A method for identifying abnormalities during the startup process of a wind turbine generator set

Abstract

The present invention discloses an abnormality identification algorithm for the start-up process of a wind turbine generator set. The start-up control method obtains the current operating state of the wind turbine generator set. If the current operating state is the standby state, it is determined whether the average speed of the wind turbine generator set in the first predetermined time period is greater than or equal to the current start-up speed, and the current start-up speed is determined by the historical operating data of the wind turbine generator set; if the average speed is greater than or equal to the current start-up speed, the generator of the wind turbine generator set is controlled to start and run. The technical solution in the embodiment of the present invention can be used to adaptively determine the start-up parameters of the set based on the historical operating data of the wind turbine generator set, and control the start-up operation of the wind turbine generator set according to the start-up parameters. The present invention uses the start-up time judgment and the external wind speed judgment to identify abnormalities, and uses two methods of offline analysis and online adjustment to effectively identify and customize parameters to optimize each unit.

Description

Technical Field

[0001] The present invention relates to an abnormal problem identification algorithm and a corresponding adjustment method applied to a variable pitch wind turbine generator set in the field of new energy, and in particular to an abnormality identification method during the start-up process of a wind turbine generator set. Background Art

[0002] Wind turbines are generally divided into several states: standby, start-up, operation, shutdown and maintenance. Under normal conditions, when a wind turbine encounters a fault or the external wind conditions are too weak to generate electricity, the shutdown strategy will be activated to switch the wind turbine from the operating state to the shutdown state. When the fault is eliminated or the external wind conditions meet the start-up requirements, the startup strategy will be activated to switch the wind turbine from the shutdown or standby state to the startup state, and then from the startup state to the power generation state when the power generation conditions are met.

[0003] However, many problems may be encountered during the startup process. For example, there is no way to identify the phenomenon that the startup time is too long or too short due to individual differences between units or unreasonable parameter settings in the startup strategy. The main control usually believes that this is due to wind conditions. For example, the wind speed is too low, etc., which causes the energy absorbed by the unit to fail to start normally and continue to be in a state of waiting for wind. Therefore, it is impossible to identify the abnormal situation that the unit should be started quickly to generate electricity but is still in the startup process.

[0004] When the unit encounters a fault during the startup process, the wind turbine control system will provide protection, but this fault protection mainly checks whether each executive component is normal. For example, during the startup process, check whether the converter system, pitch system, yaw and other subsystems are in normal condition. If a fault occurs, shut down for protection immediately. However, there is no identification method or protection measure for the startup performance of the entire unit: such as whether the startup time is too long or whether it has not been able to start normally under the same operating conditions. At this time, the wind turbine control system will only think that the wind condition of the unit is too small to meet the normal startup conditions and continue to wait. This will lead to power loss due to slow startup time or failure to start normally due to differences in units or unreasonable parameter settings. Summary of the invention

[0005] The present invention effectively identifies abnormal behaviors during the startup process of the unit through horizontal and vertical comparison of big data, and finds that there are usually several reasons for the startup time, one is the external wind conditions, the second is the individual differences of the unit, and the third is the control parameter setting problem. And through big data analysis, the abnormal startup time caused by wind conditions is eliminated, and the problem units that can be solved by control parameters are screened out, and then the control parameters of the startup process of each unit are customized to optimize, so that the unit can start normally and smoothly to avoid the loss of power generation.

[0006] The present invention adopts the following technical solution:

[0007] The technical solution is divided into four steps: data preparation, model calculation, abnormal identification and control adjustment. The first three steps are completed offline, which does not affect the operation of the fan. The last step is to formulate and complete the control parameter adjustment method and send it to the site to optimize the online operation of the unit. Data preparation work is carried out, data collection, storage and cleaning are carried out, and then model calculation is carried out on these data. After the calculation is completed, the abnormal values ​​are extracted and judged, and finally the control adjustment is carried out according to the abnormal situation.

[0008] The data preparation includes the following steps:

[0009] 1) Collect and transmit the operating data of the target wind farm, requiring no less than one month of data, which can fully experience various operating conditions and external environmental wind conditions to make the data more effective;

[0010] 2) The returned data is placed on the cloud platform for storage according to the specified location and number;

[0011] 3) Data cleaning work, including deletion of duplicate values, supplementation of missing values, and elimination of interrupted data.

[0012] The model calculation includes the following steps:

[0013] 1) Identify the startup status of each unit data. The operation data channel contains the startup status record. Normal units are divided into several states: standby, startup, power generation, shutdown and maintenance. The data marked as startup needs to be identified;

[0014] 2) Calculate the duration of each startup process Tsij of all units in the site (i=1,2...N, i represents the unit number)

[0015] (j=1,2...M, j represents the number of startups, note that the total number of startups for each unit is different). For example, for unit 1, a total of 100 startups are identified, and the end time of each startup - the start time of each startup is the startup time. That is, 100 duration values ​​of unit 1 are obtained, Ts11, Ts12, Ts13...Ts1 100 , if the 2# unit has 103 starts, then we get Ts21, Ts22, Ts23…Ts2 103 Similarly, the startup time of each unit of the site can be obtained. Assuming that the last unit is started 89 times, the total number of startups of 50 units is 5000 times, then the total number of Tsij is 100+103+…+89=5000 times;

[0016] 3) Sort the startup time Tsij of all units on site, remove the x% maximum value and X% minimum value, and turn it into Tsij'. Assume Ts11 = 40s, Ts12 = 2000s, ... Ts1 100 =70s, Ts21=60s, Ts12=80s. . . Sort the 5000 Tsij values, find the maximum and minimum values ​​that account for 1% (here x is assumed to be 1%) and filter them out. After filtering out, 4900 data remain;

[0017] 4) Calculate the average startup time of each unit and the average startup time of the whole field based on these 4900 data. Since the abnormal data has been screened out in step 3), the startup times of each unit have changed. For example, the startup time of unit 1# is Ts11, Ts13…Ts1 100 , because Ts12 is too large and has been screened out, the startup duration value of unit 1# is less than 100, but only 99. Then, the average startup duration value of unit 1# is taken, and Ts1avg = (Ts11 + Ts13 + Ts14 ... + Ts1 100 ) / 99, and similarly, the average startup time of other units is obtained, Ts2avg, Ts3avg, ... Ts50avg, and 50 values ​​are obtained, each representing the average startup time of a unit, and then the average value of these 4900 Tsij is taken to obtain the average startup time information Tsavg of the whole field;

[0018] 5) Calculate the wind speed information during the startup process based on these 4900 data and obtain 4900 Vij. Since the abnormal data has been screened out in step 3, the startup times of each unit have changed. For example, the startup wind speeds of unit 1# are V11, V13, V14, ... V1 100 , because Ts12 is too large and has been screened out, the wind speed value of unit 1# is less than 100, but only 99. Then take the average wind speed value of unit 1# and get V1avg = (V11 + V13 + V14 ... + V1 100 ) / 99, and similarly, get the average wind speeds V2avg, V3avg…V50avg of other units, and get 50 values, each representing the average startup wind speed of a unit. Then take the average value of these 4900 Vij to get the average startup wind speed information Vsavg of the whole site.

[0019] Abnormal identification includes that the startup status identification of abnormal units requires two conditions to be judged, and it is considered abnormal only when both conditions are met at the same time;

[0020] 1) Condition 1: Compare the average start-up duration of each unit and the全场 average start-up duration obtained from the above calculation results, and find the units that exceed the threshold. Exceeding the threshold means that the average start-up duration is too long or too short:

[0021] That is, TsDi = Tsiavg - Tsavg, TsDi > Ts1 or TsDi < -Ts2, where Ts1 is the upper limit duration difference and Ts2 is the lower limit duration difference. This step is to screen out the units with too long or too short start-up duration;

[0022] 2) Condition 2: Further screen the units that meet Condition 1. Compare the average start-up wind speed of each unit and the全场 average start-up wind speed obtained from the above calculation results, and determine that the wind speed distribution during the start-up process of the abnormal unit is within the normal range: That is, VDi = Viavg - Vavg, VDi < V1 and VDi > -V2, where V1 is the upper limit wind speed difference and V2 is the lower limit wind speed difference. This step eliminates the units with abnormal corresponding wind speeds among the units with too long or too short start-up duration, excluding the start-up anomalies caused by wind speed reasons.

[0023] The control adjustment includes the following steps:

[0024] 1) For the units that meet Condition 1 and Condition 2, adjust the start-up parameters. The adjustment method is mainly to adjust the start-up parameters according to TsDi, such as the start-up set speed, the initial pitch angle, and the pitch rate during the start-up process; for example, when TsDi is too large, it means that the start-up time is too long, and the initial pitch angle can be reduced, and the pitch rate during the start-up process can be increased to quickly open the pitch angle, increase the absorption of wind energy, and promote quick start-up;

[0025] Update the adjusted parameters to the unit control system for online optimization implementation;

[0026] Judge the start-up duration of the unit, find out the units with too long start-up duration, and reduce the start-up duration by changing the initial pitch angle and pitch rate during the start-up process to achieve stable and quick start-up.

[0027] The present invention judges the start-up abnormal mechanism, eliminates the start-up anomalies caused by data problems and external wind speed problems through a special algorithm, and locates the start-up anomalies caused by unit individual differences and control problems. Description of the Drawings

[0028] Figure 1 It is a flowchart of the present invention. Detailed Embodiment

[0029] The present invention is divided into four steps: data preparation, model calculation, anomaly identification, and control adjustment. The data preparation includes the following steps:

[0030] 1) Collect and transmit the operating data of the target wind farm, requiring no less than one month of data, which can fully experience various operating conditions and external environmental wind conditions to make the data more effective;

[0031] 2) The returned data is placed on the cloud platform for storage according to the specified location and number;

[0032] 3) Data cleaning work, including deletion of duplicate values, supplementation of missing values, and elimination of interrupted data.

[0033] The model calculation includes the following steps:

[0034] 1) Identify the startup status of each unit data. The operation data channel contains the startup status record. Normal units are divided into several states: standby, startup, power generation, shutdown and maintenance. The data marked as startup needs to be identified;

[0035] 2) Calculate the duration of each startup process Tsij (i=1,2...N, i represents the unit number) (j=1,2...M, j represents the number of startups, note that the total number of startups for each unit is different). For example, for unit 1, a total of 100 startups are identified, and the end time of each startup - the start time of each startup is the startup time. That is, 100 duration values ​​of unit 1 are obtained, Ts11, Ts12, Ts13...Ts1 100 , if the 2# unit has 103 starts, then we get Ts21, Ts22, Ts23…Ts2 103 Similarly, the startup time of each unit of the site can be obtained. Assuming that the last unit is started 89 times, the total number of startups of 50 units is 5000 times, then the total number of Tsij is 100+103+…+89=5000 times;

[0036] 3) Sort the startup time Tsij of all units on site, remove the x% maximum value and X% minimum value, and turn it into Tsij'. Assume Ts11 = 40s, Ts12 = 2000s, ... Ts1 100 =70s, Ts21=60s, Ts12=80s. . . Sort the 5000 Tsij values, find the maximum and minimum values ​​that account for 1% (here x is assumed to be 1%) and filter them out. After filtering out, 4900 data remain;

[0037] 4) Calculate the average startup time of each unit and the average startup time of the whole field based on these 4900 data. Since the abnormal data has been screened out in step 3), the startup times of each unit have changed. For example, the startup time of unit 1# is Ts11, Ts13…Ts1 100, because Ts12 is too large and has been screened out, the startup duration value of unit 1# is less than 100, but only 99. Then, the average startup duration value of unit 1# is taken, and Ts1avg = (Ts11 + Ts13 + Ts14 ... + Ts1 100 ) / 99, and similarly, the average startup time of other units is obtained, Ts2avg, Ts3avg, ... Ts50avg, and 50 values ​​are obtained, each representing the average startup time of a unit, and then the average value of these 4900 Tsij is taken to obtain the average startup time information Tsavg of the whole field;

[0038] 5) Calculate the wind speed information during the startup process based on these 4900 data and obtain 4900 Vij. Since the abnormal data has been screened out in step 3, the startup times of each unit have changed. For example, the startup wind speeds of unit 1# are V11, V13, V14, ... V1 100 , because Ts12 is too large and has been screened out, the wind speed value of unit 1# is less than 100, but only 99. Then take the average wind speed value of unit 1# and get V1avg = (V11 + V13 + V14 ... + V1 100 ) / 99, and similarly, get the average wind speeds V2avg, V3avg…V50avg of other units, and get 50 values, each representing the average startup wind speed of a unit. Then take the average value of these 4900 Vij to get the average startup wind speed information Vsavg of the whole site.

[0039] Abnormal identification includes that the startup status identification of abnormal units requires two conditions to be judged, and it is considered abnormal only when both conditions are met at the same time;

[0040] 1) Condition 1: Compare the average startup time of each unit obtained from the above calculation results with the average startup time of the entire field, and find out the units that exceed the threshold. Exceeding the threshold means that the average startup time of the units is too long or too short:

[0041] That is, TsDi=Tsiavg-Tsavg, TsDi>Ts1 or TsDi<-Ts2. Ts1 is the upper limit of the time difference, and Ts2 is the lower limit of the time difference. This step is to screen out units with too long or too short startup time;

[0042] 2) Condition 2: Further screen the units that meet Condition 1. Compare the average start-up wind speed of each unit obtained from the above calculation results with the average start-up wind speed of the whole field to determine that the wind speed distribution of the abnormal units during the start-up process is within the normal range: that is, VDi = Viavg - Vavg, VDi < V1 and VDi > -V2, where V1 is the upper limit wind speed difference and V2 is the lower limit wind speed difference. This step eliminates the units with abnormal corresponding wind speeds among the units with too long or too short start-up durations, excluding the start-up anomalies caused by wind speed reasons.

[0043] The control adjustment includes the following steps:

[0044] 1) For the units that meet Condition 1 and Condition 2, adjust the start-up parameters. The adjustment method is mainly to adjust the start-up parameters according to TsDi, such as the start-up set speed, the initial pitch angle, and the pitch rate during the start-up process; for example, when TsDi is too large, it means that the start-up time is too long. The initial pitch angle can be reduced, and the pitch rate during the start-up process can be increased to quickly open the pitch angle, increase the absorption of wind energy, and promote quick start-up.

[0045] Update the adjusted parameters to the unit control system for online optimization implementation;

[0046] Judge the start-up duration of the unit, find out the units with too long start-up duration, and reduce the start-up duration by changing the initial pitch angle and pitch rate during the start-up process to achieve smooth and quick start-up.

[0047] Transmit the historical operation data of the fan back to the cloud big data platform. First, through data cleaning and algorithm development, make horizontal and vertical comparisons of the time lengths of the start-up processes of the whole field of fans during the historical operation time, effectively identify the differences in the start-up processes of different units, find out the abnormal units (referring to the units with too long or too short start-up processes) during the start-up process, analyze them, and exclude the start-up problems caused by wind conditions through algorithms, locate the units that can be solved through control optimization, customize different start-up parameters for each optimizable unit, and update the control parameters to the controller of each unit (PLC - Programmable Logic Controller). The new control parameters can optimize the duration of the start-up process of the unit, enable the unit to start up normally and smoothly, and avoid power generation losses.

Claims

1. A method for identifying abnormalities during the startup process of a wind turbine generator set, characterized in that: It includes data preparation, model calculation, abnormal identification and control adjustment. The first three steps are completed offline, which does not affect the operation of the fan. The last step is to issue the control parameter adjustment method to the site after it is formulated to optimize the online operation of the unit. First, data preparation is carried out, data collection, storage and cleaning are carried out, model calculation is carried out for these data, abnormal values ​​are extracted and judged after the calculation is completed, and finally control adjustment is carried out according to the abnormal situation. Data preparation includes the following steps: 1) Collect and transmit the operating data of the target wind farm, requiring no less than one month of data, which can fully experience various operating conditions and external environmental wind conditions to make the data more effective; 2) The returned data is placed on the cloud platform for storage according to the specified location and number; 3) Data cleaning, including deletion of duplicate values, supplementation of missing values, and elimination of interrupted data; The model calculation includes the following steps: 1) Identify the startup status of each unit data. The operation data channel contains the startup status record. Normal units are divided into several states: standby, startup, power generation, shutdown and maintenance. The data marked as startup needs to be identified; 2) Calculate the duration of each startup process Tsij of all units in the site (i=1,2...N, i represents the unit number) (j=1,2...M, j represents the number of startups, note that the total number of startups for each unit is different). For example, for unit 1, a total of 100 startups are identified, and the end time of each startup - the start time of the startup is the startup time, that is, 100 duration values ​​of unit 1 are obtained, Ts11, Ts12, Ts13...Ts1 100 , if the 2# unit has 103 starts, then we get Ts21, Ts22, Ts23…Ts2 103 Similarly, the startup time of each unit of the site can be obtained. Assuming that the last unit is started 89 times, the total number of startups of 50 units is 5000 times, then the total number of Tsij is 100+103+…+89=5000 times; 3) Sort the startup time Tsij of all units on site, remove the maximum value of x% and the minimum value of x%, and become Tsij', assuming Ts11 = 40s, Ts12 = 2000s, ... Ts1 100 =70s, Ts21=60s, Ts12=80s... 5000 Tsij values ​​are sorted to find the maximum and minimum values ​​that account for 1%. Here, x is assumed to be 1% and is screened out. After screening, 4900 data remain; 4) Calculate the average startup time of each unit and the average startup time of the entire site based on these 4900 data. Since the abnormal data has been filtered out in step 3), the number of startups of each unit has changed. For example, the startup time of unit 1# is Ts11, Ts13…Ts1 100 , because Ts12 is too large and has been screened out, the startup duration value of unit 1# is less than 100, but only 99. Then, the average startup duration value of unit 1# is taken, and Ts1avg = (Ts11 + Ts13 + Ts14 ... + Ts1 100 ) / 99, and similarly, we get the average startup time of other units Ts2avg, Ts3avg, ... Ts50avg, and get 50 values, each representing the average startup time of a unit Then take the average value of these 4900 Tsij to get the average startup time information Tsavg of the whole field; 5) Calculate the wind speed information during the startup process based on these 4900 data and obtain 4900 Vij. Since the abnormal data has been screened out in step 3, the startup times of each unit have changed. For example, the startup wind speeds of unit 1# are V11, V13, V14, ... V1 100 , because Ts12 is too large and has been screened out, the wind speed value of unit 1# is less than 100, but only 99. Then the average wind speed value of unit 1# is taken, and the result is V1avg=(V11+V13+V14……+V1 100 ) / 99, and similarly, the average wind speed V2avg of other units is obtained. V3avg…V50avg, get 50 values, each representing the average startup wind speed of a unit, and then take the average value of these 4900 Vij to get the average startup wind speed information Vsavg of the whole site; Abnormal identification includes that the startup status identification of abnormal units requires two conditions to be judged, and it is considered abnormal only when both conditions are met at the same time; 1) Condition 1: Compare the average startup time of each unit obtained from the above calculation results with the average startup time of the entire field, and find out the units that exceed the threshold. Exceeding the threshold means that the average startup time of the units is too long or too short: That is, TsDi=Tsiavg-Tsavg, TsDi>Ts1 or TsDi<-Ts2, where Ts1 is the upper limit time difference and Ts2 is the lower limit time difference. This step is to screen out units with too long or too short startup time; 2) Condition 2: Further screen the units that meet Condition 1. Compare the average start-up wind speed of each unit obtained from the above calculation results with the全场平均启机风速 (the average start-up wind speed of the whole field), and determine that the wind speed distribution of the abnormal unit during the start-up process is within the normal range: that is, VDi = Viavg - Vavg, VDi < V1 and VDi > -V2, where V1 is the upper limit wind speed difference and V2 is the lower limit wind speed difference. This step eliminates the units with abnormal corresponding wind speeds among the units with too long or too short start-up durations, excluding the start-up abnormalities caused by wind speed reasons; The control adjustment includes the following steps: 1) Adjust the start-up parameters for the units that meet Condition 1 and Condition 2. The adjustment method is mainly to adjust the start-up parameters according to TsDi, such as the start-up set speed, the initial pitch angle, and the pitch rate during the start-up process; for example, when TsDi is too large, it means that the start-up time is too long, and the initial pitch angle can be reduced, and the pitch rate during the start-up process can be increased to quickly open the pitch angle, increase the absorption of wind energy, and promote a quick start-up; Update the adjusted parameters to the unit control system for online optimization implementation; 2) Judge the start-up duration of the unit, find out the units with too long start-up duration, and reduce the start-up duration by changing the initial pitch angle and pitch rate during the start-up process to achieve a smooth and quick start-up.

Images