stride of 3; convolutional layer 2 and convolutional layer 3 each have 10 filters, a convolution kernel size of 5, and a stride of 2; the fully connected layer has 30 neurons and uses a ReLU activation function; the Dropout layer randomly discards neurons with a probability of 50%; the number of output units in the output layer represents the number of classified categories, and in this task the number of categories is 2, namely electric bicycle start-up charging and non-electric bicycle start-up charging.

[0032] Furthermore, the one-dimensional convolutional neural network model after training is quantized, and the weights and activation values ​​in the model are converted from 32-bit floating point numbers to 8-bit integers. The specific process is as follows:

[0033] Full integer quantization: During the quantization process, all operations of the one-dimensional convolutional neural network model are converted to use integer execution;

[0034] Calibration Dataset: The calibration dataset is used during the quantization process to perform statistical analysis on the weights and activations of the one-dimensional convolutional neural network model;

[0035] TensorFlow Lite model conversion: After quantization is completed, the one-dimensional convolutional neural network model is converted to TensorFlow Lite format.

[0036] In a second aspect, the present invention further provides a computer-readable storage medium, wherein the storage medium stores a computer program, and the computer program implements the above method when executed by a processor.

[0037] In a third aspect, the present invention further provides an electronic terminal, comprising: a processor and a memory; the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the terminal executes the method as described above.

[0038] Compared with the prior art, the electric bicycle home charging identification method, medium and terminal provided by the present invention have at least the following beneficial effects:

[0039] The existing electric bicycle home charging identification method cannot accurately detect the charging start event under aliasing conditions, and the electric bicycle battery charging identification effect is poor, with high false alarm rate and missed alarm rate. The present invention has a simple process and convenient operation. It proposes an inverted Z-shaped sliding window event detection method. By obtaining the maximum difference in the area, it can effectively deal with the secondary power increase and jitter interference when starting charging, improve the accuracy of detection, and reduce the missed alarm rate of the electric bicycle charging start event; in addition, a double filtering system is adopted to further improve the reliability of identification through layered filtering. In the first layer of filtering, the system uses traditional feature increment threshold identification to perform preliminary filtering on signals that obviously do not meet the charging characteristics of electric bicycles, and quickly eliminate irrelevant or misleading inputs. On this basis, the second layer of filtering uses a deep neural network to further analyze the time series that has been preliminarily filtered to more accurately capture the key features of the electric bicycle charging process. This double filtering mechanism not only reduces the interference of noise and redundant information, but also improves the accuracy of the final identification and significantly reduces the false alarm rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the scheme of the present invention, a brief introduction is given below to the figures required for use in the description of the embodiments. Obviously, the figures described below are some embodiments of the present invention. For ordinary technicians in this field, other figures can be obtained based on these figures without paying any creative work.

[0041] Figure 1 A flowchart of a method for identifying home charging of an electric bicycle provided by an embodiment of the present invention;

[0042] Figure 2 A schematic diagram of an inverted Z-shaped sliding window event detection in a method for identifying home charging of an electric bicycle provided by an embodiment of the present invention;

[0043] Figure 3 A schematic diagram of the one-dimensional convolutional neural network model structure of an electric bicycle home charging identification method provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0044] In order to facilitate the understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. The preferred embodiments of the present invention are given in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present invention more thorough and comprehensive.

[0045] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0046] The present invention provides an electric bicycle home charging identification method, which is applied to the electric bicycle battery charging identification process under aliasing conditions. The electric bicycle home charging identification method comprises the following steps:

[0047] S10. Collect the voltage u(t) and current i(t) of the main meter inlet in real time, and calculate the time series of active power P and reactive power Q; S20. Calculate the maximum active power difference between the left and right windows through the inverted Z-shaped window event detection method, and judge whether the difference exceeds the set threshold, so as to determine the start-up event of the electrical equipment; S30. Use the characteristic increment threshold to identify the maximum active power difference, the maximum reactive power difference, and the ratio of the maximum active power difference to the maximum reactive power difference. If all three conditions are met at the same time, execute step S40; S40. Use a one-dimensional convolutional neural network model to analyze the output data and identify the event of starting charging of the electric bicycle.

[0048] The present invention improves the accuracy of identifying the charging of an electric bicycle battery and reduces the false alarm and missed alarm rates.

[0049] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings.

[0050] The present invention provides a method for identifying charging of an electric bicycle at home, which is applied to the process of identifying charging of an electric bicycle battery under aliasing conditions. Figures 1 to 3 In this embodiment, the electric bicycle home charging identification method includes the following steps:

[0051] S10, feature processing: real-time data collection of the voltage u(t) and current i(t) at the main electricity meter inlet, with a sampling frequency of not less than 3.2KHz, and real-time calculation of the time series of active power P and reactive power Q.

[0052] Specifically, in step S10, the calculation formula of active power P is as follows:

[0053]

[0054] Effective voltage U rms and current I rms The calculation formula is as follows:

[0055]

[0056] The calculation formulas for apparent power S and reactive power Q are as follows:

[0057] S=U rms ×I rms ;

[0058]

[0059] S20, event detection: through the inverted Z-shaped window event detection method, the sliding window calculates the maximum active power difference between the left and right windows, and determines whether the difference exceeds the set threshold, thereby determining the start-up event of the electrical equipment.

[0060] Specifically, in this embodiment, the specific steps of step S20 are as follows:

[0061] S201, calculation of the power average value of the left window: select the first M period points and calculate the power average value of the left window.

[0062] Specifically, taking the window size M=5 as an example, P left =(p1+p2+p3+p4+p5)5 / .

[0063] S202, calculation of the power average value of the first window on the right: select M periodic points and calculate the power average value starting from the first window on the right.

[0064] Specifically, P right_1 =(p6+p7+p8+p9+p10)5.

[0065] S203, movement of the right window: the right window slides with a step size of M, and the average power values ​​of the second to Nth windows are calculated in sequence.

[0066] Specifically, the moving step of the right window is M, that is, the sliding window size is M, so as to calculate the average power P of the second window on the right right_2 , the third window average power P right_3 , until the average power P of the Nth window right_n =(P M×N+1 +P M×N+2 +...+P M×N+M )M.

[0067] S204, maximum active power difference calculation: subtract the average power value of the left window from the average power value of the right window to obtain multiple differences, and take the maximum value as the maximum active power difference.

[0068] Specifically, the mean active power P in the right window is used right_1 , P right_2 ,…,P right_n Subtract the average active power of the left window and calculate the maximum active power difference P diff_max .

[0069] S205, threshold judgment: judge whether the maximum active power difference is greater than a preset threshold. In order to avoid underreporting, the threshold is usually set lower. If the maximum difference exceeds the threshold, the first identification filter is triggered, otherwise steps S201 to S204 are repeated.

[0070] Specifically, in this embodiment, the maximum active power difference P is determined diff_max Is it greater than the threshold P? th =40W, where in order to ensure that no events are missed, the threshold P th Generally, it is set too small. If it is greater than the threshold P th , it will enter the first identification filtering, otherwise it will continue to perform event detection.

[0071] S30. Use the characteristic increment threshold to identify the maximum active power difference, the maximum reactive power difference, and the ratio of the maximum active power difference to the maximum reactive power difference. The maximum reactive power difference is solved in the same way as the maximum active power difference in step S1. If all three conditions are met at the same time, execute step S40, that is, enter the second identification and filtering, otherwise return to event detection.

[0072] Specifically, in this embodiment, the following three conditions need to be met at the same time:

[0073] The characteristic increment threshold is used to identify the maximum active power difference as follows:

[0074] P diff_max >40&&P diff_max <600;

[0075] The characteristic increment threshold is used to identify the maximum reactive power difference as follows:

[0076] Q diff_max >40&&Q diff_max <600;

[0077] The ratio of the maximum active power difference to the maximum reactive power difference is identified using the characteristic increment threshold as follows:

[0078]

[0079] In this embodiment, the electric bicycle exhibits obvious capacitive load characteristics when charging, so it will produce a significant increase in reactive power. By considering the change in reactive power and its ratio to active power, the power changes caused by the electric bicycle charging start event and other electrical appliance operations can be more accurately distinguished.

[0080] In this embodiment, the threshold setting principles include:

[0081] Relax the floating range: allow power changes within a certain range to adapt to the differences in characteristics of different models of electric bicycles and chargers when starting charging.

[0082] Eliminate obvious non-electric bicycle charging events: By setting the threshold reasonably, ensure that other power change events not related to electric bicycle charging are filtered out to reduce the false alarm rate.

[0083] S40. Use a one-dimensional convolutional neural network model to deeply analyze and classify the output data in order to more accurately identify the electric bicycle start charging event.

[0084] Specifically, in this embodiment, each input data of the one-dimensional convolutional neural network model in step S40 contains 200 frequency data points, and the input data has two channels, namely active power and reactive power. These two features represent the power change information of the electric bicycle during charging, and can reflect the specific waveform characteristics when charging starts.

[0085] The output data is a binary classification result: 0 indicates that the electric bicycle is detected to be started and charged, and 1 indicates that the electric bicycle is not detected to be started and charged, which may be due to power changes caused by the operation of other electrical appliances.

[0086] Furthermore, in this embodiment, in order to meet the deployment requirements of the edge of the smart energy meter, the network model adopts a one-dimensional convolutional neural network (1D-CNN). 1D-CNN is particularly suitable for processing time series data, has the advantages of short inference time and small amount of calculation, and is very suitable for resource-constrained edge device deployment. The one-dimensional convolutional neural network model structure includes an input layer (InputLayer), a convolutional layer 1 (Conv1D), a convolutional layer 2 (Conv1D), a convolutional layer 3 (Conv1D), a flattening layer (Flatten), a fully connected layer (Dense), a Dropout layer and an output layer (OutputLayer); wherein the input layer accepts input data of shape (200,2), that is, each input has 200 time points, and each time point contains 2 features (active power and reactive power); convolutional layer 1 has 20 filters, a convolution kernel size of 7, and a stride of 3. This layer extracts local features of the input data, and applies the ReLU activation function to introduce nonlinear characteristics, batch normalization (BatchNormalization): standardize the output of the previous layer, accelerate training and increase Strengthen the stability of the model; Convolution layer 2 has 10 filters, a convolution kernel size of 5, and a stride of 2, which further extracts and compresses feature information and applies the ReLU activation function; Convolution layer 3 has 10 filters, a convolution kernel size of 5, and a stride of 2. Through multi-layer convolution, the model can capture more complex patterns in the data; the flattening layer flattens the three-dimensional output of the convolution layer into a one-dimensional vector to facilitate subsequent fully connected layer processing; the fully connected layer has 30 neurons and uses the ReLU activation function. This layer is responsible for further combining the flattened features to form a high-level representation; the Dropout layer randomly discards neurons with a probability of 50% to prevent overfitting of the model and improve generalization ability; the number of output units in the output layer represents the number of classified categories. In this task, the number of categories is 2, namely electric bicycle starting and charging and non-electric bicycle starting and charging.

[0087] Furthermore, in this embodiment, the model structure is as shown in Table 1:

[0088] Table 1

[0089] Layer Input Shape Output shape conv1d(Conv1D) (None,200,1) (None,65,20) batch_normalization(BatchN normalization) (None,65,20) (None,65,20) conv1d_1(Conv1D) (None,65,20) (None,31,10) batch_normalization_1(BatchN normalization) (None,31,10) (None,31,10) conv1d_2(Conv1D) (None,31,10) (None,14,10) batch_normalization_2 (None,14,10) (None,14,10) flatten(Flatten) (None,140) (None,140) dense(Dense) (None,140) (None,30) Dropout (None,30) (None,30) dense_1(Dense) (None,30) (None,2)

[0090] Furthermore, in this embodiment, the one-dimensional convolutional neural network model after training is quantized. The quantization process converts the weights and activation values ​​in the model from 32-bit floating point numbers (float32) to 8-bit integers (int8), thereby greatly reducing the storage space requirements and computing resource consumption of the model. The target device is usually an embedded system or edge device with limited computing power.

[0091] Specifically, in this embodiment, the specific process is as follows:

[0092] Full integer quantization: During the quantization process, all operations of the one-dimensional convolutional neural network model are converted to integer execution, which can significantly reduce the amount of computation and memory usage during model inference.

[0093] Calibration dataset: During the quantization process, a calibration dataset is used to perform statistical analysis on the weights and activations of the one-dimensional convolutional neural network model to ensure that the quantized model is as close as possible to the original floating-point model in terms of accuracy. Usually, this dataset is a small but representative sample dataset.

[0094] TensorFlow Lite model conversion: After quantization is completed, the 1D convolutional neural network model is converted to TensorFlow Lite format to run efficiently on mobile devices, microcontrollers, and other low-power devices.

[0095] In this embodiment, the advantages of the quantitative model are:

[0096] Smaller model size: The storage requirements of the quantized model are greatly reduced, making it more suitable for deployment on resource-constrained devices.

[0097] Faster inference speed: Due to the faster integer operations, the inference speed of the quantized model is significantly improved, meeting the requirements of real-time applications.

[0098] Reduced power consumption: In embedded systems, quantized models can significantly reduce the power consumption of the device, thereby extending the battery life of the device.

[0099] Through post-training quantization processing, the model not only maintains a high level of accuracy, but also runs efficiently in actual deployment, greatly improving the practicality and reliability of the system in various application scenarios.

[0100] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, any one of the methods in the present embodiment is implemented.

[0101] An embodiment of the present invention further provides an electronic terminal, comprising: a processor and a memory; the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the terminal executes any one of the methods in this embodiment.

[0102] The computer-readable storage medium in this embodiment can be understood by ordinary technicians in this field: all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to the computer program. The aforementioned computer program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk and other media that can store program codes.

[0103] The electronic terminal provided in this embodiment includes a processor, a memory, a transceiver and a communication interface. The memory and the communication interface are connected to the processor and the transceiver and complete communication with each other. The memory is used to store computer programs, the communication interface is used to communicate, and the processor and the transceiver are used to run computer programs so that the electronic terminal executes each step of the above method.

[0104] Compared with the prior art, the electric bicycle home charging identification method, medium and terminal described in the above embodiments cannot accurately detect the charging start event under aliasing conditions, and the electric bicycle battery charging identification effect is poor, and the false alarm rate and missed alarm rate are high. The present invention has a simple process and convenient operation. It proposes an inverted Z-shaped sliding window event detection method. By obtaining the maximum difference in the area, it can effectively deal with the secondary power increase and jitter interference when starting charging, improve the accuracy of detection, and reduce the missed alarm rate of the electric bicycle charging start event; in addition, a double filtering system is adopted to further improve the reliability of identification through layered filtering. In the first layer of filtering, the system adopts traditional feature increment threshold identification to perform preliminary filtering on signals that obviously do not meet the charging characteristics of electric bicycles, and quickly eliminate irrelevant or misleading inputs. On this basis, the second layer of filtering uses a deep neural network to further analyze the time series that has been preliminarily filtered to more accurately capture the key features of the electric bicycle charging process. This double filtering mechanism not only reduces the interference of noise and redundant information, but also improves the accuracy of the final identification and significantly reduces the false alarm rate.

[0105] Obviously, the embodiments described above are only preferred embodiments of the present invention, rather than all embodiments. The preferred embodiments of the present invention are shown in the accompanying drawings, but they do not limit the patent scope of the present invention. The present invention can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present invention more thorough and comprehensive. Although the present invention has been described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific embodiments, or to perform equivalent replacements for some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of the present invention, directly or indirectly used in other related technical fields, is also within the scope of patent protection of the present invention.

Claims

1. A method for identifying electric bicycles for home charging, characterized in that: The following steps are involved: S10, collect the voltage u(t) and current i(t) of the main electricity meter inlet in real time, and calculate the time series of active power P and reactive power Q; S20, using an inverted Z-shaped window event detection method, the sliding window calculates the maximum active power difference between the left and right windows, and determines whether the difference exceeds a set threshold, thereby determining a start event of the electrical device; S30, using the characteristic increment threshold to identify the maximum active power difference, the maximum reactive power difference, and the ratio of the maximum active power difference to the maximum reactive power difference. If all three conditions are met at the same time, executing step S40; S40. Use a one-dimensional convolutional neural network model to analyze the output data and identify the event of starting charging of the electric bicycle.

2. A method for identifying home charging of electric bicycles according to claim 1, characterized in that: In step S10, the calculation formula of active power P is as follows: Effective voltage U rms and current I rms The calculation formula is as follows: The calculation formulas for apparent power S and reactive power Q are as follows: S=U rms ×I rms ; 3. The method for identifying home charging of electric bicycles according to claim 1, characterized in that: The specific steps in step S20 are as follows: S201, calculation of the power average value of the left window: select the first M period points and calculate the power average value of the left window; S202, calculating the power average value of the first window on the right: selecting M periodic points, and calculating the power average value starting from the first window on the right; S203, movement of the right window: the right window slides with a step size of M, and the average power values ​​of the second to Nth windows are calculated in sequence; S204, maximum active power difference calculation: subtract the average power value of the left window from the average power value of the right window to obtain multiple differences, and take the maximum value as the maximum active power difference; S205, threshold determination: determine whether the maximum active power difference is greater than a preset threshold, if so, trigger the first identification filter, otherwise repeat steps S201 to S204.

4. The method for identifying home charging of electric bicycles according to claim 1, characterized in that: In step S30, the characteristic increment threshold is used to identify the maximum active power difference as follows: P diff_max >40&&P diff_max <600; The characteristic increment threshold is used to identify the maximum reactive power difference as follows: Q diff_max >40&&Q diff_max <600; The ratio of the maximum active power difference to the maximum reactive power difference is identified using the characteristic increment threshold as follows:

5. A method for identifying home charging of electric bicycles according to claim 4, characterized in that: The threshold setting principles include relaxing the floating range and eliminating obvious non-electric bicycle charging events, allowing power changes within a certain range to adapt to the differences in characteristics of different models of electric bicycles and chargers when charging starts, and setting thresholds to filter out other power change events that are not related to electric bicycle charging.

6. The method for identifying home charging of electric bicycles according to claim 1, characterized in that: Each input data of the one-dimensional convolutional neural network model in step S40 contains 200 frequency data points, and the input data has two channels, namely active power and reactive power; the category of the output data is a binary classification result: 0 indicates that the electric bicycle is detected to have started charging, and 1 indicates that the electric bicycle is not detected to have started charging, which may be a power change caused by the operation of other electrical appliances.

7. A method for identifying home charging of electric bicycles according to claim 6, characterized in that: The one-dimensional convolutional neural network model structure includes an input layer, a convolutional layer 1, a convolutional layer 2, a convolutional layer 3, a flattening layer, a fully connected layer, a Dropout layer and an output layer; wherein the input layer accepts input data of shape (200,2), that is, each input has 200 time points, and each time point contains 2 features, namely active power and reactive power; convolutional layer 1 has 20 filters, a convolution kernel size of 7, and a stride of 3; convolutional layer 2 and convolutional layer 3 each have 10 filters, a convolution kernel size of 5, and a stride of 2; the fully connected layer has 30 neurons and uses a ReLU activation function; the Dropout layer randomly discards neurons with a probability of 50%; the number of output units in the output layer represents the number of classified categories, and in this task the number of categories is 2, namely, electric bicycle start-up charging and non-electric bicycle start-up charging.

8. The method for identifying home charging of electric bicycles according to claim 7, characterized in that: The one-dimensional convolutional neural network model after training is quantized, and the weights and activation values ​​in the model are converted from 32-bit floating point numbers to 8-bit integers. The specific process is as follows: Full integer quantization: During the quantization process, all operations of the one-dimensional convolutional neural network model are converted to use integer execution; Calibration Dataset: The calibration dataset is used during the quantization process to perform statistical analysis on the weights and activations of the one-dimensional convolutional neural network model; TensorFlow Lite model conversion: After quantization is completed, the one-dimensional convolutional neural network model is converted to TensorFlow Lite format.

9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

10. An electronic terminal, characterized in that: include: Processor and memory; The memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the terminal executes the method according to any one of claims 1 to 8.

Images