Digitized CRPS power supply for realizing alternating current or direct current input function switching through module replacement

Abstract

A digital CRPS power supply for realizing alternating current or direct current input function switching through module replacement comprises a plurality of local CRPS power supplies which are used for providing electric energy for corresponding local servers, each local CRPS power supply is connected to a power grid, the input end of each local CRPS power supply is connected to a current and voltage detection circuit, and the output end of each local CRPS power supply is connected to a current conversion module. The current and voltage detection circuit and the current conversion module respectively realize a hot plug replacement function through corresponding plug boxes and are connected with the controller, and the controller is connected with the corresponding local server and is used for transmitting real-time load data to the corresponding local server according to the real-time load data transmitted by the corresponding local server. The load at the next moment is predicted by using the pre-trained long and short range memory network, and the current conversion module is controlled to adjust the conversion power according to the detection result of the current and voltage detection circuit. Accurate current type and power conversion of the local server is realized; and a replaced subrack mode is convenient to maintain and replace, and faults can be rapidly removed.

Description

Technical Field

[0001] The invention relates to a digital CRPS power supply with input function switching, in particular to a digital CRPS power supply with AC or DC input function switching realized by module replacement, and belongs to the field of digital power supplies. Background Art

[0002] Due to the different types of electricity and power requirements in different service areas, the digital CRPS power supply adds an AC-DC conversion module to achieve type and function matching, but it is not easy to replace, and the cost of implementing different power conversion accuracies on a fixed conversion module is slightly high. Therefore, a more flexible conversion method needs to be designed.

[0003] In addition, because the existing power supply with conversion function requires multiple power supplies, in actual use, part of the power supply is still the traditional AC supply power supply, and only the remaining part is the power supply with DC conversion function. When this part of the power supply changes due to the proportion of AC and DC power demand, it cannot quickly switch roles.

[0004] Third, if we consider sharing modules between different power sources, it will bring high cost of line connection, which is only suitable for small-scale applications. Therefore, the best way is for local power sources to only meet local needs and obey a unified power dispatching plan.

[0005] Fourth, the security of data generated during scheduling also needs to be considered. The existing technology does not consider the risk of power failure of the server due to data security. Summary of the invention

[0006] In view of the above problems in the prior art, the present invention will propose a flexible way of module replacement to meet the needs of flexible replacement of AC and DC functions. Specifically, it provides a digital CRPS power supply that can realize AC or DC input function switching by module replacement, including multiple local CRPS power supplies, which are responsible for providing power to the corresponding local servers. Each of the local CRPS power supplies is connected to the power grid, and its input end is connected to the current and voltage detection circuit, and the output end is connected to the current conversion module. The current and voltage detection circuit and the current conversion module each realize the hot-swap replacement function through the corresponding plug-in box, and are both connected to the controller. The controller is connected to the corresponding local server, and is used to predict the load at the next moment according to the real-time load data transmitted by the corresponding local server using a pre-trained long short-term memory network, and based on this, according to the detection result of the current current and voltage detection circuit, control the current conversion module to adjust the conversion power, wherein, There are multiple current and voltage detection circuits and multiple current conversion modules, which are respectively used to meet requirements of different detection accuracy and different conversion power and conversion accuracy.

[0007] Optionally, each current and voltage detection circuit and the current conversion module are twin circuits. When inserted into the corresponding plug-in box, they are connected to the switching circuit in the plug-in box. The switching circuit communicates with the controller and is used to switch between different twin circuits according to the control switching signal after the control circuit detects a fault signal.

[0008] The methods for pre-training long short-term memory networks include: The S1 controller sends a data acquisition request to the local server, and the server sends the historical load data to the controller based on the request; A long short-term memory network is constructed in the S2 controller, which is composed of multiple units corresponding to the corresponding moments of the load data. The load data is divided into a training set and a prediction set. A zero vector is input at the input end of the initial unit, and the training set at the zero moment is input into the initial unit; the first prediction data is output at the output end of the initial unit, the first prediction data is input into the second unit, the second prediction data is output in the second unit, and the second prediction data is input into the third unit, and so on. There is also a transmission layer between the units for transmitting the transmission layer information to their respective next units. A loss function is formed at the output end of each unit by comparing it with the actual load data at the corresponding moment, which is used to optimize the parameters of the network unit. After continuous training, the verification set is used to verify the accuracy, and the training is stopped when the loss function reaches the minimum.

[0009] Optionally, the frequency of real-time load data transmitted by the local server is once every 1-5 seconds, the time span of the requested historical load data is any one of one day, one week, one month, half a year, one year, and 2-3 years, and the number of the current and voltage detection circuits and the current conversion modules ranges from 2 to 5.

[0010] Preferably, a data acquisition request is issued in S1, and a specific method in which the server sends the historical load data to the controller according to the request is: S1-1 copies the historical load data randomly selected by the first random algorithm in the server to the controller; The S1-2 controller selects historical load data of multiple different time periods according to the randomly selected historical load data using a second random algorithm different from the first random algorithm, splices the selected historical load data into pass codes, thereby making multiple pass codes, and selects any number of pass codes to be sent together with the request; The S1-3 server compares all received pass codes with the pre-stored pass codes one by one. After all of them match, it starts to accept requests and randomly selects any number of pass codes and sends them to the controller along with the load data. The S1-4 controller also receives the sent pass code, compares it one by one with the multiple pass codes it has produced, and after all of them match, it receives the sent load data and copies it into the memory.

[0011] Preferably, the second random algorithm selects historical load data of multiple different time periods, and specifically, the random algorithms used for selecting historical load data of at least two time periods are different.

[0012] Optionally, the first random algorithm includes simple random sampling, and the second random algorithm includes stratified sampling, cluster sampling, systematic sampling, and multi-stage sampling.

[0013] Optionally, if the conversion power according to the prediction does not match the actual load data at the next moment and exceeds or falls below a threshold, the controller adjusts the conversion power value.

[0014] Preferably, if the load data mismatch at multiple consecutive moments exceeds or falls below a threshold, the maximum value or minimum value of the exceeding or falling below portion is selected as a compensation value for all subsequent prediction values, for correcting the conversion power.

[0015] The digital CRPS power supply provided by the present invention that realizes AC or DC input function switching by module replacement has the following beneficial effects: The long short-term memory network is used to predict power demand in real time, and the current type and power conversion are accurate in the local server. The replaceable plug-in box mode is adopted to facilitate maintenance, replacement, and quick troubleshooting. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 Schematic diagram of the digital CRPS power supply that can realize AC or DC input function switching by module replacement. Figure 2 LSTM training flow chart, Figure 3a Schematic diagram of historical load data selected by simple random sampling, Figure 3b Stratified sampling and multi-stage sampling were used to select the historical load data diagram for 15 different periods. Figure 3c Schematic diagram of multiple pass code styles. DETAILED DESCRIPTION

[0017] Figure 1The structure of a digital CRPS power supply that can realize AC or DC input function switching by module replacement is given. Taking a local area as an example, it includes a local CRPS power supply, which is responsible for providing power to the corresponding local server. The local CRPS power supply is connected to the power grid, and its input end is connected to the current and voltage detection circuit, and the output end is connected to the current conversion module. The current and voltage detection circuit and the current conversion module each realize the hot-swap replacement function through a corresponding plug-in box (not shown in the figure), and are both connected to a controller. The controller is connected to the corresponding local server and is used to predict the load at the next moment based on the real-time load data transmitted by the corresponding local server using a pre-trained long short-term memory network, and based on this, according to the detection result of the current current and voltage detection circuit, the current conversion module is controlled to adjust the conversion power, wherein, There are two current and voltage detection circuits and two current conversion modules, which are respectively used to meet requirements of different detection accuracy and different conversion power and conversion accuracy.

[0018] Among them, each current and voltage detection circuit and the current conversion module are twin circuits, that is, two sets of short circuits with exactly the same circuit structure. After being inserted into the corresponding plug-in box, the switching circuit in the plug-in box is connected, and the switching circuit communicates with the controller, and is used to switch between different twin circuits according to the control switching signal after the control circuit detects a fault signal.

[0019] These fault signals may indicate that the detection result of the detection circuit does not conform to the preset rule, or that the power value converted by the current conversion module and the target value are beyond the allowable error range.

[0020] Among them, Figure 1~Figure 3c As shown, the method of long short-term memory network pre-training includes: The S1 controller sends a data acquisition request to the local server, and the server sends the historical load data to the controller based on the request; A long short-term memory network is constructed in the S2 controller, which is composed of multiple units corresponding to the corresponding moments of the load data. The load data is divided into a training set and a prediction set. A zero vector is input at the input end of the initial unit, and the training set at the zero point is input at the initial unit; the first prediction data is output at the output end of the initial unit, the first prediction data is input into the second unit, the second prediction data is output in the second unit, and the second prediction data is input into the third unit, and so on. There is also a transmission layer between each unit, which is used to transmit the transmission layer information to each next unit. At the output end of each unit, a loss function is formed by comparing it with the actual load data at the corresponding moment ( Figure 2The first loss function and the second loss function are given as an example, where serial numbers 1, 2, and 3 are network unit serial numbers, representing the first unit, the second unit, and the third unit, respectively, and 4 represents the transmission layer) are used to optimize the parameters of the network units. After continuous training, the verification set is used to verify the accuracy. When the loss function reaches the minimum, the training is stopped.

[0021] The frequency of real-time load data transmitted by the local server is once per second, the time span of the requested historical load data is one day, and the number range of the current and voltage detection circuits and the current conversion modules is 2.

[0022] Combination Figure 3a~Figure 3c , S1 issues a data acquisition request, and the specific method for the server to send historical load data to the controller according to the request is: S1-1 copies the historical load data selected by simple random sampling in the server to the controller; Figure 3a The box in the middle shows a time period that is randomly selected.

[0023] The S1-2 controller selects 15 different periods of historical load data based on the selected historical load data using stratified sampling and multi-stage sampling, of which 10 time periods are selected by multi-stage sampling and the remaining 5 are the results of stratified sampling. Figure 3b The time points represented by the vertical line segments are the corresponding time points of the 15 selected historical load data.

[0024] Then these selected historical load data are spliced ​​into the pass code ( Figure 3c ), and then make multiple pass codes, and select any number of pass codes to send with the request. For the entire request and sending process, see Figure 1 . Figure 3c In it, multiple character strings are formed by arranging codes.

[0025] The S1-3 server compares all received pass codes with the pre-stored pass codes one by one. After all of them match, it starts to accept requests and randomly selects any number of pass codes and sends them to the controller along with the load data.

[0026] The S1-4 controller also receives the sent pass code, compares it one by one with the multiple pass codes it has produced, and after all of them match, it receives the sent load data and copies it into the memory.

[0027] If the conversion power according to the prediction and the actual load data at the next moment do not match (percentage) beyond or below the threshold, the controller adjusts the conversion power value. The threshold can be selected from 1%-5%. If the load data mismatch at three consecutive moments exceeds or falls below the threshold, the maximum or minimum value of the exceeding or falling part is selected as the compensation value for all subsequent prediction values ​​to correct the conversion power.

Claims

1. Module replacement realizes the digital CRPS power supply with AC or DC input function switching, characterized in that: It includes multiple local CRPS power supplies, which are responsible for providing power to the corresponding local servers. Each of the local CRPS power supplies is connected to the power grid, and its input end is connected to the current and voltage detection circuit, and the output end is connected to the current conversion module. The current and voltage detection circuit and the current conversion module each realize the hot-swap replacement function through the corresponding plug-in box, and are all connected to the controller. The controller is connected to the corresponding local server, and is used to predict the load at the next moment according to the real-time load data transmitted by the corresponding local server using the pre-trained long short-term memory network, and control the current conversion module to adjust the conversion power according to the detection result of the current current and voltage detection circuit. There are multiple current and voltage detection circuits and multiple current conversion modules, which are respectively used to meet requirements of different detection accuracy and different conversion power and conversion accuracy.

2. The CRPS power supply according to claim 1, characterized in that: Each current and voltage detection circuit and the current conversion module are twin circuits. When inserted into the corresponding plug-in box, they are connected to the switching circuit in the plug-in box. The switching circuit communicates with the controller and is used to switch between different twin circuits according to the controlled switching signal after the control circuit detects a fault signal.

3. The CRPS power supply according to claim 1, characterized in that: The methods for pre-training long short-term memory networks include: The S1 controller sends a data acquisition request to the local server, and the server sends the historical load data to the controller based on the request; A long short-term memory network is constructed in the S2 controller, which is composed of multiple units corresponding to the corresponding moments of the load data. The load data is divided into a training set and a prediction set. A zero vector is input at the input end of the initial unit, and the training set at the zero moment is input into the initial unit; the first prediction data is output at the output end of the initial unit, the first prediction data is input into the second unit, the second prediction data is output in the second unit, and the second prediction data is input into the third unit, and so on. There is also a transmission layer between the units for transmitting the transmission layer information to their respective next units. A loss function is formed at the output end of each unit by comparing it with the actual load data at the corresponding moment, which is used to optimize the parameters of the network unit. After continuous training, the verification set is used to verify the accuracy, and the training is stopped when the loss function reaches the minimum.

4. The CRPS power supply according to claim 3, characterized in that: The frequency of real-time load data transmitted by the local server is once every 1-5 seconds, and the time span of the requested historical load data is any one of one day, one week, one month, half a year, one year, and 2-3 years. The number of the current and voltage detection circuits and the current conversion modules ranges from 2 to 5.

5. The CRPS power supply according to claim 3 or 4, characterized in that: S1 issues a data acquisition request, and the specific method for the server to send historical load data to the controller according to the request is: S1-1 copies the historical load data randomly selected by the first random algorithm in the server to the controller; The S1-2 controller selects historical load data of multiple different time periods according to the randomly selected historical load data using a second random algorithm different from the first random algorithm, splices the selected historical load data into pass codes, thereby making multiple pass codes, and selects any number of pass codes to be sent together with the request; The S1-3 server compares all received pass codes with the pre-stored pass codes one by one. After all of them match, it starts to accept requests and randomly selects any number of pass codes and sends them to the controller along with the load data. The S1-4 controller also receives the sent pass code, compares it one by one with the multiple pass codes it has produced, and after all of them match, it receives the sent load data and copies it into the memory.

6. The CRPS power supply according to claim 5, characterized in that: The second random algorithm selects historical load data of multiple different time periods. Specifically, different random algorithms are used to select historical load data of at least two time periods.

7. The CRPS power supply according to claim 6, characterized in that: The first randomization algorithm includes, simple random sampling, and the second randomization algorithm includes stratified sampling, cluster sampling, systematic sampling, and multi-stage sampling.

8. The CRPS power supply according to claim 7, characterized in that: If the conversion power according to the prediction does not match the actual load data at the next moment and exceeds or falls below the threshold, the controller adjusts the conversion power value.

9. The CRPS power supply according to claim 8, characterized in that: If the load data mismatch at multiple consecutive moments exceeds or falls below the threshold, the maximum or minimum value of the exceeding or falling part is selected as the compensation value for all subsequent prediction values ​​to correct the conversion power.

Images