A TCS fault troubleshooting method and device based on deep learning

Abstract

The present invention provides a TCS troubleshooting method and device based on deep learning, the method comprising: obtaining initial selection parameters, and controlling the simulated TCS to pre-start based on the initial selection parameters to generate a pre-start result; when the pre-start result is a successful start, inputting the initial selection parameters into a target deep learning model with complete learning, obtaining a first recommended optimization parameter, and controlling the TCS to start based on the first recommended optimization parameter; when the pre-start result is a start failure, obtaining start failure data, determining a second recommended optimization parameter based on the start failure data and normal start data, and controlling the TCS to start based on the second recommended optimization parameter. The present invention can improve the start success rate and start efficiency of TCS.

Description

Technical Field

[0001] The present invention relates to the technical field of step-down compensation solid-state soft starting, and in particular to a TCS fault troubleshooting method and device based on deep learning. Background Art

[0002] The step-down compensated solid-state soft starter (TCS) is used to realize the soft starting of the motor.

[0003] Since the starting effect of TCS is greatly affected by the user's power grid, motor and load parameters, and although these parameters are generally designed by the manufacturer at the beginning of the design, there are more or less differences between them and the actual parameter values ​​of the on-site units. This leads to the situation that TCS may be repeatedly started with load several times in the early stage of commissioning or in later use, resulting in a low starting success rate. In addition, since there are many parameters that affect the starting success rate of TCS, such as a large damper opening causing excessive load, a small system short-circuit capacity causing a large busbar voltage drop, a reactive power removal voltage that is too low causing reactive power removal too early, and a starting time setting that is too short, etc., it is often impossible to quickly troubleshoot the TCS faults if the user or after-sales personnel only rely on their own personal experience to judge the cause of each starting failure, resulting in low TCS fault troubleshooting efficiency. Moreover, the TCS in the prior art only indicates that a fault has occurred, and the fault alarm interface that pops up on the touch screen is relatively professional. Even general after-sales personnel need to conduct careful research and analysis, and go through one or more trial and error to find the real problem. This is not conducive to timely and accurate on-site troubleshooting of the fault, and it cannot directly indicate the specific fault location, further reducing the troubleshooting efficiency of the TCS.

[0004] Therefore, there is an urgent need to provide a TCS troubleshooting method and device based on deep learning to solve the technical problems of low troubleshooting efficiency and low starting success rate in the prior art. Summary of the invention

[0005] In view of this, it is necessary to provide a TCS troubleshooting method and device based on deep learning to solve the technical problems of low troubleshooting efficiency and low starting success rate existing in the prior art.

[0006] On the one hand, the present invention provides a TCS troubleshooting method based on deep learning, comprising:

[0007] Acquire initial selection parameters, and control the simulated TCS to perform pre-start based on the initial selection parameters to generate a pre-start result;

[0008] When the pre-start result is a successful start, the initial selection parameters are input into a fully learned target deep learning model to obtain first recommended optimization parameters, and the TCS is controlled to start based on the first recommended optimization parameters;

[0009] When the pre-start result is a start failure, start failure data is acquired, second recommended optimization parameters are determined according to the start failure data and normal start data, and the TCS is controlled to start based on the second recommended optimization parameters.

[0010] In some possible implementations, the controlling the simulated TCS to perform pre-start based on the initial selection parameters and generating a pre-start result includes:

[0011] Determining whether the initial selection parameters are complete;

[0012] When the initial selection parameters are complete, controlling the simulated TCS to perform pre-start based on the initial selection parameters to generate a pre-start result;

[0013] When the initial selection parameters are incomplete, an incomplete parameter prompt message is generated.

[0014] In some possible implementations, before inputting the initial selection parameters into the fully learned deep learning model, the method further includes:

[0015] Step 1: Build the deep learning model to be learned;

[0016] Step 2: determine whether the number of successful starts of the simulated TCS is greater than the preset number of learning times;

[0017] Step three: when the number of successful startups of the simulated TCS is greater than the preset number of learning times, the deep learning model to be learned learns the initial selection parameters corresponding to each successful startup to obtain the target deep learning model; when the number of successful startups of the simulated TCS is less than or equal to the preset number of learning times, the initial selection parameters are obtained again, and the simulated TCS is started according to the initial selection parameters, and steps two to three are repeated.

[0018] In some possible implementations, the starting fault data includes a plurality of starting fault sub-data, and the normal starting data includes a plurality of normal starting sub-data corresponding to the plurality of starting fault sub-data; and determining the second recommended optimization parameter according to the starting fault data and the normal starting data includes:

[0019] determining actual starting fault sub-data among the plurality of starting fault sub-data based on a preset troubleshooting sequence and the plurality of normal starting sub-data;

[0020] A second recommended optimization parameter is determined according to the actual starting failure sub-data and the corresponding normal starting sub-data.

[0021] In some possible implementations, determining the second recommended optimization parameter according to the actual starting fault sub-data and the corresponding normal starting sub-data includes:

[0022] Determine an actual fault difference between the actual starting fault sub-data and the corresponding normal starting sub-data;

[0023] adjusting the actual starting fault sub-data according to the actual fault difference to generate actual starting adjustment sub-data;

[0024] The second recommended optimization parameter is generated according to the actual startup adjustment sub-data.

[0025] In some possible implementations, the multiple starting fault sub-data include fault occurrence time, starting current, starting voltage, line break fault, voltage threshold, current threshold, cut-off voltage threshold, cut-off reactive power time and starting time, and the multiple normal starting sub-data include theoretical starting current, theoretical starting voltage, theoretical voltage threshold, theoretical current threshold, theoretical cut-off voltage threshold, theoretical cut-off reactive power time and theoretical starting time.

[0026] In some possible implementations, the preset troubleshooting order is: fault occurrence time troubleshooting, starting overcurrent troubleshooting, starting overvoltage troubleshooting and starting time troubleshooting. The troubleshooting order of the fault occurrence time troubleshooting is line break fault troubleshooting, voltage threshold troubleshooting and current threshold troubleshooting; the troubleshooting order of the starting overcurrent troubleshooting is cut-off voltage threshold troubleshooting and cut-off reactive power time troubleshooting; the starting overvoltage troubleshooting includes cut-off voltage threshold troubleshooting.

[0027] In some possible implementations, the TCS troubleshooting method based on deep learning further includes:

[0028] The first recommended optimization parameter and / or the second recommended optimization parameter are displayed on a touch screen of the TCS.

[0029] In some possible implementations, the TCS troubleshooting method based on deep learning further includes:

[0030] generating a first optimization parameter annotation based on the first recommended optimization parameter and / or generating a second optimization parameter annotation based on the second recommended optimization parameter;

[0031] The first optimization parameter annotation and / or the second optimization parameter annotation are displayed on a touch screen of the TCS.

[0032] On the other hand, the present invention also provides a TCS fault troubleshooting device based on deep learning, comprising:

[0033] A pre-start unit, used to obtain initial selection parameters, and control the simulated TCS to perform pre-start based on the initial selection parameters to generate a pre-start result;

[0034] A first recommended optimization parameter determination unit, configured to input the initial selection parameters into a target deep learning model that has been fully learned when the pre-start result is a successful start, and obtain first recommended optimization parameters, and control the TCS to start based on the first recommended optimization parameters;

[0035] The second recommended optimization parameter determination unit is used to obtain starting failure data when the pre-start result is a starting failure, determine second recommended optimization parameters according to the starting failure data and normal starting data, and control the TCS to start based on the second recommended optimization parameters.

[0036] The beneficial effect of adopting the above embodiment is: the TCS troubleshooting method based on deep learning provided by the present invention performs pre-starting by setting up a simulated TCS based on the initial selection parameters to generate a pre-starting result, and based on the success or failure of the pre-starting result, generates a first recommended optimization parameter and a second recommended optimization parameter respectively, which can improve the starting success rate of TCS.

[0037] Furthermore, when the pre-start result is a successful start, the initial selection parameters are input into the fully learned target deep learning model to obtain the first recommended optimization parameters, which can improve the accuracy and rationality of the obtained first recommended optimization parameters and further improve the start-up success rate of TCS.

[0038] Furthermore, when the pre-start result is a starting failure, the second recommended optimization parameters are determined based on the starting failure data and the normal starting data. There is no need to manually find the cause of the fault, and the TCS can be directly controlled to start according to the second recommended optimization parameters. There is no need to spend time manually finding the fault. In addition, the fault can be directly troubleshooted through the difference between the second recommended optimization control parameters and the initial selection parameters, thereby improving the efficiency of troubleshooting. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0040] Figure 1A schematic diagram of an embodiment of a TCS fault troubleshooting method based on deep learning provided by the present invention;

[0041] Figure 2 For the present invention Figure 1 A schematic flow chart of an embodiment of S101;

[0042] Figure 3 A schematic diagram of an embodiment of a flow chart for generating a target deep learning model provided by the present invention;

[0043] Figure 4 For the present invention Figure 1 A schematic flow chart of an embodiment of S103;

[0044] Figure 5 For the present invention Figure 4 A schematic flow chart of an embodiment of S402;

[0045] Figure 6 A schematic diagram of a flow chart of an embodiment of a preset troubleshooting sequence provided by the present invention;

[0046] Figure 7 A schematic diagram of a flow chart of an embodiment of generating optimization parameter annotations provided by the present invention;

[0047] Figure 8 A schematic structural diagram of an embodiment of a TCS troubleshooting device based on deep learning provided by the present invention. DETAILED DESCRIPTION

[0048] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0049] In the description of the embodiments of the present invention, "and / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent three situations: A exists alone, A and B exist at the same time, and B exists alone.

[0050] It should be understood that the schematic drawings are not drawn to scale. The flow chart used in the present invention shows the operations implemented according to some embodiments of the present invention. It should be understood that the operations of the flow chart can be implemented out of order, and the steps without logical context can be reversed or implemented simultaneously. In addition, those skilled in the art can add one or more other operations to the flow chart under the guidance of the content of the present invention, and can also remove one or more operations from the flow chart.

[0051] Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor systems and / or microcontroller systems.

[0052] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0053] The embodiments of the present invention provide a TCS fault troubleshooting method and device based on deep learning, which are described below respectively.

[0054] Before showing the embodiments, TCS is first explained. TCS refers to a step-down compensation solid-state soft start device, which is used to realize the soft start of the motor, and mainly includes a step-down device, a reactive generator, a step-down control cabinet and a reactive control cabinet. The parallel circuit of the motor and the reactive generator is connected to the power grid through the step-down device, and the reactive generator provides part of the reactive power consumed by the motor to reduce the starting current, and reduces the bus current by reducing the terminal voltage.

[0055] Figure 1 A flowchart of an embodiment of the TCS troubleshooting method based on deep learning provided by the present invention is shown in FIG. Figure 1 As shown in the figure, the TCS troubleshooting method based on deep learning includes:

[0056] S101, obtaining initial selection parameters, and controlling the simulated TCS to perform pre-start based on the initial selection parameters to generate a pre-start result;

[0057] S102, when the pre-start result is a successful start, inputting the initial selection parameters into the target deep learning model that has been fully learned, obtaining first recommended optimization parameters, and controlling the TCS to start based on the first recommended optimization parameters;

[0058] S103: When the pre-start result is a starting failure, obtaining starting failure data, determining second recommended optimization parameters according to the starting failure data and normal starting data, and controlling the TCS to start based on the second recommended optimization parameters.

[0059] Compared with the prior art, the deep learning-based TCS troubleshooting method provided in an embodiment of the present invention performs pre-starting by setting a simulated TCS based on initial selection parameters to generate a pre-starting result, and generates a first recommended optimization parameter and a second recommended optimization parameter based on the success or failure of the pre-starting result, respectively, thereby improving the starting success rate of TCS.

[0060] Furthermore, when the pre-start result is a successful start, the initial selection parameters are input into the fully learned target deep learning model to obtain the first recommended optimization parameters, which can improve the accuracy and rationality of the obtained first recommended optimization parameters and further improve the start-up success rate of TCS.

[0061] Furthermore, when the pre-start result is a starting failure, the second recommended optimization parameters are determined based on the starting failure data and the normal starting data. There is no need to manually find the cause of the fault, and the TCS can be directly controlled to start according to the second recommended optimization parameters. There is no need to spend time manually finding the fault. In addition, the fault can be directly troubleshooted through the difference between the second recommended optimization control parameters and the initial selection parameters, thereby improving the efficiency of troubleshooting.

[0062] It should be understood that the normal start data in step S103 may be design values ​​in the early stage of TCS design, and may be directly obtained from the design manual or technical data of TCS.

[0063] In order to avoid the situation that the simulated TCS cannot be started successfully due to incomplete initial selection parameter settings, thereby causing fault troubleshooting errors, in some embodiments of the present invention, such as Figure 2 As shown, step S101 includes:

[0064] S201, judging whether the initial selection parameters are complete;

[0065] S202, when the initial selection parameters are complete, controlling the simulated TCS to perform pre-start based on the initial selection parameters to generate a pre-start result;

[0066] S203: When the initial selection parameters are incomplete, generate a prompt message indicating that the parameters are incomplete.

[0067] The embodiment of the present invention can ensure the reliability of the simulated TCS pre-start result by judging the integrity of the initial selection parameters, and avoid the unsuccessful pre-start due to incomplete initial selection parameters. In addition, when the initial selection parameters are incomplete, the embodiment of the present invention generates an incomplete parameter prompt message to prompt the operator, which can improve the efficiency of the TCS fault troubleshooting method based on deep learning.

[0068] It should be understood that the parameter incomplete prompt information can be at least one of a variety of forms such as sound, light, pop-up window, etc.

[0069] In order to improve the reliability of the first recommended optimization parameters, in some embodiments of the present invention, Figure 3 As shown, before step S102, it also includes:

[0070] S301, constructing a deep learning model to be learned;

[0071] S302, determining whether the number of successful starts of the simulated TCS is greater than the preset number of learning times;

[0072] S303. When the number of successful startups of the simulated TCS is greater than the preset number of learning times, the deep learning model to be learned learns the initial selection parameters corresponding to each successful startup to obtain the target deep learning model; when the number of successful startups of the simulated TCS is less than or equal to the preset number of learning times, the initial selection parameters are obtained again, and the simulated TCS is started according to the initial selection parameters, and steps S302 to S303 are repeated.

[0073] The embodiment of the present invention is arranged that when the number of successful startups of the simulated TCS is greater than the preset number of learning times, the deep learning model to be learned learns the initial selection parameters to obtain the target deep learning model, thereby increasing the number of samples of the deep learning model to be learned, and improving the accuracy and reliability of the target deep learning model, thereby improving the reliability of the first recommended optimization parameters, and further achieving the purpose of improving the startup success rate of TCS.

[0074] It should be understood that the preset number of learning times can be set or adjusted according to actual conditions or experience, and is not specifically limited here.

[0075] In some embodiments of the present invention, the starting fault data includes a plurality of starting fault sub-data, and the normal starting data includes a plurality of normal starting sub-data corresponding to the plurality of starting fault sub-data; Figure 4 As shown, step S103 includes:

[0076] S401, determining actual starting fault sub-data among a plurality of starting fault sub-data based on a preset troubleshooting sequence and a plurality of normal starting sub-data;

[0077] S402: Determine a second recommended optimization parameter according to the actual starting failure sub-data and the corresponding normal starting sub-data.

[0078] The embodiment of the present invention can improve the efficiency of determining the actual starting fault sub-data by determining the actual starting fault sub-data based on a preset troubleshooting sequence, thereby improving the efficiency of determining the second recommended optimization parameter, and further improving the efficiency of TCS starting success.

[0079] In a specific embodiment of the present invention, Figure 5 As shown, step S402 includes:

[0080] S501, determining an actual fault difference between actual starting fault sub-data and corresponding normal starting sub-data;

[0081] S502, adjusting the actual starting fault sub-data according to the actual fault difference to generate actual starting adjustment sub-data;

[0082] S503: Generate second recommended optimization parameters according to the actual startup adjustment sub-data.

[0083] The embodiment of the present invention generates actual starting adjustment sub-data by adjusting actual starting fault sub-data according to actual fault difference, and generates second recommended optimization parameters according to the actual starting adjustment sub-data. The TCS can be started directly by the second recommended optimization parameters, thereby improving the starting success rate of the TCS.

[0084] In a specific embodiment of the present invention, the actual starting fault sub-data is the fault reactive cut-off voltage, and the voltage value of the fault reactive cut-off voltage is 50V, and the reactive cut-off voltage of normal starting is 78V. The actual starting adjustment sub-data is the actual adjusted reactive cut-off voltage, that is: the fault reactive cut-off voltage is modified from 50V to 78V, that is to say: the reactive cut-off voltage in the second recommended optimization parameter is 78V.

[0085] In a specific embodiment of the present invention, multiple starting fault sub-data include fault occurrence time, starting current, starting voltage, line break fault, voltage threshold, current threshold, cut-off voltage threshold, cut-off reactive power time and starting time, and multiple normal starting sub-data include theoretical starting current, theoretical starting voltage, theoretical voltage threshold, theoretical current threshold, theoretical cut-off voltage threshold, theoretical cut-off reactive power time and theoretical starting time.

[0086] Based on the above-mentioned starting fault sub-data, in a specific embodiment of the present invention, the preset troubleshooting order is: fault occurrence time troubleshooting, starting overcurrent troubleshooting, starting overvoltage troubleshooting and starting time troubleshooting. The troubleshooting order of the fault occurrence time troubleshooting is line break fault troubleshooting, starting voltage troubleshooting and starting current troubleshooting; the troubleshooting order of the starting overcurrent troubleshooting is cut-off voltage threshold troubleshooting and cut-off reactive power time troubleshooting; the starting overvoltage troubleshooting includes cut-off voltage threshold troubleshooting.

[0087] like Figure 6As shown, when simulating a TCS starting failure, first determine whether the fault occurs within 2 seconds of starting. If it occurs within 2 seconds, first determine whether it is a wire break fault. If it is a wire break fault, generate a prompt message to prompt the staff to check the sensor and determine the specific location of the wire break through the sensor. If it is not a wire break fault, determine whether it is a starting voltage signal fault. If it is a starting voltage signal fault, recommend a suitable starting voltage threshold. If it is not a starting voltage signal fault, determine whether it is a starting current signal fault. If it is a starting current signal fault, recommend a suitable starting current threshold. If the fault does not occur within 2 seconds of starting, first determine whether it is a starting overcurrent. If it is a starting overcurrent, then compare and check whether the cut-off voltage threshold setting is reasonable. If the cut-off voltage threshold setting is unreasonable, recommend a suitable cut-off voltage threshold. If the cut-off voltage threshold setting is reasonable, then determine whether the reactive power cut-off time is too early. If the reactive power cut-off time is too early, recommend a suitable reactive power cut-off time. If there is no starting overcurrent, then determine whether the starting is overvoltage. If there is starting overvoltage, then compare and check whether the cut-off voltage threshold setting is reasonable. If it is unreasonable, recommend a suitable cut-off voltage threshold. If there is no starting overvoltage, then determine whether the starting is timed out. If the starting is timed out, then compare each normal starting sub-data and starting fault sub-data to generate the second optimized recommended parameters.

[0088] In order to facilitate reminding the operator to learn the first recommended optimization parameter and the second recommended optimization parameter, in some embodiments of the present invention, the TCS fault troubleshooting method based on deep learning further includes:

[0089] The first recommended optimization parameter and / or the second recommended optimization parameter are displayed on the touch screen of the TCS.

[0090] In the embodiment of the present invention, by displaying the first recommended optimization parameter and / or the second recommended optimization parameter on the touch screen of the TCS, the operator can intuitively know the parameters for successful startup of the TCS for reference.

[0091] Furthermore, in order to facilitate the operator to know the difference between the first recommended optimization parameter and the second recommended optimization parameter and the initial selection parameter, in some embodiments of the present invention, Figure 7 As shown, the TCS troubleshooting method based on deep learning also includes:

[0092] S701, generating a first optimization parameter annotation based on a first recommended optimization parameter and / or generating a second optimization parameter annotation based on a second recommended optimization parameter;

[0093] S702: Display the first optimization parameter annotation and / or the second optimization parameter annotation on the touch screen of the TCS.

[0094] The embodiment of the present invention generates the first optimization parameter annotation and the second optimization parameter annotation, which can help the operator understand the difference between the first recommended optimization parameter and the second recommended optimization parameter and the initial selection parameter, and facilitate the operator to debug the TCS later.

[0095] In a specific embodiment of the present invention: the first optimization parameter annotation and the second optimization parameter annotation include but are not limited to: the name of the adjusted parameter and the adjusted value. For example: increasing the system short-circuit capacity of the TCS by 50MVA, etc.

[0096] In order to better implement the TCS troubleshooting method based on deep learning in the embodiment of the present invention, on the basis of the TCS troubleshooting method based on deep learning, the embodiment of the present invention also provides a TCS troubleshooting device based on deep learning, such as Figure 8 As shown, the TCS fault troubleshooting device 800 based on deep learning includes:

[0097] The pre-start unit 801 is used to obtain initial selection parameters, and control the simulated TCS to perform pre-start based on the initial selection parameters to generate a pre-start result;

[0098] A first recommended optimization parameter determination unit 802 is used to input the initial selection parameters into the target deep learning model that has been fully learned when the pre-start result is a successful start, and obtain the first recommended optimization parameters, and control the TCS to start based on the first recommended optimization parameters;

[0099] The second recommended optimization parameter determination unit 803 is used to obtain the starting fault data when the pre-start result is a starting fault, determine the second recommended optimization parameter according to the starting fault data and the normal starting data, and control the TCS to start based on the second recommended optimization parameter.

[0100] The deep learning-based TCS troubleshooting device 800 provided in the above embodiment can implement the technical solution described in the above-mentioned deep learning-based TCS troubleshooting method embodiment. The specific implementation principles of the above-mentioned modules or units can refer to the corresponding contents in the above-mentioned deep learning-based TCS troubleshooting method embodiment, which will not be repeated here.

[0101] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware (such as a processor, a controller, etc.) through a computer program, and the computer program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0102] The above is a detailed introduction to the TCS troubleshooting method and device based on deep learning provided by the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for technical personnel in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A TCS troubleshooting method based on deep learning, characterized in that: include: Acquire initial selection parameters, and control the simulated TCS to perform pre-start based on the initial selection parameters to generate a pre-start result; When the pre-start result is a successful start, the initial selection parameters are input into a fully learned target deep learning model to obtain first recommended optimization parameters, and the TCS is controlled to start based on the first recommended optimization parameters; When the pre-start result is a starting fault, acquiring starting fault data, determining a second recommended optimization parameter according to the starting fault data and normal starting data, and controlling the TCS to start based on the second recommended optimization parameter; The starting fault data includes a plurality of starting fault sub-data, and the normal starting data includes a plurality of normal starting sub-data corresponding to the plurality of starting fault sub-data; and determining the second recommended optimization parameter according to the starting fault data and the normal starting data includes: determining actual starting fault sub-data among the plurality of starting fault sub-data based on a preset troubleshooting sequence and the plurality of normal starting sub-data; A second recommended optimization parameter is determined according to the actual starting failure sub-data and the corresponding normal starting sub-data.

2. The TCS troubleshooting method based on deep learning according to claim 1 is characterized in that: The controlling the simulated TCS to perform pre-start based on the initial selection parameters and generating a pre-start result comprises: Determining whether the initial selection parameters are complete; When the initial selection parameters are complete, controlling the simulated TCS to perform pre-start based on the initial selection parameters to generate a pre-start result; When the initial selection parameters are incomplete, an incomplete parameter prompt message is generated.

3. The TCS troubleshooting method based on deep learning according to claim 1 is characterized in that: Before inputting the initial selection parameters into the complete deep learning model, the method further includes: Step 1: Build the deep learning model to be learned; Step 2: determine whether the number of successful starts of the simulated TCS is greater than the preset number of learning times; Step three: when the number of successful startups of the simulated TCS is greater than the preset number of learning times, the deep learning model to be learned learns the initial selection parameters corresponding to each successful startup to obtain the target deep learning model; when the number of successful startups of the simulated TCS is less than or equal to the preset number of learning times, the initial selection parameters are obtained again, and the simulated TCS is started according to the initial selection parameters, and steps two to three are repeated.

4. The TCS troubleshooting method based on deep learning according to claim 1 is characterized in that: The determining of the second recommended optimization parameter according to the actual starting fault sub-data and the corresponding normal starting sub-data includes: Determine an actual fault difference between the actual starting fault sub-data and the corresponding normal starting sub-data; adjusting the actual starting fault sub-data according to the actual fault difference to generate actual starting adjustment sub-data; The second recommended optimization parameter is generated according to the actual startup adjustment sub-data.

5. The TCS troubleshooting method based on deep learning according to claim 1 is characterized in that: The multiple starting fault sub-data include fault occurrence time, starting current, starting voltage, line break fault, voltage threshold, current threshold, cut-off voltage threshold, cut-off reactive power time and starting time; the multiple normal starting sub-data include theoretical starting current, theoretical starting voltage, theoretical voltage threshold, theoretical current threshold, theoretical cut-off voltage threshold, theoretical cut-off reactive power time and theoretical starting time.

6. The TCS troubleshooting method based on deep learning according to claim 5 is characterized in that: The preset troubleshooting order is: fault occurrence time troubleshooting, starting overcurrent troubleshooting, starting overvoltage troubleshooting and starting time troubleshooting. The troubleshooting order of the fault occurrence time troubleshooting is line break fault troubleshooting, starting voltage troubleshooting and starting current troubleshooting; the troubleshooting order of the starting overcurrent troubleshooting is cut-off voltage threshold troubleshooting and cut-off reactive power time troubleshooting; the starting overvoltage troubleshooting includes cut-off voltage threshold troubleshooting.

7. The TCS troubleshooting method based on deep learning according to claim 1 is characterized in that: The TCS troubleshooting method based on deep learning also includes: The first recommended optimization parameter and / or the second recommended optimization parameter are displayed on a touch screen of the TCS.

8. The TCS troubleshooting method based on deep learning according to claim 7 is characterized in that: The TCS troubleshooting method based on deep learning also includes: generating a first optimization parameter annotation based on the first recommended optimization parameter and / or generating a second optimization parameter annotation based on the second recommended optimization parameter; The first optimization parameter annotation and / or the second optimization parameter annotation are displayed on a touch screen of the TCS.

9. A TCS fault troubleshooting device based on deep learning, characterized in that: include: A pre-start unit, used to obtain initial selection parameters, and control the simulated TCS to perform pre-start based on the initial selection parameters to generate a pre-start result; A first recommended optimization parameter determination unit, configured to input the initial selection parameters into a target deep learning model that has been fully learned when the pre-start result is a successful start, and obtain first recommended optimization parameters, and control the TCS to start based on the first recommended optimization parameters; a second recommended optimization parameter determination unit, configured to, when the pre-start result is a starting failure, obtain starting failure data, determine a second recommended optimization parameter according to the starting failure data and normal starting data, and control the TCS to start based on the second recommended optimization parameter; The starting fault data includes a plurality of starting fault sub-data, and the normal starting data includes a plurality of normal starting sub-data corresponding to the plurality of starting fault sub-data; and determining the second recommended optimization parameter according to the starting fault data and the normal starting data includes: determining actual starting fault sub-data among the plurality of starting fault sub-data based on a preset troubleshooting sequence and the plurality of normal starting sub-data; A second recommended optimization parameter is determined according to the actual starting failure sub-data and the corresponding normal starting sub-data.

Images