AC and DC Multi-Channel Power Supply Controllers

Abstract

The present application discloses an AC and DC multi-channel power supply controller, which combines a deep learning algorithm based on artificial intelligence to locally extract and locally enhance the time domain fluctuation correlation features of the time series data of the equipment load, and introduces a feature distribution reverse mask clustering analysis mechanism to aggregate the implicit feature patterns of the equipment load timing correlation in each local time domain, thereby simulating and characterizing the dynamic load characteristics of the entire equipment, and based on this, automatically generates control instructions to guide the switching of DC power supply.

Description

Technical Field

[0001] The present application relates to the field of intelligent control, and more specifically, to an AC and DC multi-channel power supply controller. Background Art

[0002] In systems that supply power to AC loads, AC power is usually used as the power source for the AC loads. In some application scenarios, DC power can be added as a backup power source. However, configuring a DC power source in the system requires adding more supporting equipment, which increases the complexity of the system, and the AC power source and DC power source cannot be switched uninterruptedly. Summary of the invention

[0003] In order to solve the above technical problems, this application is proposed. The embodiment of the present application provides an AC and DC multi-channel power supply controller, which combines a deep learning algorithm based on artificial intelligence to locally extract and locally enhance the time domain fluctuation correlation features of the time series data of the device load, and introduces a feature distribution reverse mask clustering analysis mechanism to aggregate the implicit feature patterns of the device load timing correlation in each local time domain, thereby simulating and characterizing the dynamic load characteristics of the device as a whole, and based on this, automatically generating control instructions to guide the switching of DC power supply.

[0004] According to one aspect of the present application, an AC and DC multi-channel power supply controller is provided, which includes: a control component, a DC power supply circuit, an AC power supply circuit, a frequency converter and an AC load component; the DC power supply circuit is used to convert the input DC power into AC power and then input it into the frequency converter, and the AC power supply circuit is used to input the AC power into the frequency converter; the control component is respectively connected to the DC power supply circuit and the AC power supply circuit for communication, and is used to control the DC power supply circuit and the AC power supply circuit to select one of them to supply power to the frequency converter; the output end of the frequency converter is electrically connected to the AC load component, and the frequency converter is used to frequency modulate the input AC power and then input it into the AC load component; it is characterized in that the control component includes: An equipment load acquisition module is used to acquire the time series of the equipment load; an equipment load local timing correlation feature capture module is used to extract the equipment load local timing correlation features and enhance the equipment load local timing correlation features on the time series of the equipment load to obtain a sequence of enhanced equipment load gradient local timing correlation feature vectors; a clustering fusion representation module is used to input the sequence of enhanced equipment load gradient local timing correlation feature vectors into a feature distribution reverse mask clustering analysis network to obtain an equipment load gradient local timing correlation feature cluster representation vector; a control instruction generation module is used to determine a control instruction based on the equipment load gradient local timing correlation feature cluster representation vector, and the control instruction is used to indicate whether to switch to DC power supply.

[0005] Compared with the prior art, the present application provides an AC and DC multi-channel power supply controller, which combines a deep learning algorithm based on artificial intelligence to locally extract and locally enhance the time domain fluctuation correlation features of the time series data of the equipment load, and introduces a feature distribution reverse mask clustering analysis mechanism to aggregate the implicit feature patterns of the equipment load timing correlation in each local time domain, thereby simulating and characterizing the dynamic load characteristics of the entire equipment, and based on this, automatically generates control instructions to guide the switching of DC power supply. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] 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.

[0007] Figure 1 is a block diagram of an AC and DC multi-channel power supply controller according to an embodiment of the present application.

[0008] Figure 2 4 is a system architecture diagram of an AC and DC multi-channel power supply controller according to an embodiment of the present application.

[0009] Figure 3 4 is a block diagram of a control component in an AC and DC multi-channel power supply controller according to an embodiment of the present application.

[0010] Figure 4 It is a block diagram of a module for capturing local timing correlation characteristics of device loads in an AC and DC multi-channel power supply controller according to an embodiment of the present application. DETAILED DESCRIPTION

[0011] 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.

[0012] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0013] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0014] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0015] 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.

[0016] In actual application scenarios, when choosing between a DC power supply circuit and an AC power supply circuit to power the inverter, a fixed threshold method is usually used. For example, if the AC power supply voltage drops below a certain threshold, the system may switch to the DC power supply to maintain the stable operation of the load device. However, this method ignores the complex real-time demand changes of the equipment load and may not be able to flexibly adjust the power supply type, resulting in the inability to fully utilize the advantages of each power supply. Therefore, a more intelligent control solution is expected.

[0017] In the technical solution of the present application, an AC and DC multi-channel power supply controller is proposed. Figure 1 is a block diagram of an AC and DC multi-channel power supply controller according to an embodiment of the present application. Figure 2 1 is a system architecture diagram of an AC and DC multi-channel power supply controller according to an embodiment of the present application. Figure 1 and Figure 2 As shown, the AC and DC multi-channel power supply controller 300 according to the embodiment of the present application includes: a control component 310; a DC power supply circuit 320; an AC power supply circuit 330; a frequency converter 340 and an AC load component 350.

[0018] In particular, the control component 310. The control component is respectively connected to the DC power supply circuit and the AC power supply circuit, and is used to control the DC power supply circuit and the AC power supply circuit to select one of them to supply power to the inverter. In particular, in a specific example of the present application, Figure 3As shown, the control component 310 includes: an equipment load acquisition module 311, which is used to acquire the time series of the equipment load; an equipment load local timing correlation feature capture module 312, which is used to extract the equipment load local timing correlation feature and enhance the equipment load local timing correlation feature on the time series of the equipment load to obtain a sequence of enhanced equipment load gradient local timing correlation feature vectors; a clustering fusion representation module 313, which is used to input the sequence of enhanced equipment load gradient local timing correlation feature vectors into a feature distribution reverse mask clustering analysis network to obtain an equipment load gradient local timing correlation feature cluster representation vector; a control instruction generation module 314, which is used to determine a control instruction based on the equipment load gradient local timing correlation feature cluster representation vector, and the control instruction is used to indicate whether to switch to DC power supply.

[0019] Specifically, the device load acquisition module 311 is used to acquire the time series of the device load. Here, acquiring the time series of the device load can learn the dynamic trend of the device load changing over time and understand the operation mode and power consumption characteristics of the device. For example, in an embodiment of the present application, the device load data of mechanical devices such as heaters, motors, and air compressors can be collected to capture the load change status of these mechanical devices.

[0020] Specifically, the device load local time series correlation feature capture module 312 is used to extract the device load local time series correlation feature and enhance the device load local time series correlation feature of the device load time series to obtain a sequence of enhanced device load gradient local time series correlation feature vectors. In particular, in a specific example of the present application, Figure 4 As shown, the device load local time series correlation feature capture module 312 includes: a load gradient calculation unit 3121, which is used to calculate the gradient value of the device load at each time point in the time series of the device load to obtain a time series of the device load gradient value; a device load local time domain fluctuation feature extraction unit 3122, which is used to input the time series of the device load gradient value into a device load local time domain fluctuation feature extractor based on a 1D-CNN model to obtain a sequence of device load gradient local time series correlation feature vectors; a device load local time domain fluctuation feature enhancement unit 3123, which is used to input the sequence of device load gradient local time series correlation feature vectors into a feature enhancement module based on significance-globality to obtain a sequence of enhanced device load gradient local time series correlation feature vectors.

[0021] More specifically, the load gradient calculation unit 3121 is used to calculate the gradient value of the device load at each time point in the time series of the device load to obtain a time series of the device load gradient value. It should be understood that by calculating the gradient value of the device load, the increase and decrease speed of the load can be revealed, helping to identify changes in the operating state of the device, such as startup, normal operation, deceleration or stop. It should be understood that the change pattern of the gradient value can more directly reflect the energy consumption change characteristics of the device, providing a deeper insight into the trend and pattern of device energy consumption changes.

[0022] More specifically, the device load local time domain fluctuation feature extraction unit 3122 is used to input the time series of the device load gradient value into the device load local time domain fluctuation feature extractor based on the 1D-CNN model to obtain a sequence of local time series associated feature vectors of the device load gradient. That is, in the technical solution of the present application, the 1D-CNN model is used to construct a device load local time domain fluctuation feature extractor to automatically extract the local time domain fluctuation features in the time series of the device load gradient value. It is worth mentioning that the one-dimensional convolutional neural network (1D-CNN) is a deep learning model for processing sequence data. Unlike traditional convolutional neural networks (CNNs) that focus on processing two-dimensional image data, 1D-CNN is specifically designed to process one-dimensional sequence data, such as text, time series data, audio signals, etc. The convolution layer in the 1D-CNN model performs a convolution operation on the input sequence by sliding a convolution kernel to extract features at different positions. After the convolution layer, a pooling layer (such as maximum pooling or average pooling) is usually added to reduce the dimension of the feature and retain the most important information. Activation functions (such as ReLU) are usually applied to the output of the convolutional layer to introduce nonlinear characteristics. In addition, one or more fully connected layers are usually added between the convolutional layer and the final output to integrate the features extracted by the convolutional layer and generate the final output.

[0023] More specifically, the device load local time domain fluctuation feature enhancement unit 3123 is used to input the sequence of the local time series correlation feature vectors of the device load gradient into the feature enhancement module based on significance-globality to obtain the sequence of the enhanced device load gradient local time series correlation feature vectors. Since there is a large amount of redundant information in the time series of the device load gradient value, although the device load local time domain fluctuation feature extractor based on the 1D-CNN model can capture the local time series implicit correlation feature pattern contained therein, it cannot focus on the key information in the feature distribution to a certain extent, which may lead to a lack of accuracy in the characterization of the device load state. Therefore, in the technical solution of the present application, it is expected that the sequence of the local time series correlation feature vectors of the device load gradient will be input into the feature enhancement module based on significance-globality to obtain the sequence of the enhanced device load gradient local time series correlation feature vectors. Among them, the feature enhancement module based on significance-globality calculates the maximum value of each device load gradient local time series associated feature vector in the sequence of the device load gradient local time series associated feature vector to extract significance information in the feature distribution, and calculates the average value of each device load gradient local time series associated feature vector in the sequence of the device load gradient local time series associated feature vector to extract global information in the feature distribution, and then further convolutional coding, nonlinear activation and fusion are performed on the significance information and the global information to obtain the weights for enhancing important information and filtering redundant or unimportant information in the sequence of the device load gradient local time series associated feature vector. In this way, the sequence of the device load gradient local time series associated feature vector is enhanced based on significance information and global information, providing a richer feature representation for the final power control decision.

[0024] In an embodiment of the present application, the sequence of the local temporal association feature vectors of the device load gradient is input into a feature enhancement module based on significance-globality to obtain the sequence of the enhanced local temporal association feature vectors of the device load gradient, including: extracting the local temporal association significant feature vectors of the device load gradient from the sequence of the local temporal association feature vectors of the device load gradient; extracting the local temporal association global feature vectors of the device load gradient from the sequence of the local temporal association feature vectors of the device load gradient; performing neighborhood encoding and activation on the local temporal association significant feature vectors of the device load gradient and the local temporal association global feature vectors of the device load gradient to obtain the local temporal association significant feature context-associated activation feature vector of the device load gradient and the local temporal association global feature context-associated activation feature vector of the device load gradient; fusing the local temporal association significant feature context-associated activation feature vector of the device load gradient After activating the feature vector and the device load gradient local time series association-global feature context association activation feature vector to obtain the device load gradient local time series association significant-global feature context association activation feature vector, nonlinear activation is performed on the device load gradient local time series association significant-global feature context association activation feature vector to obtain the device load gradient local time series association significant-global feature context association weight feature vector; using the device load gradient local time series association significant-global feature context association weight feature vector as the weight vector, positionally dot multiplying each device load gradient local time series association feature vector in the sequence of device load gradient local time series association feature vectors, and adding the sequence obtained by positional dot multiplication to the sequence of device load gradient local time series association feature vectors element by element to obtain the sequence of enhanced device load gradient local time series association feature vectors.

[0025] Among them, the process of extracting the device load gradient local timing association significant feature vector of the sequence of the device load gradient local timing association feature vector includes: calculating the maximum value of each device load gradient local timing association feature vector in the sequence of the device load gradient local timing association feature vector to obtain the device load gradient local timing association significant feature vector composed of each said maximum value; and the process of extracting the device load gradient local timing association global feature vector of the sequence of the device load gradient local timing association feature vector includes: calculating the average value of each device load gradient local timing association feature vector in the sequence of the device load gradient local timing association feature vector to obtain the device load gradient local timing association global feature vector composed of each said average value. More specifically, the local temporal association significant feature vector of the device load gradient and the local temporal association global feature vector of the device load gradient are neighborhood encoded and activated to obtain the local temporal association significant feature context-associated activation feature vector of the device load gradient and the local temporal association global feature context-associated activation feature vector of the device load gradient, including: performing one-dimensional convolution encoding, nonlinear activation and linear transformation on the local temporal association significant feature vector of the device load gradient to obtain the local temporal association significant feature context-associated activation feature vector of the device load gradient; performing one-dimensional convolution encoding, nonlinear activation and linear transformation on the local temporal association global feature vector of the device load gradient to obtain the local temporal association global feature context-associated activation feature vector of the device load gradient.

[0026] In a specific example of the present application, after fusing the local temporal correlation significant feature context-associated activation feature vector of the device load gradient and the local temporal correlation global feature context-associated activation feature vector of the device load gradient to obtain the local temporal correlation significant-global feature context-associated activation feature vector of the device load gradient, the local temporal correlation significant-global feature context-associated activation feature vector of the device load gradient is nonlinearly activated to obtain the local temporal correlation significant-global feature context-associated weight feature vector of the device load gradient, including: after respectively passing the local temporal correlation significant-global feature context-associated activation feature vector of the device load gradient through the tanh activation function and the sigmoid activation function, the two obtained vectors are point multiplied to obtain the local temporal correlation significant-global feature context-associated weight feature vector of the device load gradient. More specifically, after respectively passing the local temporal correlation significant-global feature context-associated activation feature vector of the device load gradient through the tanh activation function and the sigmoid activation function according to the following formula, the two obtained vectors are point multiplied to obtain the local temporal correlation significant-global feature context-associated weight feature vector of the device load gradient; wherein the formula is: ;in, is the sequence of local time series correlation feature vectors of the equipment load gradient, is the global eigenvector of the local time series correlation of the equipment load gradient, is the significant feature vector of the local temporal correlation of the equipment load gradient, It means taking the average value, Indicates taking the maximum value, represents a one-dimensional convolution operation, express Activation function, and are the first weight matrix and the second weight matrix for linear transformation, and They are the device load gradient local temporal correlation global feature context correlation activation feature vector and the device load gradient local temporal correlation significant feature context correlation activation feature vector, express function, express Activation function, represents dot product, is a sequence of local time series correlation feature vectors of the load gradient of the enhancement device.

[0027] It is worth mentioning that in other specific examples of the present application, the time series of the device load can also be used to extract local timing correlation features of the device load and enhance the local timing correlation features of the device load to obtain a sequence of enhanced local timing correlation feature vectors of the device load gradient, for example: input the time series of the device load; perform a sliding window operation on the device load time series data, and select the data in the local time window as input; extract features from the data in each local time window, and use statistical features (such as mean, variance, maximum value, minimum value, etc.) or frequency domain features (such as Fourier transform) and other methods to extract timing correlation features; apply a filter to smooth the timing data or remove high-frequency components to reduce noise interference; convert the timing data to another time domain, such as wavelet transform, to obtain better feature representation; calculate the gradient of the device load time series to capture the rate information of load change; and compose the enhanced local timing correlation feature vectors of the device load into a sequence to obtain a sequence of enhanced local timing correlation feature vectors of the device load gradient.

[0028] Specifically, the cluster fusion representation module 313 is used to input the sequence of the enhanced device load gradient local time series correlation feature vectors into the feature distribution reverse mask cluster analysis network to obtain the device load gradient local time series correlation feature cluster representation vector. Considering that the local time domain fluctuation characteristics of the device load may not be able to represent the overall load state, in the technical solution of the present application, it is expected that the sequence of the enhanced device load gradient local time series correlation feature vectors will be input into the feature distribution reverse mask cluster analysis network to aggregate the local load change feature representation, thereby obtaining the device load gradient local time series correlation feature cluster representation vector.

[0029] In an embodiment of the present application, the specific process of inputting the sequence of enhanced device load gradient local time series association feature vectors into the feature distribution reverse mask clustering analysis network to obtain the device load gradient local time series association feature cluster representation vector includes: first calculating the positional mean vector of the sequence of enhanced device load gradient local time series association feature vectors to cluster the sequence of enhanced device load gradient local time series association feature vectors to obtain the device load gradient local time series association cluster center vector. Here, the device load gradient local time series association cluster center vector can be regarded as a representative feature of the device load change state under each local time span. Then, the semantic correlation coefficient between each enhanced device load gradient local time series association feature vector in the sequence of enhanced device load gradient local time series association feature vectors and the device load gradient local time series association cluster center vector is calculated to obtain a sequence of semantic correlation coefficients. That is, the correlation and similarity between each enhanced device load gradient local time series association feature vector and the device load gradient local time series association cluster center vector are quantified by calculating the semantic correlation coefficient; specifically, the process of calculating the semantic correlation coefficient between each enhanced device load gradient local time series association feature vector in the sequence of the enhanced device load gradient local time series association feature vector and the device load gradient local time series association cluster center vector to obtain the sequence of semantic correlation coefficients includes: dividing the point product result between the transposed vector of the i-th enhanced device load gradient local time series association feature vector and the device load gradient local time series association cluster center vector by the point product result between the second norm of the i-th enhanced device load gradient local time series association feature vector and the second norm of the device load gradient local time series association cluster center vector to obtain the i-th semantic correlation coefficient. After that, taking the reciprocal of each semantic correlation coefficient in the sequence of the semantic correlation coefficient to obtain the sequence of semantic anti-correlation coefficients; and using the Sigmoid function to normalize the sequence of the semantic anti-correlation coefficients to obtain the sequence of semantic anti-correlation weight coefficients. It should be understood that, through such processing, information that is not similar or relevant to the local temporal association cluster center vector of the device load gradient in the high-dimensional semantic feature space is amplified. Next, using each semantic anti-correlation weight coefficient in the sequence of semantic anti-correlation weight coefficients as a weight, each enhanced device load gradient local temporal association feature vector in the sequence of enhanced device load gradient local temporal association feature vectors is weighted to obtain a sequence of device load gradient local temporal association suppression feature vectors; and calculating the positional difference between the sequence of enhanced device load gradient local temporal association feature vectors and the sequence of device load gradient local temporal association suppression feature vectors to obtain a sequence of device load gradient local temporal association importance highlighting feature vectors. That is, the feature distribution that has a low correlation with the overall device load state representation is suppressed.Finally, the position mean vector of the sequence of the local time series correlation importance highlighting feature vectors of the equipment load gradient is calculated to obtain the local time series correlation feature cluster representation vector of the equipment load gradient. In this way, a mask is set for the feature distribution with lower importance during the aggregation process to reversely optimize the overall time series feature pattern expression of the equipment load.

[0030] In summary, in the above embodiment, the sequence of the enhanced device load gradient local time series correlation feature vectors is input into the feature distribution reverse mask cluster analysis network to obtain the device load gradient local time series correlation feature cluster representation vector, including: using the feature distribution reverse mask cluster analysis network to perform cluster analysis on the sequence of the enhanced device load gradient local time series correlation feature vectors using the following formula to obtain the device load gradient local time series correlation feature cluster representation vector; wherein the formula is: ;in, is a sequence of local temporal correlation feature vectors of enhanced device load gradients, , , and are respectively the first enhanced device load gradient local timing association feature vector, the second enhanced device load gradient local timing association feature vector, the i-th enhanced device load gradient local timing association feature vector and the N-th enhanced device load gradient local timing association feature vector in the sequence of the enhanced device load gradient local timing association feature vector, The value of is the number of enhanced device load gradient local time series associated feature vectors in the sequence of enhanced device load gradient local time series associated feature vectors, is the local time series correlation cluster center vector of the equipment load gradient, represents the transpose of a vector, is the i-th semantic correlation coefficient, represents the two-norm of a vector, is the i-th semantic anti-correlation coefficient, is the i-th semantic anti-correlation weight coefficient, represents an exponential function with a natural constant as base, is the sequence of local temporal correlation suppression feature vectors of device load gradients, , , and are respectively the first device load gradient local timing association suppression feature vector, the second device load gradient local timing association suppression feature vector, the i-th device load gradient local timing association suppression feature vector and the N-th device load gradient local timing association suppression feature vector in the sequence of the device load gradient local timing association suppression feature vector, is a sequence of feature vectors highlighting the importance of local temporal correlation of equipment load gradient, , , and They are the first device load gradient local timing correlation importance highlighting feature vector, the second device load gradient local timing correlation importance highlighting feature vector, the i-th device load gradient local timing correlation importance highlighting feature vector and the N-th device load gradient local timing correlation importance highlighting feature vector, It is the cluster representation vector of the local time series correlation feature of the equipment load gradient.

[0031] Specifically, the control instruction generation module 314 is used to determine a control instruction based on the device load gradient local time series correlation feature cluster representation vector, and the control instruction is used to indicate whether to switch to DC power supply. In particular, in a specific example of the present application, the device load gradient local time series correlation feature cluster representation vector is input into a classifier-based power controller to obtain a control instruction, and the control instruction is used to indicate whether to switch to DC power supply.

[0032] In a preferred embodiment, inputting the local timing association feature cluster representation vector of the device load gradient into a classifier-based power controller to obtain a control instruction specifically includes: passing the local timing association feature cluster representation vector of the device load gradient through a probabilistic activation function to obtain a probabilistic local timing association feature cluster representation vector of the device load gradient; determining a switching probability value indicating whether the local timing association feature cluster representation vector of the device load gradient is switched to DC power supply obtained by the classifier-based power controller; performing point subtraction between a unit feature vector and the probabilistic local timing association feature cluster representation vector of the device load gradient, using the power function as the base with the switching probability value as an exponent, and further performing point multiplication with the probabilistic local timing association feature cluster representation vector of the device load gradient to obtain a probabilistic local timing association feature cluster representation vector of the device load gradient. The invention relates to a device load gradient local timing association feature clustering convergence limiting feature vector; taking the absolute value of each eigenvalue of the probabilistic device load gradient local timing association feature clustering representation vector as the base, calculating a power function with the inverse of the switching probability value as the exponent, and summing all eigenvalues ​​of the probabilistic device load gradient local timing association feature clustering representation vector to obtain a device load gradient local timing association feature clustering convergence bias eigenvalue; performing dot addition on the product of the probabilistic device load gradient local timing association feature clustering convergence limiting feature vector and the device load gradient local timing association feature clustering convergence bias eigenvalue and a weight as a hyperparameter to obtain an optimized device load gradient local timing association feature clustering representation vector; passing the optimized device load gradient local timing association feature clustering representation vector through the classifier-based power supply controller to obtain the control instruction.

[0033] The probabilistic device load gradient local time series correlation feature clustering representation vector is recorded as , the switching probability value is , then the optimized equipment load gradient local time series correlation feature clustering representation vector It is expressed as: ;in, is the probabilistic clustering representation vector of the local temporal correlation feature of the equipment load gradient, represents the unit eigenvector with all eigenvalues ​​being one, is the weight as a hyperparameter, is the first representation vector of the probabilistic local temporal correlation feature clustering of the equipment load gradient eigenvalues, represents the switching probability value, is a power function with the switching probability value as an exponent to calculate the eigenvalue of each position in the eigenvector, It means point multiplication by position. It means to make a difference by position. It means adding by position. is the cluster representation vector of the optimized local time series correlation feature of the equipment load gradient.

[0034] Here, the sequence of local timing correlation feature vectors of the device load gradient expresses the one-dimensional local timing correlation characteristics of the device load gradient value. Therefore, after performing significance-global feature enhancement and distribution reverse mask clustering based on the timing feature distribution, the obtained local timing correlation feature cluster representation vector of the device load gradient will also have a priori-posteriori class probability causal association missing based on timing feature distribution enhancement and reverse mask clustering relative to the sequence of local timing correlation feature vectors of the device load gradient, causing a feature distribution regression mode shift, resulting in an overflow of the control instruction obtained by the local timing correlation feature cluster representation vector of the device load gradient through a classifier-based power controller, affecting its accuracy.

[0035] Accordingly, in the technical solution of the present application, for the local timing correlation feature cluster representation vector of the device load gradient, based on the convergence limit of the Cauchy-Hadamard form power series probability distribution, the feature set of the local timing correlation feature cluster representation vector of the device load gradient is represented by a formal power series based on the convergence probability, and a set convergence bounded bias based on each eigenvalue of the local timing correlation feature cluster representation vector of the device load gradient is added as a convergence radius representation, so as to avoid the invalid overflow of the classification result caused by the deviation of the distribution pattern of the local timing correlation feature cluster representation vector of the device load gradient due to the gradient propagation of the model parameters based on the convergence probability during the regression process of the classifier-based power controller under a unified probability convergence limit standard, thereby improving the accuracy of the control instructions obtained by the classifier-based power controller.

[0036] In particular, the DC power supply circuit 320 is used to convert the input DC power into AC power and then input it into the frequency converter.

[0037] In particular, the AC power circuit 330 is used to input AC power into the frequency converter.

[0038] In particular, the frequency converter 340 and the AC load component 350. The output end of the frequency converter is electrically connected to the AC load component, and the frequency converter is used to frequency-modulate the input AC power and then input it to the AC load component.

[0039] As described above, the AC and DC multi-way power supply controller 300 according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with an AC and DC multi-way power supply control algorithm. In a possible implementation, the AC and DC multi-way power supply controller 300 according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the AC and DC multi-way power supply controller 300 can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the AC and DC multi-way power supply controller 300 can also be one of the many hardware modules of the wireless terminal.

[0040] Alternatively, in another example, the AC and DC multi-way power supply controller 300 and the wireless terminal may also be separate devices, and the AC and DC multi-way power supply controller 300 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0041] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. An AC and DC multi-channel power supply controller, comprising a control component, a DC power supply circuit, an AC power supply circuit, a frequency converter and an AC load component; the DC power supply circuit is used to convert the input DC power into AC power and then input it into the frequency converter, and the AC power supply circuit is used to input AC power into the frequency converter; the control component is respectively connected to the DC power supply circuit and the AC power supply circuit for communication, and is used to control the DC power supply circuit and the AC power supply circuit to select one of them to supply power to the frequency converter; the output end of the frequency converter is electrically connected to the AC load component, and the frequency converter is used to frequency-modulate the input AC power and then input it into the AC load component; characterized in that The control component includes: an equipment load acquisition module, which is used to acquire the time series of the equipment load; an equipment load local timing correlation feature capture module, which is used to extract the equipment load local timing correlation features and enhance the equipment load local timing correlation features on the time series of the equipment load to obtain a sequence of enhanced equipment load gradient local timing correlation feature vectors; a cluster fusion representation module, which is used to input the sequence of enhanced equipment load gradient local timing correlation feature vectors into a feature distribution reverse mask cluster analysis network to obtain an equipment load gradient local timing correlation feature cluster representation vector; a control instruction generation module, which is used to determine a control instruction based on the equipment load gradient local timing correlation feature cluster representation vector, and the control instruction is used to indicate whether to switch to direct current power supply.

2. The AC and DC multi-channel power supply controller according to claim 1, characterized in that: The device load local time series correlation feature capture module includes: a load gradient calculation unit, which is used to calculate the gradient value of the device load at each time point in the time series of the device load to obtain a time series of the device load gradient value; a device load local time domain fluctuation feature extraction unit, which is used to input the time series of the device load gradient value into a device load local time domain fluctuation feature extractor based on the 1D-CNN model to obtain a sequence of device load gradient local time series correlation feature vectors; and a device load local time domain fluctuation feature enhancement unit, which is used to input the sequence of device load gradient local time series correlation feature vectors into a feature enhancement module based on significance-globality to obtain a sequence of enhanced device load gradient local time series correlation feature vectors.

3. The AC and DC multi-channel power supply controller according to claim 2, characterized in that: The device load local time domain fluctuation feature enhancement unit includes: a saliency information extraction subunit, used to extract a device load gradient local time series associated saliency feature vector of a sequence of the device load gradient local time series associated feature vectors; a global information extraction subunit, used to extract a device load gradient local time series associated global feature vector of a sequence of the device load gradient local time series associated feature vectors; a neighborhood encoding and activation subunit, used to perform neighborhood encoding and activation on the device load gradient local time series associated saliency feature vector and the device load gradient local time series associated global feature vector to obtain a device load gradient local time series associated saliency feature context associated activation feature vector and a device load gradient local time series associated global feature context associated activation feature vector; a weight determination subunit, used to fuse the device load gradient local time series associated saliency feature context associated activation feature vector and the device load gradient local time series associated global feature vector. After preparing the load gradient local timing association global feature context association activation feature vector to obtain the device load gradient local timing association significant-global feature context association activation feature vector, the device load gradient local timing association significant-global feature context association activation feature vector is nonlinearly activated to obtain the device load gradient local timing association significant-global feature context association weight feature vector; a weight application subunit is used to use the device load gradient local timing association significant-global feature context association weight feature vector as the weight vector, perform positional dot multiplication on each device load gradient local timing association feature vector in the sequence of the device load gradient local timing association feature vector, and add the sequence obtained by positional dot multiplication to the sequence of the device load gradient local timing association feature vectors element by element to obtain the sequence of the enhanced device load gradient local timing association feature vectors.

4. The AC and DC multi-channel power supply controller according to claim 3, characterized in that: The saliency information extraction subunit is used to calculate the maximum value of each device load gradient local time series associated feature vector in the sequence of the device load gradient local time series associated feature vector to obtain the device load gradient local time series associated saliency feature vector composed of each of the maximum values.

5. The AC and DC multi-channel power supply controller according to claim 4, characterized in that: The global information extraction subunit is used to calculate the average value of each device load gradient local time series correlation feature vector in the sequence of the device load gradient local time series correlation feature vector to obtain the device load gradient local time series correlation global feature vector composed of each of the average values.

6. The AC and DC multi-channel power supply controller according to claim 5, characterized in that: The neighborhood coding and activation subunit is used to: perform one-dimensional convolution coding, nonlinear activation and linear transformation on the local temporal association significant feature vector of the device load gradient to obtain the local temporal association significant feature context-associated activation feature vector of the device load gradient; perform one-dimensional convolution coding, nonlinear activation and linear transformation on the local temporal association global feature vector of the device load gradient to obtain the local temporal association global feature context-associated activation feature vector of the device load gradient.

7. The AC and DC multi-channel power supply controller according to claim 6, characterized in that: The weight determination subunit is used to: after respectively passing the device load gradient local temporal correlation significant-global feature context correlation activation feature vector through the tanh activation function and the sigmoid activation function, perform dot multiplication on the two vectors obtained to obtain the device load gradient local temporal correlation significant-global feature context correlation weight feature vector.

8. The AC and DC multi-channel power supply controller according to claim 7, characterized in that: The clustering fusion representation module includes: a clustering unit, which is used to calculate the position mean vector of the sequence of the enhanced device load gradient local time series association feature vector to cluster the sequence of the enhanced device load gradient local time series association feature vector to obtain the device load gradient local time series association cluster center vector; a semantic relevance measurement unit, which is used to calculate the semantic correlation coefficient between each enhanced device load gradient local time series association feature vector in the sequence of the enhanced device load gradient local time series association feature vector and the device load gradient local time series association cluster center vector to obtain a sequence of semantic correlation coefficients; a semantic anti-correlation measurement unit, which is used to take the inverse of each semantic correlation coefficient in the sequence of the semantic correlation coefficient to obtain a sequence of semantic anti-correlation coefficients; and a semantic anti-correlation normalization unit, which is used to normalize the sequence of the semantic anti-correlation coefficients using a Sigmoid function. The method comprises the following steps: a first step of calculating the position difference between the sequence of enhanced device load gradient local time series association feature vectors and the sequence of device load gradient local time series association suppression feature vectors to obtain a sequence of device load gradient local time series association importance highlighting feature vectors; and a second step of calculating the position difference between the sequence of enhanced device load gradient local time series association feature vectors and the sequence of device load gradient local time series association suppression feature vectors to obtain a sequence of device load gradient local time series association importance highlighting feature vectors. The method comprises the following steps: a first step of calculating the position difference between the sequence of enhanced device load gradient local time series association feature vectors and the sequence of device load gradient local time series association suppression feature vectors to obtain a sequence of device load gradient local time series association importance highlighting feature vectors. The method comprises the following steps: a first step of calculating the position difference between the sequence of enhanced device load gradient local time series association feature vectors and the sequence of device load gradient local time series association suppression feature vectors to obtain a sequence of device load gradient local time series association importance highlighting feature vectors. The method comprises the following steps: a first step of calculating the position mean vector of the sequence of device load gradient local time series association importance highlighting feature vectors to obtain a clustering representation vector of the device load gradient local time series association feature.

9. The AC and DC multi-channel power supply controller according to claim 8, characterized in that: The semantic relevance measurement unit is used to: divide the point product result between the transposed vector of the i-th enhanced device load gradient local time series association feature vector and the device load gradient local time series association cluster center vector by the point product result between the second norm of the i-th enhanced device load gradient local time series association feature vector and the second norm of the device load gradient local time series association cluster center vector to obtain the i-th semantic relevance coefficient.

10. The AC and DC multi-channel power supply controller according to claim 9, characterized in that: The control instruction generation module is used to: input the local time series correlation feature cluster representation vector of the equipment load gradient into a classifier-based power supply controller to obtain the control instruction.

Images