A method and system for inverter failure prediction
By calculating the reliability of the inverter fault prediction model and combining it with the comprehensive reliability judgment, an early warning signal is sent, which solves the problem of low accuracy in inverter fault prediction and achieves more accurate fault prediction and reduced power outage time.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HUANENG NINGXIA ENERGY CO LTD LINGWULONGQIAO BRANCH
- Filing Date
- 2023-07-13
- Publication Date
- 2026-07-21
AI Technical Summary
Existing inverter fault prediction methods are not very accurate, resulting in untimely fault detection and increased unnecessary power outage time.
By calculating the first and second credibility of the fault prediction model and combining the comprehensive credibility to judge the fault prediction result, different early warning signals are sent to improve the prediction accuracy.
This greatly improves the accuracy of inverter fault prediction, reduces unnecessary power outage time, and ensures stable inverter operation.
Smart Images
Figure CN117150219B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of fault prediction technology, and in particular to a method and system for predicting inverter faults. Background Technology
[0002] Inverters, as a crucial component of wind power systems, operate at high frequencies year-round and are prone to malfunctions. Failure to effectively handle such sudden failures can jeopardize the stable operation of the wind power system and endanger lives. Therefore, to ensure the inverters operate reliably and prevent malfunctions and losses, equipment maintenance personnel must be able to monitor the equipment's operation, accurately predict potential future failures, and take preventative measures in advance.
[0003] Existing fault prediction methods are all based on data prediction models, but the results are not always accurate and cannot accurately determine the inverter's operating status, leading to untimely detection of inverter faults and increased unnecessary power outage time.
[0004] Therefore, improving the accuracy of inverter fault prediction is a technical problem that needs to be solved. Summary of the Invention
[0005] To address the problems existing in the prior art, the purpose of this invention is to provide an inverter fault prediction method and system, which improves the accuracy of inverter fault prediction by calculating a first confidence level of the fault prediction model, calculating a second confidence level of the output fault information, and judging the accuracy of the fault prediction result based on the first and second confidence levels.
[0006] To achieve the above objectives, the present invention provides an inverter fault prediction method, the method comprising:
[0007] Collect inverter operating data for a preset time period, preprocess the operating data, and output standardized data;
[0008] The standardized data is input into the trained fault prediction model, and the fault prediction model outputs fault prediction information.
[0009] Calculate the second confidence level of the fault prediction information, and calculate the comprehensive confidence level based on the first confidence level and the second confidence level. If the comprehensive confidence level is not less than a preset confidence level threshold, send an early warning signal. The first confidence level is obtained based on the fault prediction model.
[0010] In some embodiments of this application, the trained fault prediction model includes:
[0011] Acquire multiple sets of historical fault data within a preset time period. Each set of historical fault data is set with corresponding historical fault information, which includes historical fault time, historical fault type, and historical fault probability. The historical fault time is the time required from the start of the fault to the inverter stopping working.
[0012] The historical fault data is preprocessed to output standardized fault data, and the standardized fault data is divided into training set and test set according to a preset ratio.
[0013] The training set is input into a preset fault prediction model to obtain a trained fault prediction model. The test set is input into the fault prediction model, and the fault prediction model outputs fault information. The first confidence level of the fault prediction model is calculated based on the output fault information, wherein the fault information includes fault time, fault type and fault probability.
[0014] In some embodiments of this application, calculating the first confidence level of the fault prediction model based on the output fault information includes:
[0015] The first confidence level of the fault prediction model is calculated based on the fault time, fault type and fault probability in the fault information and the corresponding historical fault time, historical fault type and historical fault probability in the historical fault information.
[0016] Determine the fault type correlation coefficient Q based on the fault type and historical fault type; determine the fault time deviation T and fault time deviation adjustment coefficient W1 based on the fault time and historical fault time; determine the fault probability difference X and fault probability difference adjustment coefficient W2 based on the fault probability and historical fault probability.
[0017] Pre-set a threshold Qmin for the number of contacts, and select the appropriate formula for calculating the first confidence level based on the relationship between the correlation coefficient of the fault type and the threshold of the correlation coefficient;
[0018] When Q≤Qmin, it is directly determined that the first confidence level is less than the preset first confidence level;
[0019] When Q > Qmin, the formula for calculating the first confidence level is:
[0020]
[0021] Wherein, P1 is the first confidence level, A1 is the preset weight of historical fault type, A2 is the preset weight of historical fault time, and A3 is the preset weight of historical fault probability.
[0022] When the first confidence level is less than the preset first confidence level, the number of samples in the training set is expanded, and the fault prediction model continues to be trained until the first confidence level is not less than the preset first confidence level, at which point the fault prediction model stops training.
[0023] In some embodiments of this application, calculating the second confidence level of the fault prediction information includes:
[0024] The fault prediction information includes fault prediction probability H1, fault prediction type H2, and fault prediction time H3.
[0025] The formula for calculating the second level of credibility is:
[0026]
[0027] Where P2 is the second confidence level, α1 is the weight corresponding to the fault prediction probability, α2 is the weight corresponding to the fault prediction type, α3 is the weight corresponding to the fault prediction time, and y is a preset constant.
[0028] In some embodiments of this application, when sending a warning signal if the overall confidence level is not less than a preset confidence level threshold, the following are included:
[0029] The formula for calculating the overall credibility is as follows:
[0030] P0 = (P1 + P2) * t;
[0031] Where P0 is the overall credibility and t is the error coefficient of the overall credibility;
[0032] When P0 is less than the preset confidence threshold, an alarm signal is sent and the operating data of adjacent time periods are collected to perform fault prediction.
[0033] When P0 is not less than the preset confidence threshold, the corresponding early warning signal is selected according to different fault prediction times.
[0034] A first preset fault prediction time T1, a second preset fault prediction time T2, a third preset fault prediction time T3, and a fourth preset fault prediction time T4 are preset, and T1 < T2 < T3 < T4; a first preset warning signal R1, a second preset warning signal R2, a third preset warning signal R3, and a fourth preset warning signal R4 are also preset, and R1 < R2 < R3 < R4.
[0035] The corresponding early warning signal is selected based on the relationship between the current fault prediction time T0 and the preset fault prediction time.
[0036] When T1 < T0 < T2, the fourth preset fault signal R4 is selected as the current warning signal;
[0037] When T2 < T0 < T3, the third preset fault signal R3 is selected as the current warning signal;
[0038] When T3 < T0 < T4, the second preset fault signal R2 is selected as the current warning signal;
[0039] When T4 < T0, the first preset fault signal R1 is selected as the current warning signal.
[0040] In some embodiments of this application, an inverter fault prediction system is also included:
[0041] The processing module is used to collect inverter operating data for a preset time period, preprocess the operating data, and output standardized data. The operating data includes phase voltage, phase current, line voltage, line current, and the temperature of each phase IGBT.
[0042] The prediction module is used to input the standardized data into the trained fault prediction model, and the fault prediction model outputs fault prediction information.
[0043] The calculation module is used to calculate the second credibility of the fault prediction information, calculate the comprehensive credibility based on the first credibility and the second credibility, and send an early warning signal if the comprehensive credibility is not less than a preset credibility threshold. The first credibility is obtained based on the fault prediction model.
[0044] In some embodiments of this application, the fault prediction model includes:
[0045] Acquire multiple sets of historical fault data within a preset time period. Each set of historical fault data is set with corresponding historical fault information, which includes historical fault time, historical fault type, and historical fault probability. The historical fault time is the time required from the start of the fault to the inverter stopping working.
[0046] The historical fault data is preprocessed to output standardized fault data, and the standardized fault data is divided into training set and test set according to a preset ratio.
[0047] The training set is input into a preset fault prediction model to obtain a trained fault prediction model. The test set is input into the fault prediction model, and the fault prediction model outputs fault information. The first confidence level of the fault prediction model is calculated based on the output historical fault information, wherein the fault information includes fault time, fault type and fault probability.
[0048] In some embodiments of this application, the calculation module is further configured to calculate the first confidence level of the fault prediction model, and calculate the first confidence level of the fault prediction model based on the fault time, fault type and fault probability in the fault information and the historical fault time, historical fault type and historical fault probability in the corresponding historical fault information.
[0049] Determine the fault type correlation coefficient Q based on the fault type and historical fault type; determine the fault time deviation T and fault time deviation adjustment coefficient W1 based on the fault time and historical fault time; determine the fault probability difference X and fault probability difference adjustment coefficient W2 based on the fault probability and historical fault probability.
[0050] Pre-set a threshold Qmin for the number of contacts, and select the appropriate formula for calculating the first confidence level based on the relationship between the correlation coefficient of the fault type and the threshold of the correlation coefficient;
[0051] When Q≤Qmin, it is directly determined that the first confidence level is less than the preset first confidence level;
[0052] When Q > Qmin, the formula for calculating the first confidence level is:
[0053]
[0054] Wherein, P1 is the first confidence level, A1 is the preset weight of historical fault type, A2 is the preset weight of historical fault time, and A3 is the preset weight of historical fault probability.
[0055] When the first confidence level is less than the preset first confidence level, the number of samples in the training set is expanded, and the fault prediction model continues to be trained until the first confidence level is not less than the preset first confidence level, at which point the fault prediction model stops training.
[0056] In some embodiments of this application, when the calculation module calculates the second confidence level of the fault prediction information, it includes:
[0057] The fault prediction information includes fault prediction probability H1, fault prediction type H2, and fault prediction time H3.
[0058] The formula for calculating the second level of credibility is:
[0059]
[0060] Where P2 is the second confidence level, α1 is the weight corresponding to the fault prediction probability, α2 is the weight corresponding to the fault prediction type, α3 is the weight corresponding to the fault prediction time, and y is a preset constant.
[0061] In some embodiments of this application, the calculation module is used to calculate a comprehensive credibility based on a first credibility and a second credibility;
[0062] The formula for calculating the overall credibility is as follows:
[0063] P0 = (P1 + P2) * t;
[0064] Where P0 is the overall credibility and t is the error coefficient of the overall credibility;
[0065] The calculation module is used to send an alarm signal when P0 is less than a preset confidence threshold, and to collect the operation data of adjacent time periods to perform fault prediction.
[0066] The calculation module is also used to select a corresponding early warning signal based on different fault prediction times when P0 is not less than a preset confidence threshold.
[0067] The calculation module is pre-set with a first preset fault prediction time T1, a second preset fault prediction time T2, a third preset fault prediction time T3, and a fourth preset fault prediction time T4, and T1 < T2 < T3 < T4; it is also pre-set with a first preset warning signal R1, a second preset warning signal R2, a third preset warning signal R3, and a fourth preset warning signal R4, and R1 < R2 < R3 < R4.
[0068] The corresponding early warning signal is selected based on the relationship between the current fault prediction time T0 and the preset fault prediction time.
[0069] When T1 < T0 < T2, the fourth preset fault signal R4 is selected as the current warning signal;
[0070] When T2 < T0 < T3, the third preset fault signal R3 is selected as the current warning signal;
[0071] When T3 < T0 < T4, the second preset fault signal R2 is selected as the current warning signal;
[0072] When T4 < T0, the first preset fault signal R1 is selected as the current warning signal.
[0073] This invention provides an inverter fault prediction method and system, which has the following advantages compared with the prior art:
[0074] The first confidence level is calculated on the trained fault prediction model, and the second confidence level is calculated on the fault information output by the fault prediction model. The comprehensive confidence level is obtained based on the first and second confidence levels. When the comprehensive confidence level is not less than the preset confidence level threshold, different early warning signals are sent according to the fault prediction time in the fault information. This greatly improves the accuracy of inverter fault prediction, makes accurate judgments on the inverter's operating status, and reduces unnecessary power outage time. Attached Figure Description
[0075] Figure 1 A flowchart illustrating an inverter fault prediction method according to an embodiment of the present invention is shown.
[0076] Figure 2 A schematic diagram of an inverter fault prediction system according to an embodiment of the present invention is shown. Detailed Implementation
[0077] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.
[0078] In the description of this application, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.
[0079] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.
[0080] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.
[0081] The following is a description of preferred embodiments of the present invention in conjunction with the accompanying drawings.
[0082] like Figure 1 As shown, an embodiment of the present invention discloses an inverter fault prediction method, the method comprising:
[0083] Step S101: Collect inverter operation data for a preset time period, preprocess the operation data, and output standardized data;
[0084] Step S102: Input the standardized data into the trained fault prediction model, and the fault prediction model outputs fault prediction information;
[0085] Step S103: Calculate the second credibility of the fault prediction information, calculate the comprehensive credibility based on the first credibility and the second credibility, and send an early warning signal if the comprehensive credibility is not less than a preset credibility threshold. The first credibility is obtained based on the fault prediction model.
[0086] In this embodiment, the inverter's operating data includes phase voltage, phase current, line voltage, line current, and the temperature of each phase IGBT. The preprocessing specifically involves standardizing the operating data and converting the numerical format of the operating data to facilitate subsequent calculations. The fault prediction information includes fault prediction time, fault prediction type, and fault prediction probability.
[0087] In some embodiments of this application, the trained fault prediction model includes:
[0088] Acquire multiple sets of historical fault data within a preset time period. Each set of historical fault data is set with corresponding historical fault information, which includes historical fault time, historical fault type, and historical fault probability. The historical fault time is the time required from the start of the fault to the inverter stopping working.
[0089] The historical fault data is preprocessed to output standardized fault data, and the standardized fault data is divided into training set and test set according to a preset ratio.
[0090] The training set is input into a preset fault prediction model to obtain a trained fault prediction model. The test set is input into the fault prediction model, and the fault prediction model outputs fault information. The first confidence level of the fault prediction model is calculated based on the output fault information, wherein the fault information includes fault time, fault type and fault probability.
[0091] In this embodiment, the training framework of the preset fault prediction model includes a DQN network model. Based on historical fault data, a deep reinforcement learning algorithm is used to train the preset fault prediction model. After standardizing the historical fault data, the historical fault data is uniformly converted in numerical format and divided into a training set and a test set in a 7:3 ratio. The training set is input into the preset fault prediction model to obtain the trained fault prediction model. The test set is input into the fault prediction model to output fault information. The first confidence level of the fault prediction model is calculated based on the fault information.
[0092] In some embodiments of this application, calculating the first confidence level of the fault prediction model based on the output fault information includes:
[0093] The first confidence level of the fault prediction model is calculated based on the fault time, fault type and fault probability in the fault information and the corresponding historical fault time, historical fault type and historical fault probability in the historical fault information.
[0094] Determine the fault type correlation coefficient Q based on the fault type and historical fault type; determine the fault time deviation T and fault time deviation adjustment coefficient W1 based on the fault time and historical fault time; determine the fault probability difference X and fault probability difference adjustment coefficient W2 based on the fault probability and historical fault probability.
[0095] Pre-set a threshold Qmin for the number of contacts, and select the appropriate formula for calculating the first confidence level based on the relationship between the correlation coefficient of the fault type and the threshold of the correlation coefficient;
[0096] When Q≤Qmin, it is directly determined that the first confidence level is less than the preset first confidence level;
[0097] When Q > Qmin, the formula for calculating the first confidence level is:
[0098]
[0099] Wherein, P1 is the first confidence level, A1 is the preset weight of historical fault type, A2 is the preset weight of historical fault time, and A3 is the preset weight of historical fault probability.
[0100] When the first confidence level is less than the preset first confidence level, the number of samples in the training set is expanded, and the fault prediction model continues to be trained until the first confidence level is not less than the preset first confidence level, at which point the fault prediction model stops training.
[0101] In this embodiment, the correlation coefficient of the fault type is obtained according to the correlation parsing model. When the correlation coefficient of the fault type is less than the correlation coefficient threshold, the first confidence of the fault prediction model does not need to be calculated. The fault prediction model is directly trained to expand the training set until the correlation coefficient is not less than the correlation coefficient threshold. The first confidence is calculated according to the above calculation formula. The first confidence and the preset first confidence are compared. When the first confidence is not less than the preset first confidence, the fault prediction model can predict the running data. When the first confidence is less than the preset first confidence, the training set is expanded and the fault prediction model is trained to continue until the first confidence is not less than the preset first confidence.
[0102] In some embodiments of this application, calculating the second confidence level of the fault prediction information includes:
[0103] The fault prediction information includes fault prediction probability H1, fault prediction type H2, and fault prediction time H3.
[0104] The formula for calculating the second level of credibility is:
[0105]
[0106] Where P2 is the second confidence level, α1 is the weight corresponding to the fault prediction probability, α2 is the weight corresponding to the fault prediction type, α3 is the weight corresponding to the fault prediction time, and y is a preset constant.
[0107] In some embodiments of this application, when sending a warning signal if the overall confidence level is not less than a preset confidence level threshold, the following are included:
[0108] The formula for calculating the overall credibility is as follows:
[0109] P0 = (P1 + P2) * t;
[0110] Where P0 is the overall credibility and t is the error coefficient of the overall credibility;
[0111] When P0 is less than the preset confidence threshold, an alarm signal is sent and the operating data of adjacent time periods are collected to perform fault prediction.
[0112] When P0 is not less than the preset confidence threshold, the corresponding early warning signal is selected according to different fault prediction times.
[0113] A first preset fault prediction time T1, a second preset fault prediction time T2, a third preset fault prediction time T3, and a fourth preset fault prediction time T4 are preset, and T1 < T2 < T3 < T4; a first preset warning signal R1, a second preset warning signal R2, a third preset warning signal R3, and a fourth preset warning signal R4 are also preset, and R1 < R2 < R3 < R4.
[0114] The corresponding early warning signal is selected based on the relationship between the current fault prediction time T0 and the preset fault prediction time.
[0115] When T1 < T0 < T2, the fourth preset fault signal R4 is selected as the current warning signal;
[0116] When T2 < T0 < T3, the third preset fault signal R3 is selected as the current warning signal;
[0117] When T3 < T0 < T4, the second preset fault signal R2 is selected as the current warning signal;
[0118] When T4 < T0, the first preset fault signal R1 is selected as the current warning signal.
[0119] In this embodiment, when the overall confidence level is less than a preset confidence level threshold, the operation data of adjacent time periods within a preset time period are collected for prediction. When the overall confidence level is not less than the preset confidence level threshold, different early warning signals are selected based on the fault prediction time. When the fault prediction time is shorter, a more urgent early warning signal is sent.
[0120] In some embodiments of this application, an inverter fault prediction system is also included:
[0121] The processing module is used to collect inverter operating data for a preset time period, preprocess the operating data, and output standardized data.
[0122] The prediction module is used to input the standardized data into the trained fault prediction model, and the fault prediction model outputs fault prediction information.
[0123] The calculation module is used to calculate the second credibility of the fault prediction information, calculate the comprehensive credibility based on the first credibility and the second credibility, and send an early warning signal if the comprehensive credibility is not less than a preset credibility threshold. The first credibility is obtained based on the fault prediction model.
[0124] In some embodiments of this application, the fault prediction model includes:
[0125] Acquire multiple sets of historical fault data within a preset time period. Each set of historical fault data is set with corresponding historical fault information, which includes historical fault time, historical fault type, and historical fault probability. The historical fault time is the time required from the start of the fault to the inverter stopping working.
[0126] The historical fault data is preprocessed to output standardized fault data, and the standardized fault data is divided into training set and test set according to a preset ratio.
[0127] The training set is input into a preset fault prediction model to obtain a trained fault prediction model. The test set is input into the fault prediction model, and the fault prediction model outputs fault information. The first confidence level of the fault prediction model is calculated based on the output historical fault information, wherein the fault information includes fault time, fault type and fault probability.
[0128] In some embodiments of this application, the calculation module is further configured to calculate the first confidence level of the fault prediction model, and calculate the first confidence level of the fault prediction model based on the fault time, fault type and fault probability in the fault information and the historical fault time, historical fault type and historical fault probability in the corresponding historical fault information.
[0129] Determine the fault type correlation coefficient Q based on the fault type and historical fault type; determine the fault time deviation T and fault time deviation adjustment coefficient W1 based on the fault time and historical fault time; determine the fault probability difference X and fault probability difference adjustment coefficient W2 based on the fault probability and historical fault probability.
[0130] Pre-set a threshold Qmin for the number of contacts, and select the appropriate formula for calculating the first confidence level based on the relationship between the correlation coefficient of the fault type and the threshold of the correlation coefficient;
[0131] When Q≤Qmin, it is directly determined that the first confidence level is less than the preset first confidence level;
[0132] When Q > Qmin, the formula for calculating the first confidence level is:
[0133]
[0134] Wherein, P1 is the first confidence level, A1 is the preset weight of historical fault type, A2 is the preset weight of historical fault time, and A3 is the preset weight of historical fault probability.
[0135] When the first confidence level is less than the preset first confidence level, the number of samples in the training set is expanded, and the fault prediction model continues to be trained until the first confidence level is not less than the preset first confidence level, at which point the fault prediction model stops training.
[0136] In some embodiments of this application, when the calculation module calculates the second confidence level of the fault prediction information, it includes:
[0137] The fault prediction information includes fault prediction probability H1, fault prediction type H2, and fault prediction time H3.
[0138] The formula for calculating the second level of credibility is:
[0139]
[0140] Where P2 is the second confidence level, α1 is the weight corresponding to the fault prediction probability, α2 is the weight corresponding to the fault prediction type, α3 is the weight corresponding to the fault prediction time, and y is a preset constant.
[0141] In some embodiments of this application, the calculation module is used to calculate a comprehensive credibility based on a first credibility and a second credibility;
[0142] The formula for calculating the overall credibility is as follows:
[0143] P0 = (P1 + P2) * t;
[0144] Where P0 is the overall credibility and t is the error coefficient of the overall credibility;
[0145] The calculation module is used to send an alarm signal when P0 is less than a preset confidence threshold, and to collect the operation data of adjacent time periods to perform fault prediction.
[0146] The calculation module is also used to select a corresponding early warning signal based on different fault prediction times when P0 is not less than a preset confidence threshold.
[0147] The calculation module is pre-set with a first preset fault prediction time T1, a second preset fault prediction time T2, a third preset fault prediction time T3, and a fourth preset fault prediction time T4, and T1 < T2 < T3 < T4; it is also pre-set with a first preset warning signal R1, a second preset warning signal R2, a third preset warning signal R3, and a fourth preset warning signal R4, and R1 < R2 < R3 < R4.
[0148] The corresponding early warning signal is selected based on the relationship between the current fault prediction time T0 and the preset fault prediction time.
[0149] When T1 < T0 < T2, the fourth preset fault signal R4 is selected as the current warning signal;
[0150] When T2 < T0 < T3, the third preset fault signal R3 is selected as the current warning signal;
[0151] When T3 < T0 < T4, the second preset fault signal R2 is selected as the current warning signal;
[0152] When T4 < T0, the first preset fault signal R1 is selected as the current warning signal.
[0153] In summary, this invention discloses an inverter fault prediction method and system. The method includes: step S101: collecting inverter operating data for a preset time period, preprocessing the operating data, and outputting standardized data; step S102: inputting the standardized data into a trained fault prediction model, which outputs fault prediction information; step S103: calculating the second confidence level of the fault prediction information, calculating the comprehensive confidence level based on the first and second confidence levels, and sending an early warning signal if the comprehensive confidence level is not less than a preset confidence level threshold. The first confidence level is obtained based on the fault prediction model. This invention calculates the first confidence level on the trained fault prediction model, calculates the second confidence level on the fault information output by the fault prediction model, and obtains the comprehensive confidence level based on the first and second confidence levels. When the comprehensive confidence level is not less than a preset confidence level threshold, different early warning signals are sent based on the fault prediction time in the fault information, greatly improving the accuracy of inverter fault prediction, accurately judging the inverter's operating status, and reducing unnecessary power outage time.
[0154] In the description of the above embodiments, specific features, structures, materials, or characteristics may be combined in any suitable manner in one or more embodiments or examples.
[0155] Although the invention has been described above with reference to embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of the invention. In particular, features in the embodiments disclosed herein can be combined with each other in any manner, provided there is no structural conflict. The omission of all such combinations in this specification is merely for brevity and resource conservation. Therefore, the invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.
[0156] It will be understood by those skilled in the art that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for predicting inverter faults, characterized in that, include: Collect inverter operating data for a preset time period, preprocess the operating data, and output standardized data; The standardized data is input into the trained fault prediction model, and the fault prediction model outputs fault prediction information. Calculate the second confidence level of the fault prediction information, calculate the comprehensive confidence level based on the first confidence level and the second confidence level, and send an early warning signal if the comprehensive confidence level is not less than a preset confidence level threshold. The first confidence level is obtained based on the fault prediction model. The trained fault prediction model includes: Acquire multiple sets of historical fault data within a preset time period. Each set of historical fault data is set with corresponding historical fault information, which includes historical fault time, historical fault type, and historical fault probability. The historical fault time is the time required from the start of the fault to the inverter stopping working. The historical fault data is preprocessed to output standardized fault data, and the standardized fault data is divided into training set and test set according to a preset ratio. The training set is input into a preset fault prediction model to obtain a trained fault prediction model. The test set is input into the fault prediction model, and the fault prediction model outputs fault information. The first confidence level of the fault prediction model is calculated based on the output fault information, wherein the fault information includes fault time, fault type and fault probability. When calculating the first confidence level of the fault prediction model based on the output fault information, the following is included: The first confidence level of the fault prediction model is calculated based on the fault time, fault type and fault probability in the fault information and the corresponding historical fault time, historical fault type and historical fault probability in the historical fault information. Determine the fault type correlation coefficient Q based on the fault type and historical fault type; determine the fault time deviation T and fault time deviation adjustment coefficient W1 based on the fault time and historical fault time; determine the fault probability difference X and fault probability difference adjustment coefficient W2 based on the fault probability and historical fault probability. Pre-set a threshold Qmin for the number of contacts, and select the appropriate formula for calculating the first confidence level based on the relationship between the correlation coefficient of the fault type and the threshold of the correlation coefficient; When Q≤Qmin, it is directly determined that the first confidence level is less than the preset first confidence level; When Q > Qmin, the formula for calculating the first confidence level is: Wherein, P1 is the first confidence level, A1 is the preset weight of historical fault type, A2 is the preset weight of historical fault time, and A3 is the preset weight of historical fault probability. When the first confidence level is less than the preset first confidence level, the number of samples in the training set is expanded, and the fault prediction model continues to be trained until the first confidence level is not less than the preset first confidence level, at which point the fault prediction model stops training. Calculating the second confidence level of the fault prediction information includes: The fault prediction information includes fault prediction probability H1, fault prediction type H2, and fault prediction time H3. The formula for calculating the second level of credibility is: P2 represents the second level of confidence. The weights corresponding to the fault prediction probabilities. 2 represents the weight corresponding to the fault prediction type. 3 represents the weight corresponding to the fault prediction time, and y is a preset constant.
2. The inverter fault prediction method according to claim 1, characterized in that, If the overall credibility is not less than the preset credibility threshold, when sending a warning signal, it includes: The formula for calculating the overall credibility is as follows: P0=(P1+P2) t; Where P0 is the overall credibility and t is the error coefficient of the overall credibility; When P0 is less than the preset confidence threshold, an alarm signal is sent and the operating data of adjacent time periods are collected to perform fault prediction. When P0 is not less than the preset confidence threshold, the corresponding early warning signal is selected according to different fault prediction times. A first preset fault prediction time T1, a second preset fault prediction time T2, a third preset fault prediction time T3, and a fourth preset fault prediction time T4 are preset, and T1 < T2 < T3 < T4; a first preset warning signal R1, a second preset warning signal R2, a third preset warning signal R3, and a fourth preset warning signal R4 are also preset, and R1 < R2 < R3 < R4. The corresponding early warning signal is selected based on the relationship between the current fault prediction time T0 and the preset fault prediction time. When T1 < T0 < T2, the fourth preset warning signal R4 is selected as the current warning signal; When T2 < T0 < T3, the third preset warning signal R3 is selected as the current warning signal; When T3 < T0 < T4, the second preset warning signal R2 is selected as the current warning signal; When T4 < T0, the first preset warning signal R1 is selected as the current warning signal.
3. An inverter fault prediction system, characterized in that, include: The processing module is used to collect inverter operating data for a preset time period, preprocess the operating data, and output standardized data. The operating data includes phase voltage, phase current, line voltage, line current, and the temperature of each phase IGBT. The prediction module is used to input the standardized data into the trained fault prediction model, and the fault prediction model outputs fault prediction information. The calculation module is used to calculate the second credibility of the fault prediction information, calculate the comprehensive credibility based on the first credibility and the second credibility, and send an early warning signal if the comprehensive credibility is not less than a preset credibility threshold. The first credibility is obtained according to the fault prediction model. The fault prediction model includes: Acquire multiple sets of historical fault data within a preset time period. Each set of historical fault data is set with corresponding historical fault information, which includes historical fault time, historical fault type, and historical fault probability. The historical fault time is the time required from the start of the fault to the inverter stopping working. The historical fault data is preprocessed to output standardized fault data, and the standardized fault data is divided into training set and test set according to a preset ratio. The training set is input into a preset fault prediction model to obtain a trained fault prediction model. The test set is input into the fault prediction model, and the fault prediction model outputs fault information. The first confidence level of the fault prediction model is calculated based on the output historical fault information, wherein the fault information includes fault time, fault type and fault probability. The calculation module is also used to calculate the first confidence level of the fault prediction model, and to calculate the first confidence level of the fault prediction model based on the fault time, fault type and fault probability in the fault information and the historical fault time, historical fault type and historical fault probability in the corresponding historical fault information. Determine the fault type correlation coefficient Q based on the fault type and historical fault type; determine the fault time deviation T and fault time deviation adjustment coefficient W1 based on the fault time and historical fault time; determine the fault probability difference X and fault probability difference adjustment coefficient W2 based on the fault probability and historical fault probability. Pre-set a threshold Qmin for the number of contacts, and select the appropriate formula for calculating the first confidence level based on the relationship between the correlation coefficient of the fault type and the threshold of the correlation coefficient; When Q≤Qmin, it is directly determined that the first confidence level is less than the preset first confidence level; When Q > Qmin, the formula for calculating the first confidence level is: Wherein, P1 is the first confidence level, A1 is the preset weight of historical fault type, A2 is the preset weight of historical fault time, and A3 is the preset weight of historical fault probability. When the first confidence level is less than the preset first confidence level, the number of samples in the training set is expanded, and the fault prediction model continues to be trained until the first confidence level is not less than the preset first confidence level, at which point the fault prediction model stops training. When the calculation module calculates the second confidence level of the fault prediction information, it includes: The fault prediction information includes fault prediction probability H1, fault prediction type H2, and fault prediction time H3. The formula for calculating the second level of credibility is: P2 represents the second level of confidence. The weights corresponding to the fault prediction probabilities. 2 represents the weight corresponding to the fault prediction type. 3 represents the weight corresponding to the fault prediction time, and y is a preset constant.
4. The inverter fault prediction system according to claim 3, characterized in that, The calculation module is used to calculate the overall credibility based on the first credibility and the second credibility. The formula for calculating the overall credibility is as follows: P0=(P1+P2) t; Where P0 is the overall credibility and t is the error coefficient of the overall credibility; The calculation module is used to send an alarm signal when P0 is less than a preset confidence threshold, and to collect the operation data of adjacent time periods to perform fault prediction. The calculation module is also used to select a corresponding early warning signal based on different fault prediction times when P0 is not less than a preset confidence threshold. The calculation module is pre-set with a first preset fault prediction time T1, a second preset fault prediction time T2, a third preset fault prediction time T3, and a fourth preset fault prediction time T4, and T1 < T2 < T3 < T4; it is also pre-set with a first preset warning signal R1, a second preset warning signal R2, a third preset warning signal R3, and a fourth preset warning signal R4, and R1 < R2 < R3 < R4. The corresponding early warning signal is selected based on the relationship between the current fault prediction time T0 and the preset fault prediction time. When T1 < T0 < T2, the fourth preset warning signal R4 is selected as the current warning signal; When T2 < T0 < T3, the third preset warning signal R3 is selected as the current warning signal; When T3 < T0 < T4, the second preset warning signal R2 is selected as the current warning signal; When T4 < T0, the first preset warning signal R1 is selected as the current warning signal.