Power factor adaptive control method for charging system

Abstract

The invention relates to the field of intelligent control, and provides a power factor adaptive control method for a charging system, and the method employs a signal processing and analysis algorithm based on deep learning to carry out the wavelet analysis and time-frequency feature extraction of an input current signal and an input voltage signal. Therefore, the duty ratio of the PFC driving signal is automatically obtained according to the global interactive aggregation representation characteristic between the input current signal and the input voltage signal. Therefore, the dynamic information of the power grid and the load can be more accurately captured so as to quickly respond to the voltage fluctuation and load change of the power grid and reduce the phase difference between the input current and the input voltage, thereby reducing the unstable risk of the charging system caused by the drop of the power factor, and further maintaining the stability of the power grid. Therefore, the automation degree of power factor control is improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent control, and more specifically, to a power factor adaptive control method for a charging system. Background Art

[0002] With the popularity of electric vehicles (EVs), hybrid electric vehicles (HEVs), and portable electronic devices, the demand for efficient and reliable charging systems is growing. These systems need to be able to maintain high efficiency under different load conditions while minimizing the impact on the power grid. Power factor (PF) is an important indicator of the efficiency of effective power transmission in AC circuits. Improving power factor means less energy loss and higher energy utilization.

[0003] However, traditional power factor controllers usually use fixed control strategies, such as fixed duty cycle or time delay. In actual applications, load and grid conditions often change. Specifically, when the load of an electric vehicle charging station suddenly increases or decreases, the fixed control strategy may not be able to quickly adjust to the optimal state, resulting in a decrease in the power factor, which in turn affects the efficiency and stability of the system. In addition, when the grid voltage fluctuates, the traditional power factor controller may not be able to make immediate adjustments due to its preset operating parameters, resulting in an increase in the phase difference between the output current and the input voltage, and a decrease in the power factor. In this case, the system will not only waste more energy, but may also impose additional burdens on the grid, such as increasing harmonic pollution.

[0004] Therefore, an optimized power factor control method for a charging system is desired. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present application provides a power factor adaptive control method for a charging system, which uses a signal processing and analysis algorithm based on deep learning to perform wavelet analysis and time-frequency feature extraction of input current signals and input voltage signals, so as to automatically obtain the duty ratio of the PFC drive signal according to the global interactive aggregation representation characteristics between the input current signal and the input voltage signal. In this way, the dynamic information of the power grid and the load can be captured more accurately, so as to quickly respond to power grid voltage fluctuations and load changes, reduce the phase difference between the input current and the input voltage, thereby reducing the risk of instability of the charging system caused by a decrease in power factor, and then maintain the stability of the power grid, thereby improving the degree of automation of power factor control.

[0006] According to one aspect of the present application, a power factor adaptive control method for a charging system is provided, which includes:

[0007] Acquiring an input current signal collected by a current detection unit;

[0008] Acquire an input voltage signal collected by a voltage detection unit;

[0009] Extracting time-frequency features of the input current signal and the input voltage signal to obtain an input current time-frequency graph and an input voltage time-frequency graph;

[0010] Inputting the input current time-frequency graph into a current time-frequency feature extractor to obtain an input current time-frequency feature vector;

[0011] Inputting the input voltage time-frequency graph into a voltage time-frequency feature extractor to obtain an input voltage time-frequency feature vector;

[0012] The input current time-frequency feature vector and the input voltage time-frequency feature vector are input into a feature decoupling local interaction fine-grained aggregation analysis network to obtain an input current-input voltage decoupling enhanced interaction global representation vector;

[0013] Based on the input current-input voltage decoupling enhancement interaction global representation vector, a duty ratio value of the PFC drive signal is obtained.

[0014] In the above-mentioned power factor adaptive control method for the charging system, time-frequency features of the input current signal and the input voltage signal are extracted to obtain input current time-frequency diagram and input voltage time-frequency diagram, including: wavelet analysis of the input current signal and the input voltage signal to obtain the input current time-frequency diagram and the input voltage time-frequency diagram.

[0015] In the above-mentioned power factor adaptive control method for a charging system, the input current time-frequency diagram is input into a current time-frequency feature extractor to obtain an input current time-frequency feature vector, including: inputting the input current time-frequency diagram into a current time-frequency feature extractor based on a first Vi T model to obtain the input current time-frequency feature vector.

[0016] In the above-mentioned power factor adaptive control method for a charging system, the input voltage time-frequency diagram is input into a voltage time-frequency feature extractor to obtain an input voltage time-frequency feature vector, including: inputting the input voltage time-frequency diagram into a voltage time-frequency feature extractor based on a second Vi T model to obtain the input voltage time-frequency feature vector.

[0017] In the above-mentioned power factor adaptive control method for the charging system, the input current time-frequency feature vector and the input voltage time-frequency feature vector are input into a feature decoupling local interaction fine-grained aggregation analysis network to obtain an input current-input voltage decoupling enhanced interaction global representation vector, including: feature decoupling the input current time-frequency feature vector and the input voltage time-frequency feature vector to obtain a sequence of input current local time-frequency feature vectors and a sequence of input voltage local time-frequency feature vectors; calculating the local semantic association interaction vector between each group of corresponding input current local time-frequency feature vectors and input voltage local time-frequency feature vectors in the sequence of the input current local time-frequency feature vectors and the sequence of the input voltage local time-frequency feature vectors to obtain the input current-input voltage local time-frequency semantic association A sequence of interaction vectors; calculating the fine-grained attention weights of each input current-input voltage local time-frequency semantic association interaction vector in the sequence of input current-input voltage local time-frequency semantic association interaction vectors to obtain a sequence of input current-input voltage local time-frequency fine-grained attention weights; inputting the sequence of input current-input voltage local time-frequency fine-grained attention weights into a random deactivation module to obtain a sequence of input current-input voltage local time-frequency fine-grained attention weights after deactivation; using the sequence of input current-input voltage local time-frequency fine-grained attention weights after deactivation as a weight distribution, calculating the weighted sum of the sequence of input current-input voltage local time-frequency semantic association interaction vectors to obtain the input current-input voltage decoupling reinforcement interaction global representation vector.

[0018] In the above-mentioned power factor adaptive control method for a charging system, the local semantic association interaction vector between each group of corresponding input current local time-frequency feature vectors and input voltage local time-frequency feature vectors in the sequence of the input current local time-frequency feature vectors and the sequence of the input voltage local time-frequency feature vectors is calculated to obtain a sequence of input current-input voltage local time-frequency semantic association interaction vectors, including: respectively calculating the position difference, position dot multiplication and position addition between the input current local time-frequency feature vector and the input voltage local time-frequency feature vector to obtain the input current-input voltage local time-frequency semantic association interaction vector. The input current-input voltage local time-frequency difference vector, the input current-input voltage local time-frequency point product vector and the input current-input voltage local time-frequency sum vector are cascaded and then one-dimensional convolutional encoded to obtain an input current-input voltage local time-frequency multi-dimensional fusion vector; the input current-input voltage local time-frequency multi-dimensional fusion vector is subjected to local window-based maximum pooling processing to obtain the input current-input voltage local time-frequency semantic association interaction vector.

[0019] In the above-mentioned power factor adaptive control method for the charging system, the fine-grained attention weights of each input current-input voltage local time-frequency semantic association interaction vector in the sequence of input current-input voltage local time-frequency semantic association interaction vectors are calculated to obtain a sequence of input current-input voltage local time-frequency fine-grained attention weights, including: multiplying each input current-input voltage local time-frequency semantic association interaction vector in the sequence of input current-input voltage local time-frequency semantic association interaction vectors with a weight matrix, and then adding the obtained vector to the bias vector by position to obtain a sequence of input current-input voltage local time-frequency semantic association interaction modulation vectors; inputting the sequence of input current-input voltage local time-frequency semantic association interaction modulation vectors into a tanh function for activation processing to obtain a sequence of input current-input voltage local time-frequency semantic association interaction activation vectors; multiplying each input current-input voltage local time-frequency semantic association interaction activation vector in the sequence of input current-input voltage local time-frequency semantic association interaction activation vectors with a weight coefficient vector to obtain a sequence of input current-input voltage local time-frequency fine-grained attention weights.

[0020] In the above-mentioned power factor adaptive control method for a charging system, obtaining a duty ratio value of a PFC drive signal based on the input current-input voltage decoupling enhanced interaction global representation vector includes: performing duty ratio decoding processing of the PFC drive signal on the input current-input voltage decoupling enhanced interaction global representation vector to obtain the duty ratio value of the PFC drive signal.

[0021] In the above-mentioned power factor adaptive control method for a charging system, the input current-input voltage decoupling enhanced interaction global representation vector is subjected to a duty cycle decoding process of a PFC drive signal to obtain a duty cycle value of the PFC drive signal, including: inputting the input current-input voltage decoupling enhanced interaction global representation vector into a decoder-based PFC driver to obtain a duty cycle decoding value of the PFC drive signal.

[0022] In the above-mentioned power factor adaptive control method for a charging system, the input current-input voltage decoupling enhanced interaction global representation vector is input into a decoder-based PFC driver to obtain a decoded duty ratio decoding value of a PFC drive signal, including: multiplying a decoding weight matrix of a decoder with the input current-input voltage decoupling enhanced interaction global representation vector to obtain a decoded input current-input voltage decoupling enhanced interaction global representation vector, and accumulating and summing all eigenvalues ​​of the decoded input current-input voltage decoupling enhanced interaction global representation vector to obtain a decoded duty ratio decoding value of the PFC drive signal.

[0023] This application has significant technical effects due to the adoption of the above technical solutions:

[0024] The power factor adaptive control method for a charging system provided in the present application adopts a signal processing and analysis algorithm based on deep learning to perform wavelet analysis and time-frequency feature extraction of input current signals and input voltage signals, thereby automatically obtaining the duty ratio value of the PFC drive signal according to the global interactive aggregation representation characteristics between the input current signal and the input voltage signal. In this way, the dynamic information of the power grid and the load can be captured more accurately, so as to quickly respond to power grid voltage fluctuations and load changes, reduce the phase difference between the input current and the input voltage, thereby reducing the risk of instability of the charging system caused by a decrease in power factor, and then maintaining the stability of the power grid, thereby improving the degree of automation of power factor control. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0026] Figure 1 4 is a flow chart of a power factor adaptive control method for a charging system according to an embodiment of the present application.

[0027] Figure 2 Schematic diagram of data flow of a power factor adaptive control method for a charging system according to an embodiment of the present application.

[0028] Figure 3 In the power factor adaptive control method for a charging system according to an embodiment of the present application, the input current time-frequency feature vector and the input voltage time-frequency feature vector are input into a feature decoupling local interaction fine-grained aggregation analysis network to obtain a flow chart of an input current-input voltage decoupling enhanced interaction global representation vector.

[0029] Figure 4 In the power factor adaptive control method for a charging system according to an embodiment of the present application, a flowchart is provided for calculating the local semantic association interaction vector between each corresponding group of input current local time-frequency feature vectors and input voltage local time-frequency feature vectors in the sequence of the input current local time-frequency feature vectors and the sequence of the input voltage local time-frequency feature vectors to obtain a sequence of input current-input voltage local time-frequency semantic association interaction vectors.

[0030] Figure 5The present invention is a flowchart of calculating the fine-grained attention weights of each input current-input voltage local time-frequency semantic association interaction vector in the sequence of input current-input voltage local time-frequency semantic association interaction vectors in the power factor adaptive control method for a charging system according to an embodiment of the present application to obtain a sequence of input current-input voltage local time-frequency fine-grained attention weights. DETAILED DESCRIPTION

[0031] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0032] In the description of the embodiments of the present application, the term "including" and similar terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "based at least in part on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc. may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0033] It should be noted that the modifications of "one" and "plurality" mentioned in this application are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0034] With the widespread use of electric vehicles (EVs), hybrid electric vehicles (HEVs) and various portable electronic devices, the market demand for efficient and stable charging solutions is increasing. These solutions must maintain high efficiency under diverse load conditions and minimize the impact on the power network. Power factor (PF) is a key parameter for evaluating the effective power transmission efficiency in AC systems. Its optimization can reduce energy loss and improve energy efficiency.

[0035] However, currently common power factor control strategies often adopt static methods, such as setting a fixed duty cycle or time delay. This approach has limitations in practical applications because the load and grid conditions change dynamically. For example, in the case of a sudden increase or decrease in the load at an electric vehicle charging station, the static control strategy may not be able to respond in time, resulting in a decrease in the power factor, affecting the efficiency and stability of the entire system. Similarly, when the grid voltage fluctuates, the traditional power factor controller may not be able to adjust in time because it relies on preset operating parameters, which increases the phase difference between the output current and the input voltage and reduces the power factor. This not only increases energy waste, but may also cause additional pressure on the grid, such as increasing the risk of harmonic pollution.

[0036] Based on this, the present application proposes a power factor adaptive control method for a charging system, which adjusts the working state of the inductor and capacitor by adding a PFC circuit after the rectifier bridge to make the input voltage and input current phase consistent, thereby controlling the power factor of the power supply, so that the power supply device can use power resources more efficiently and reduce the negative impact on the power grid. Specifically, the PFC circuit is composed of a PFC driver, a current detection unit, a voltage detection unit, a comparison unit and a controller.

[0037] In particular, in a PFC circuit, the duty cycle is a key control parameter, which affects the on and off time of the switching device (such as MOSFET or IGBT) in the circuit, and thus affects the working efficiency of the circuit and the effect of power factor correction. Among them, the PFC driver controls the switching element by adjusting the duty cycle of its drive (PWM) signal, so that the circuit can maintain a high power factor under different load conditions. For example, in a boost PFC circuit, the PFC driver adjusts the duty cycle of the PWM signal according to the information of the feedback loop to maintain the stability of the output voltage and ensure that the input current follows the change of the input voltage, so that the input current waveform is close to a sine wave and the current distortion is reduced.

[0038] Accordingly, the technical concept of the present application is to collect input current signals by the current detection unit, and collect input voltage signals by the voltage detection unit, and use a signal processing and analysis algorithm based on deep learning to perform wavelet analysis and time-frequency feature extraction on the input current signal and the input voltage signal, so as to automatically obtain the duty ratio of the PFC drive signal according to the global interaction aggregation representation characteristics between the input current signal and the input voltage signal. In this way, the dynamic information of the power grid and the load can be captured more accurately, so as to quickly respond to power grid voltage fluctuations and load changes, reduce the phase difference between the input current and the input voltage, thereby reducing the risk of instability of the charging system caused by the decrease in power factor, and then maintain the stability of the power grid, thereby improving the automation level of power factor control.

[0039] Figure 1 4 is a flow chart of a power factor adaptive control method for a charging system according to an embodiment of the present application. Figure 2 FIG. 1 is a data flow diagram of a power factor adaptive control method for a charging system according to an embodiment of the present application. Figure 1 and Figure 2As shown, according to the embodiment of the present application, the power factor adaptive control method for the charging system includes: S110, obtaining an input current signal collected by a current detection unit; S120, obtaining an input voltage signal collected by a voltage detection unit; S130, performing time-frequency feature extraction on the input current signal and the input voltage signal to obtain an input current time-frequency diagram and an input voltage time-frequency diagram; S140, inputting the input current time-frequency diagram into a current time-frequency feature extractor to obtain an input current time-frequency feature vector; S150, inputting the input voltage time-frequency diagram into a voltage time-frequency feature extractor to obtain an input voltage time-frequency feature vector; S160, inputting the input current time-frequency feature vector and the input voltage time-frequency feature vector into a feature decoupling local interaction fine-grained aggregation analysis network to obtain an input current-input voltage decoupling enhanced interaction global representation vector; S170, obtaining a duty ratio value of a PFC drive signal based on the input current-input voltage decoupling enhanced interaction global representation vector.

[0040] In step S110 and step S120, the input current signal collected by the current detection unit and the input voltage signal collected by the voltage detection unit are obtained respectively. It should be understood that the input current signal contains the current amplitude, current waveform, phase, and frequency information of the input current of the charging system; the input voltage signal contains the voltage amplitude, voltage waveform, phase, and frequency information of the input voltage of the charging system. Specifically, the input current signal and the input voltage signal contain the key characteristic information of the operation of the charging system. By analyzing these characteristic information, the duty ratio decoding value of the PFC drive signal can be obtained. This decoding value directly determines the working state of the PFC circuit, thereby realizing the dynamic adjustment and optimization of the power factor.

[0041] In step S130, the input current signal and the input voltage signal are subjected to time-frequency feature extraction to obtain an input current time-frequency diagram and an input voltage time-frequency diagram. Specifically, in an embodiment of the present application, the input current signal and the input voltage signal are subjected to time-frequency feature extraction to obtain an input current time-frequency diagram and an input voltage time-frequency diagram, including: performing wavelet analysis on the input current signal and the input voltage signal to obtain the input current time-frequency diagram and the input voltage time-frequency diagram. It should be understood that the input current signal and the input voltage signal have information on time-frequency features, and these feature information are crucial for understanding the dynamic characteristics and change patterns of the signal. Based on this, in the technical solution of the present application, the input current signal and the input voltage signal are subjected to wavelet analysis to obtain an input current time-frequency diagram and an input voltage time-frequency diagram. That is, wavelet analysis is a powerful time-frequency analysis tool that can provide detailed information of the signal in both the time and frequency domains. By performing wavelet analysis on the input current and the input voltage, the time-frequency characteristics of the input current and voltage signals can be more accurately captured and understood, providing support for subsequent data processing.

[0042] In step S140, the input current time-frequency graph is input into a current time-frequency feature extractor to obtain an input current time-frequency feature vector. Specifically, in an embodiment of the present application, the input current time-frequency graph is input into a current time-frequency feature extractor to obtain an input current time-frequency feature vector, including: inputting the input current time-frequency graph into a current time-frequency feature extractor based on a first Vi T model to obtain the input current time-frequency feature vector. It should be understood that the input current time-frequency graph contains a variety of frequency components and time-frequency characteristics, which contain key and important frequency feature information at different time points, and the Vi T model uses a self-attention mechanism, which enables it to identify the most important features in the input data and give them more weight, which is very useful for capturing key time-frequency features in current signals. Therefore, in the technical solution of the present application, the input current time-frequency graph is input into a current time-frequency feature extractor based on the first Vi T model to capture and refine key time-frequency features to obtain an input current time-frequency feature vector.

[0043] In step S150, the input voltage time-frequency graph is input into a voltage time-frequency feature extractor to obtain an input voltage time-frequency feature vector. Specifically, in an embodiment of the present application, the input voltage time-frequency graph is input into a voltage time-frequency feature extractor to obtain an input voltage time-frequency feature vector, including: inputting the input voltage time-frequency graph into a voltage time-frequency feature extractor based on a second Vi T model to obtain the input voltage time-frequency feature vector. It should be understood that, considering that the input voltage time-frequency graph contains the changes of the input voltage signal at different times and frequencies, based on this, in the technical solution of the present application, the input voltage time-frequency graph is input into a voltage time-frequency feature extractor based on a second Vi T model to extract and mine out key time-frequency feature information in the input voltage signal to obtain an input voltage time-frequency feature vector.

[0044] In step S160, the input current time-frequency feature vector and the input voltage time-frequency feature vector are input into a feature decoupled local interaction fine-grained aggregation analysis network to obtain an input current-input voltage decoupled enhanced interaction global representation vector. Accordingly, considering that the mutual interactions and dependencies between the input current time-frequency feature vector and the input voltage time-frequency feature vector in different local time periods are different, and in order to effectively integrate the time-frequency characteristics of current and voltage, extract the mutual relationship and interaction information between them, and obtain a more comprehensive power supply characteristic representation, in the technical solution of the present application, a feature decoupled local interaction fine-grained aggregation analysis network is introduced to perform fine-grained local interaction fusion on the input current time-frequency feature vector and the input voltage time-frequency feature vector to obtain an input current-input voltage decoupled enhanced interaction global representation vector.

[0045] Specifically, the feature decoupled local interaction fine-grained aggregation analysis network first decomposes the current and voltage time-frequency feature information in an entire time period into a sequence of input current local time-frequency feature vectors and a sequence of input voltage local time-frequency feature vectors by feature decoupling the input current time-frequency feature vector and the input voltage time-frequency feature vector, so as to perform a more detailed analysis of the current and voltage time-frequency features in each time period, capture more subtle changes and patterns, and enhance the subsequent dynamic processing capabilities of local feature details. Then, the local semantic association between the sequence of decoupled input current local time-frequency feature vectors and the sequence of input voltage local time-frequency feature vectors is calculated to reveal the intrinsic connection and interdependence between the current and voltage time-frequency features, forming a richer and more meaningful sequence of input current-input voltage local time-frequency semantic association interaction vectors. In particular, the input current-input voltage local time-frequency semantic association interaction vector shows the implicit connection between voltage and current in the local time period, providing a quantitative framework to observe how features influence each other at the detail level, which is extremely critical for identifying specific patterns and organizational structures in data. Subsequently, the fine-grained attention weight of each semantic association interaction vector is calculated, that is, the fine-grained attention weight provides a quantitative measure of the importance of each feature, so as to identify and highlight the most important features in a specific task or decision, thereby improving the quality of feature representation and the prediction accuracy of the model. Subsequently, each fine-grained attention weight is sent to a random deactivation unit to promote the sparsification of the weights and obtain a sequence of input current-input voltage local time-frequency fine-grained attention weights after deactivation. It should be understood that this process effectively reduces the possibility of overfitting by randomly "deactivating" some weights, while ensuring that the information of the key current-voltage exchange feature is maintained. Finally, using the fine-grained attention weight sequence after deactivation as the weight distribution, the sequence of input current-input voltage local time-frequency semantic association interaction vectors is weighted summed to synthesize local interaction information, thereby extracting the most representative feature interaction information and generating a global representation vector of input current-input voltage decoupling reinforcement interaction.

[0046] Specifically, Figure 3 The present invention is a flowchart of a power factor adaptive control method for a charging system according to an embodiment of the present application, wherein the input current time-frequency feature vector and the input voltage time-frequency feature vector are input into a feature decoupling local interaction fine-grained aggregation analysis network to obtain an input current-input voltage decoupling enhanced interaction global representation vector. Figure 3As shown, the input current time-frequency feature vector and the input voltage time-frequency feature vector are input into a feature decoupling local interaction fine-grained aggregation analysis network to obtain an input current-input voltage decoupling enhanced interaction global representation vector, including: S161, feature decoupling the input current time-frequency feature vector and the input voltage time-frequency feature vector to obtain a sequence of input current local time-frequency feature vectors and a sequence of input voltage local time-frequency feature vectors; S162, calculating the local semantic association interaction vector between each group of corresponding input current local time-frequency feature vectors and input voltage local time-frequency feature vectors in the sequence of input current local time-frequency feature vectors and the sequence of input voltage local time-frequency feature vectors to obtain a sequence of input current-input voltage local time-frequency semantic association interaction vectors; S163 , calculate the fine-grained attention weights of each input current-input voltage local time-frequency semantic association interaction vector in the sequence of input current-input voltage local time-frequency semantic association interaction vectors to obtain a sequence of input current-input voltage local time-frequency fine-grained attention weights; S164, input the sequence of input current-input voltage local time-frequency fine-grained attention weights into a random deactivation module to obtain a sequence of input current-input voltage local time-frequency fine-grained attention weights after deactivation processing; S165, using the sequence of input current-input voltage local time-frequency fine-grained attention weights after the deactivation processing as the weight distribution, calculate the weighted sum of the sequence of input current-input voltage local time-frequency semantic association interaction vectors to obtain the input current-input voltage decoupling reinforcement interaction global representation vector.

[0047] More specifically, Figure 4 The present invention is a flowchart of calculating the local semantic association interaction vector between each corresponding input current local time-frequency feature vector and input voltage local time-frequency feature vector in the sequence of the input current local time-frequency feature vector and the sequence of the input voltage local time-frequency feature vector to obtain a sequence of input current-input voltage local time-frequency semantic association interaction vectors in the power factor adaptive control method for the charging system according to the embodiment of the present application. Figure 4As shown, calculating the local semantic association interaction vector between each group of corresponding input current local time-frequency feature vectors and input voltage local time-frequency feature vectors in the sequence of the input current local time-frequency feature vectors and the sequence of the input voltage local time-frequency feature vectors to obtain a sequence of input current-input voltage local time-frequency semantic association interaction vectors, including: S1621, respectively calculating the position difference, position dot multiplication and position addition between the input current local time-frequency feature vector and the input voltage local time-frequency feature vector to obtain the input current-input voltage local time-frequency difference vector, the input current-input voltage local time-frequency difference vector, the input current-input voltage local time-frequency difference vector, the input current-input voltage local time-frequency difference vector, the input current-input voltage local time-frequency difference vector, the input current-input voltage local time-frequency difference vector, the input voltage-input voltage local time-frequency semantic association interaction vector, and the input voltage local time-frequency difference vector. The voltage local time-frequency point product vector and the input current-input voltage local time-frequency sum vector; S1622, cascading the input current-input voltage local time-frequency difference vector, the input current-input voltage local time-frequency point product vector and the input current-input voltage local time-frequency sum vector, and performing one-dimensional convolution encoding to obtain an input current-input voltage local time-frequency multi-dimensional fusion vector; S1623, performing local window-based maximum pooling processing on the input current-input voltage local time-frequency multi-dimensional fusion vector to obtain the input current-input voltage local time-frequency semantic association interaction vector.

[0048] More specifically, Figure 5 The present invention is a flowchart of calculating the fine-grained attention weights of each input current-input voltage local time-frequency semantic association interaction vector in the sequence of input current-input voltage local time-frequency semantic association interaction vectors in the power factor adaptive control method for a charging system according to an embodiment of the present application to obtain a sequence of input current-input voltage local time-frequency fine-grained attention weights. Figure 5 As shown, the fine-grained attention weights of each input current-input voltage local time-frequency semantic association interaction vector in the sequence of input current-input voltage local time-frequency semantic association interaction vectors are calculated to obtain a sequence of input current-input voltage local time-frequency fine-grained attention weights, including: S1631, after multiplying each input current-input voltage local time-frequency semantic association interaction vector in the sequence of input current-input voltage local time-frequency semantic association interaction vectors with a weight matrix, the obtained vector is added to the bias vector by position to obtain a sequence of input current-input voltage local time-frequency semantic association interaction modulation vectors; S1632, the sequence of input current-input voltage local time-frequency semantic association interaction modulation vectors is input into a tanh function for activation processing to obtain a sequence of input current-input voltage local time-frequency semantic association interaction activation vectors; S1633, each input current-input voltage local time-frequency semantic association interaction activation vector in the sequence of input current-input voltage local time-frequency semantic association interaction activation vectors is multiplied by a weight coefficient vector to obtain a sequence of input current-input voltage local time-frequency fine-grained attention weights.

[0049] In the embodiment of the present application, specifically, the input current time-frequency feature vector and the input voltage time-frequency feature vector are input into a feature decoupling local interaction fine-grained aggregation analysis network to obtain an input current-input voltage decoupling enhanced interaction global representation vector, including: the input current time-frequency feature vector and the input voltage time-frequency feature vector are input into a feature decoupling local interaction fine-grained aggregation analysis network, and processed with the following local interaction fine-grained aggregation to obtain the input current-input voltage decoupling enhanced interaction global representation vector; wherein the local interaction fine-grained aggregation formula is:

[0050] Decouple(f1)={f 11 ,f 12 ,...,f 1i ,...,f 1n}

[0051] Decouple(f2)={f 21 ,f 22 ,...,f 2i ,...,f 2n}

[0052]

[0053] E i =v T tanh(W r f int (f 1i ,f 2i )+b r )

[0054] W i = dropout(E i )

[0055] V int =∑ i f int (f 1i ,f 2i )·W i

[0056] Wherein, f1 and f2 are the input current time-frequency characteristic vector and the input voltage time-frequency characteristic vector, respectively, Decouple(·) is the characteristic decoupling operation, and f 1n and f 2nare the nth input current local time-frequency feature vector and the nth input voltage local time-frequency feature vector in the sequence of the input current local time-frequency feature vector and the sequence of the input voltage local time-frequency feature vector, respectively, n represents the number of feature vectors in the sequence of the input current local time-frequency feature vector and the sequence of the input voltage local time-frequency feature vector, f 1i and f 2i are respectively the i-th local time-frequency characteristic vector of the input current and the i-th local time-frequency characteristic vector of the input voltage, ⊙ and are position difference, position dot multiplication and position addition respectively, [·,·,·] is cascade processing, conv1D(·) is a one-dimensional convolutional encoding operation, MaxPool(·) is a maximum pooling operation, and f int (f 1i ,f 2i ) is the i-th input current-input voltage local time-frequency semantic association interaction vector in the sequence of the input current-input voltage local time-frequency semantic association interaction vector, W r is the weight matrix, b r is the bias vector, tanh(·) is the tanh function, v T is the weight coefficient vector, E i is the i-th input current-input voltage local time-frequency fine-grained attention weight in the sequence of the input current-input voltage local time-frequency fine-grained attention weights, dropout(·) represents the deactivation process, W i is the i-th input current-input voltage local time-frequency fine-grained attention weight after deactivation in the sequence of the input current-input voltage local time-frequency fine-grained attention weight after deactivation, V int is the global representation vector of the input current-input voltage decoupling enhancement interaction.

[0057] In step S170, based on the input current-input voltage decoupling enhanced interaction global representation vector, the duty ratio value of the PFC drive signal is obtained. Specifically, in an embodiment of the present application, based on the input current-input voltage decoupling enhanced interaction global representation vector, the duty ratio value of the PFC drive signal is obtained, including: performing a PFC drive signal duty ratio decoding process on the input current-input voltage decoupling enhanced interaction global representation vector to obtain the duty ratio value of the PFC drive signal. In an example of the present application, performing a PFC drive signal duty ratio decoding process on the input current-input voltage decoupling enhanced interaction global representation vector to obtain the duty ratio value of the PFC drive signal includes: inputting the input current-input voltage decoupling enhanced interaction global representation vector into a decoder-based PFC driver to obtain a PFC drive signal duty ratio decoding value. That is, the input current-input voltage decoupling enhanced interaction global representation vector obtained by fine-grained local interaction using the input current time-frequency feature vector and the input voltage time-frequency feature vector is decoded, so as to automatically obtain the duty ratio value of the PFC drive signal. It should be understood that the duty cycle decoding value of the PFC drive signal refers to the ratio of the conduction time of the switch element (such as MOSFET) in the PFC circuit in one cycle to the entire cycle time. This ratio is a key parameter for adjusting the power factor and directly affects the working efficiency of the power supply and the quality of the current waveform. By dynamically adjusting the duty cycle, the PFC driver can improve the waveform of the input current, making it closer to the waveform of the input voltage, reducing the phase difference, thereby improving the power factor of the entire power system, achieving energy saving and reducing the impact of electrical equipment on the power grid, and thus providing the stability and reliability of the power grid. In this way, the dynamic information of the power grid and the load can be captured more accurately, so as to quickly respond to grid voltage fluctuations and load changes, reduce the phase difference between the output current and the input voltage, thereby reducing the risk of instability of the charging system due to a decrease in the power factor, and then maintain the stability of the power grid, thereby improving the automation level of power factor control.

[0058] In a preferred example of the present application, the input current-input voltage decoupling enhanced interaction global representation vector is input into a decoder-based PFC driver to obtain a decoded duty ratio decoded value of the PFC drive signal, including: multiplying a decoding weight matrix of the decoder with the input current-input voltage decoupling enhanced interaction global representation vector to obtain a decoded input current-input voltage decoupling enhanced interaction global representation vector, and accumulating and summing all eigenvalues ​​of the decoded input current-input voltage decoupling enhanced interaction global representation vector to obtain a decoded duty ratio decoded value of the PFC drive signal.

[0059] Specifically, in the technical solution of the present application, the decoder is used to decode and regress the input current-input voltage decoupling enhanced interaction global representation vector using the following formula to obtain the duty ratio decoding value of the PFC drive signal, wherein the formula is Wherein, X is the global representation vector of the input current-input voltage decoupling reinforcement interaction, Y is the decoded value of the duty cycle of the PFC drive signal, and W is the decoding weight matrix of the decoder. Represents matrix multiplication.

[0060] Preferably, considering that the input current time-frequency feature vector and the input voltage time-frequency feature vector respectively represent the signal waveforms of the input current signal and the input voltage signal based on the global image semantic space domain association features of the local image semantic space domain distribution context, when performing feature decoupling local interaction fine-grained aggregation analysis on them, due to the feature decoupling local interaction differences caused by the differences in the local-global distribution correspondence of the image semantic features, analysis semantic aggregation fine-grained alignment conflicts will occur, thereby causing the loss of aggregation key information and affecting the expression effect of the input current-input voltage decoupling enhanced interaction global representation vector.

[0061] Based on this, in the preferred example, when the input current-input voltage decoupling enhancement interaction global representation vector is input into the decoder-based PFC driver to obtain the duty cycle decoding value of the PFC drive signal, the input current-input voltage decoupling enhancement interaction global representation vector is optimized. Specifically, the optimization process includes the following steps: clustering all eigenvalues ​​of the input current-input voltage decoupling enhancement interaction global representation vector based on the distance between eigenvalues, such as the difference absolute value distance, and determining the ratio of the number of features of the clustering features to the number of features of the input current-input voltage decoupling enhancement interaction global representation vector; arranging the clustering features into an input current-input voltage decoupling enhancement interaction global clustering vector, and dividing the second norm of the input current-input voltage decoupling enhancement interaction global clustering vector by the second norm of the input current-input voltage decoupling enhancement interaction global representation vector to obtain a first input current-input voltage decoupling enhancement interaction global clustering related weight value; dividing the first norm of the input current-input voltage decoupling enhancement interaction global clustering vector by a first power value exponential to the ratio value A second power value of the ratio value as an exponent of a norm of the input current-input voltage decoupling reinforcement interaction global representation vector is used to obtain a second input current-input voltage decoupling reinforcement interaction global clustering related weight value; for each eigenvalue of the input current-input voltage decoupling reinforcement interaction global clustering vector, multiply it by the inverse of the difference between the first and second input current-input voltage decoupling reinforcement interaction global clustering related weight values ​​to obtain the optimized eigenvalue of the input current-input voltage decoupling reinforcement interaction global clustering vector; for each eigenvalue outside the cluster in the input current-input voltage decoupling reinforcement interaction global representation vector, multiply it by the inverse of the sum of the first and second input current-input voltage decoupling reinforcement interaction global clustering related weight values ​​to obtain the optimized out-of-class eigenvalue of the input current-input voltage decoupling reinforcement interaction global representation vector; the optimized eigenvalue of the input current-input voltage decoupling reinforcement interaction global clustering vector and the optimized out-of-class eigenvalue of the input current-input voltage decoupling reinforcement interaction global representation vector are combined to form an optimized input current-input voltage decoupling reinforcement interaction global representation vector.

[0062] That is, the optimization process is expressed as:

[0063]

[0064] where f is each eigenvalue of the global clustering vector of the input current-input voltage decoupling reinforcement interaction, V n is the global representation vector of the input current-input voltage decoupling reinforcement interaction, is the input current-input voltage decoupling reinforcement interaction global clustering vector, n is the feature number of the input current-input voltage decoupling reinforcement interaction global representation vector, k is the feature number of the input current-input voltage decoupling reinforcement interaction global clustering vector, λ is the ratio of the feature number of the input current-input voltage decoupling reinforcement interaction global clustering vector to the feature number of the input current-input voltage decoupling reinforcement interaction global representation vector, represents the clustering feature set corresponding to the global clustering vector of the input current-input voltage decoupling enhancement interaction, ||·||2 and represent the λth power of the second norm and the first norm of the vector respectively, and f' is each eigenvalue of the global representation vector of the optimized input current-input voltage decoupling enhancement interaction.

[0065] Here, in order to avoid the loss of key suffix semantic information relative to the original feature set as a whole due to aggregation conflict caused by the input current-input voltage decoupling and enhanced interaction global representation vector being based on aggregated features, the clustering ratio of the number of features of the input current-input voltage decoupling and enhanced interaction global clustering vector relative to the number of features of the input current-input voltage decoupling and enhanced interaction global representation vector is used as a decision function to perform an adversarial decision on the set absolute representation of the one-norm of the input current-input voltage decoupling and enhanced interaction global clustering vector and the input current-input voltage decoupling and enhanced interaction global representation vector, and the clustering ratio of the number of features of the input current-input voltage decoupling and enhanced interaction global clustering vector and the input current are respectively compared with the input current-input voltage decoupling and enhanced interaction global clustering vector and the input current -The clustering intrinsic conflict representation of the two norms of the input voltage decoupling enhanced interaction global representation vector is positively and negatively interacted to construct the optimized input current-input voltage decoupling enhanced interaction global representation vector based on the aggregated features and the solid alignment guardrail of the original feature set as a whole, so as to achieve the mitigation of the harmful information loss intention of the optimized input current-input voltage decoupling enhanced interaction global representation vector based on the aggregated risk transferability, and improve the expression effect of the optimized input current-input voltage decoupling enhanced interaction global representation vector, so as to improve the accuracy of the duty ratio decoding value of the PFC drive signal obtained by the input current-input voltage decoupling enhanced interaction global representation vector input to the decoder-based PFC driver. In this way, the dynamic information of the power grid and the load can be captured more accurately, so as to quickly respond to the voltage fluctuation and load change of the power grid, reduce the phase difference between the output current and the input voltage, thereby reducing the instability risk of the charging system caused by the decrease in power factor, and then maintain the stability of the power grid, so as to improve the automation level of power factor control.

[0066] In summary, the power factor adaptive control method for the charging system based on the embodiment of the present application is explained, which uses a signal processing and analysis algorithm based on deep learning to perform wavelet analysis and time-frequency feature extraction of the input current signal and the input voltage signal, so as to automatically obtain the duty ratio of the PFC drive signal according to the global interactive aggregation representation characteristics between the input current signal and the input voltage signal. In this way, the dynamic information of the power grid and the load can be captured more accurately, so as to quickly respond to the voltage fluctuations and load changes of the power grid, reduce the phase difference between the input current and the input voltage, thereby reducing the risk of instability of the charging system caused by the decrease in power factor, and then maintain the stability of the power grid, thereby improving the degree of automation of power factor control.

[0067] The above is only a preferred embodiment of the present application and does not constitute any form of limitation to the present application. Although the present application has been disclosed as a preferred embodiment as above, it is not intended to limit the present application. Any technician familiar with the profession can make some changes or modifications to equivalent embodiments of the technical contents disclosed above without departing from the scope of the technical solution of the present application. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present application without departing from the content of the technical solution of the present application still falls within the scope of the technical solution of the present application.

Claims

1. A power factor adaptive control method for a charging system, characterized in that: include: Acquiring an input current signal collected by a current detection unit; Acquire an input voltage signal collected by a voltage detection unit; Extracting time-frequency features of the input current signal and the input voltage signal to obtain an input current time-frequency graph and an input voltage time-frequency graph; Inputting the input current time-frequency graph into a current time-frequency feature extractor to obtain an input current time-frequency feature vector; Inputting the input voltage time-frequency graph into a voltage time-frequency feature extractor to obtain an input voltage time-frequency feature vector; The input current time-frequency feature vector and the input voltage time-frequency feature vector are input into a feature decoupling local interaction fine-grained aggregation analysis network to obtain an input current-input voltage decoupling enhanced interaction global representation vector; Based on the input current-input voltage decoupling enhancement interaction global representation vector, a duty ratio value of the PFC drive signal is obtained.

2. The power factor adaptive control method for a charging system according to claim 1, characterized in that: Extracting time-frequency features of the input current signal and the input voltage signal to obtain an input current time-frequency diagram and an input voltage time-frequency diagram, including: performing wavelet analysis on the input current signal and the input voltage signal to obtain the input current time-frequency diagram and the input voltage time-frequency diagram.

3. The power factor adaptive control method for a charging system according to claim 2, characterized in that: Inputting the input current time-frequency diagram into a current time-frequency feature extractor to obtain an input current time-frequency feature vector includes: inputting the input current time-frequency diagram into a current time-frequency feature extractor based on a first ViT model to obtain the input current time-frequency feature vector.

4. The power factor adaptive control method for a charging system according to claim 3, characterized in that: Inputting the input voltage time-frequency graph into a voltage time-frequency feature extractor to obtain an input voltage time-frequency feature vector includes: inputting the input voltage time-frequency graph into a voltage time-frequency feature extractor based on a second ViT model to obtain the input voltage time-frequency feature vector.

5. The power factor adaptive control method for a charging system according to claim 4, characterized in that: The input current time-frequency feature vector and the input voltage time-frequency feature vector are input into a feature decoupling local interaction fine-grained aggregation analysis network to obtain an input current-input voltage decoupling enhanced interaction global representation vector, including: Performing feature decoupling on the input current time-frequency feature vector and the input voltage time-frequency feature vector to obtain a sequence of input current local time-frequency feature vectors and a sequence of input voltage local time-frequency feature vectors; Calculating the local semantic association interaction vector between each corresponding group of input current local time-frequency feature vectors and input voltage local time-frequency feature vectors in the sequence of the input current local time-frequency feature vectors and the sequence of the input voltage local time-frequency feature vectors to obtain a sequence of input current-input voltage local time-frequency semantic association interaction vectors; Calculating the fine-grained attention weight of each input current-input voltage local time-frequency semantic association interaction vector in the sequence of input current-input voltage local time-frequency semantic association interaction vectors to obtain a sequence of input current-input voltage local time-frequency fine-grained attention weights; Inputting the sequence of input current-input voltage local time-frequency fine-grained attention weights into a random dropout module to obtain a sequence of input current-input voltage local time-frequency fine-grained attention weights after dropout processing; The sequence of local time-frequency fine-grained attention weights of input current-input voltage after the deactivation process is used as the weight distribution, and the weighted sum of the sequence of local time-frequency semantic association interaction vectors of input current-input voltage is calculated to obtain the global representation vector of the input current-input voltage decoupling reinforcement interaction.

6. The power factor adaptive control method for a charging system according to claim 5, characterized in that: Calculating the local semantic association interaction vector between each corresponding group of input current local time-frequency feature vectors and input voltage local time-frequency feature vectors in the sequence of the input current local time-frequency feature vectors and the sequence of the input voltage local time-frequency feature vectors to obtain a sequence of input current-input voltage local time-frequency semantic association interaction vectors, including: Respectively calculating the position difference, position point multiplication and position addition between the input current local time-frequency characteristic vector and the input voltage local time-frequency characteristic vector to obtain an input current-input voltage local time-frequency difference vector, an input current-input voltage local time-frequency point multiplication vector and an input current-input voltage local time-frequency sum vector; The input current-input voltage local time-frequency difference vector, the input current-input voltage local time-frequency dot product vector and the input current-input voltage local time-frequency sum vector are cascaded and then one-dimensionally convolutionally encoded to obtain an input current-input voltage local time-frequency multi-dimensional fusion vector; The input current-input voltage local time-frequency multi-dimensional fusion vector is subjected to a local window-based maximum pooling process to obtain the input current-input voltage local time-frequency semantic association interaction vector.

7. The power factor adaptive control method for a charging system according to claim 6, characterized in that: Calculating the fine-grained attention weight of each input current-input voltage local time-frequency semantic association interaction vector in the sequence of the input current-input voltage local time-frequency semantic association interaction vector to obtain a sequence of input current-input voltage local time-frequency fine-grained attention weights, including: After multiplying each input current-input voltage local time-frequency semantic association interaction vector in the sequence of the input current-input voltage local time-frequency semantic association interaction vectors by a weight matrix, the obtained vector is added to the bias vector by position to obtain a sequence of input current-input voltage local time-frequency semantic association interaction modulation vectors; Inputting the sequence of input current-input voltage local time-frequency semantic association interaction modulation vectors into a tanh function for activation processing to obtain a sequence of input current-input voltage local time-frequency semantic association interaction activation vectors; Each input current-input voltage local time-frequency semantic association interaction activation vector in the sequence of the input current-input voltage local time-frequency semantic association interaction activation vectors is multiplied by a weight coefficient vector to obtain a sequence of the input current-input voltage local time-frequency fine-grained attention weights.

8. The power factor adaptive control method for a charging system according to claim 7 is characterized in that: Based on the input current-input voltage decoupling enhanced interaction global representation vector, obtaining a duty ratio value of a PFC drive signal comprises: performing a duty ratio decoding process of the PFC drive signal on the input current-input voltage decoupling enhanced interaction global representation vector to obtain the duty ratio value of the PFC drive signal.

9. The power factor adaptive control method for a charging system according to claim 8 is characterized in that , performing a PFC driving signal duty decoding process on the input current-input voltage decoupling enhanced interaction global representation vector to obtain a duty ratio value of the PFC driving signal, including: inputting the input current-input voltage decoupling enhanced interaction global representation vector into a decoder-based PFC driver to obtain a PFC driving signal duty ratio decoding value.

10. The power factor adaptive control method for a charging system according to claim 9 is characterized in that: Inputting the input current-input voltage decoupling enhanced interaction global representation vector into a decoder-based PFC driver to obtain a decoded duty ratio of a PFC drive signal, comprising: multiplying a decoding weight matrix of a decoder with the input current-input voltage decoupling enhanced interaction global representation vector to obtain a decoded input current-input voltage decoupling enhanced interaction global representation vector, and accumulating and summing all eigenvalues ​​of the decoded input current-input voltage decoupling enhanced interaction global representation vector to obtain a decoded duty ratio of the PFC drive signal.

Images