Centrifugal compressor surge line online updating and abnormal value monitoring method

Abstract

The present application belongs to the technical field of centrifugal compressor control, and relates to a centrifugal compressor surge line online updating and abnormal value monitoring method, comprising: S1, starting the centrifugal compressor; S2, establishing a theoretical surge line; S3, obtaining a surge point in a partition; S4, calculating an equivalent distance D; S5, calculating the standard deviation of the equivalent distance of all surge points σ ; S6, judging whether the standard deviation of the equivalent distance is greater than a threshold value θ, and if not, returning to step S3; if yes, executing S7; S7, judging whether the number of recorded surge points is greater than m, and if yes, comparing the equivalent distance D of the surge points to the theoretical surge line, eliminating the surge point with the largest equivalent distance D, and establishing a surge line; if not, returning to step S3 for continuous monitoring. The present application can eliminate the abnormal value of the surge point in the process of establishing the surge line, establish an accurate surge line, and update the surge line in real time.

Description

Technical Field

[0001] This invention relates to a method for online updating of surge lines and monitoring of abnormal values ​​in centrifugal compressors, belonging to the field of control technology for centrifugal compressors. Background Technology

[0002] Surge in a centrifugal compressor refers to the sudden, intense vibrations that sometimes occur during operation, accompanied by significant fluctuations in the flow and pressure of the gas medium, a periodic, muffled "whooshing" sound, and a loud "whooshing" noise caused by airflow fluctuations in the pipeline network. This phenomenon is known as the surge condition of a centrifugal compressor. Mild surge, due to the pulsation and periodic oscillation of the airflow, causes large fluctuations in supply parameters (pressure, flow, etc.), significantly reducing system stability and the operating efficiency of the centrifugal compressor. Severe surge causes intense blade vibration, intensifies noise, and may even lead to friction and collision between moving and stationary parts, causing bending deformation of the compressor shaft and serious damage to the equipment.

[0003] The surge line, also known as the surge boundary line, is as follows: Figure 1 The diagram shows the line connecting the surge points of a centrifugal compressor. The compressor's operating point should ideally be to the right of the surge line to avoid surge. However, the actual surge line differs from the theoretical surge line, and the deviation from the rated operating point increases with the actual surge line. Furthermore, the actual surge line is located to the right of the theoretical surge line. If the theoretical surge line were used for protection, the centrifugal compressor would experience surge when operating in region A between the theoretical and actual surge lines. Therefore, the theoretical surge line cannot be directly used in actual equipment; a new actual surge curve needs to be established.

[0004] Currently, the methods for determining the surge line of centrifugal compressors on the market generally include the following:

[0005] 1) Reserve a 5%-20% protection margin directly based on the theoretical surge line, and use this as the actual surge line for equipment protection. Figure 1 As shown. The larger the protection margin reserved by this method, the farther the operating point of the centrifugal compressor is from the surge line, and the safer the equipment operation;

[0006] The disadvantage is that if the theoretical 5%-20% margin line is used for protection, the centrifugal compressor cannot operate in region B between the actual surge line and the theoretical 5%-20% margin line. This method narrows the operating range of the centrifugal compressor and makes the adjustable range too small. When the operating point of the equipment under partial load is not within the operating range of the centrifugal compressor, bypassing can only be done through the hot gas bypass valve, resulting in increased energy consumption.

[0007] 2) Various advanced intelligent algorithms, such as neural networks and genetic algorithms. This method is applicable to two situations:

[0008] A. Standardized products: Before a standard product leaves the factory, a large number of experiments are required to obtain a massive amount of experimental data. Then, regression analysis is performed on various models. This method can achieve relatively accurate surge prevention for the product and maximize the operating range of the centrifugal compressor.

[0009] B. The control system connects to the centrifugal compressor equipment or project on the cloud platform through an IoT gateway. This method can achieve accurate prediction and protection of the equipment throughout its entire life cycle using advanced intelligent algorithms, and maximize the operating range and operating efficiency.

[0010] The drawback is that for non-standard products and projects based on data confidentiality, where devices are not allowed to connect to the cloud platform, the intelligent algorithm cannot be used.

[0011] 3) Test on the test bench before leaving the factory, measure several or a dozen surge points, and then use data regression software such as Excel to regress the surge line equation. Then input the equation into the PLC for surge line protection. This method can regress the surge line more accurately.

[0012] shortcoming:

[0013] A. When there are a large number of surge points to measure, it will consume a lot of manpower and resources;

[0014] B. When the number of surge points measured is small, the surge line is inaccurate;

[0015] C. Factory test data cannot fully simulate the actual situation, making model updates difficult;

[0016] D. If the factory test values ​​are inaccurate, it will affect the operation of the equipment throughout its entire life cycle;

[0017] E. In Excel software, many trendline regression models have poor accuracy and cannot intuitively display the model residuals, making model selection inconvenient.

[0018] F. The high cost of setting up the test bench system and the long testing time result in increased product costs and extended delivery time.

[0019] In summary, current methods for determining the surge line of centrifugal compressors generally suffer from inaccuracies and a lack of real-time updates. This is because different equipment conditions at different times can affect the surge line. Therefore, surge point data must be monitored and the surge line updated in real time for more accurate surge control.

[0020] Therefore, a method for online updating of surge lines and monitoring of outliers in centrifugal compressors is needed, which can realize real-time monitoring and updating of surge point data, so that surge control can be more accurate. Summary of the Invention

[0021] The purpose of this invention is to provide a new technical solution to improve or solve the technical problems existing in the prior art as described above.

[0022] The technical solution provided by this invention is as follows: A method for online updating of surge line and monitoring of outliers in a centrifugal compressor, comprising a centrifugal compressor system, wherein the centrifugal compressor system includes a centrifugal compressor, and the method for online updating of surge line and monitoring of outliers includes the following steps:

[0023] S1. Start the centrifugal compressor;

[0024] S2. Establish the theoretical surge line of the centrifugal compressor;

[0025] S3. Obtain the surge point of the centrifugal compressor by partitioning;

[0026] S4. Calculate the equivalent distance D from each surge point to the theoretical surge line;

[0027] S5. Calculate the standard deviation σ of the equivalent distance of all surge points;

[0028] S6. Determine whether the standard deviation σ of the equivalent distance is greater than the threshold θ. If not, return to step S3 to continue monitoring; if yes, proceed to S7.

[0029] S7. Determine whether the total number of recorded surge points is greater than m. If so, compare the equivalent distance D from the surge point to the theoretical surge line, remove the surge point with the largest equivalent distance D, and establish the surge line. If not, return to step S3 to continue monitoring.

[0030] Compared with the prior art, the technical solution provided by the present invention has the following beneficial effects: the present invention can remove outlier values ​​of surge points during the process of establishing surge lines, establish accurate surge lines, and update surge lines in real time, thereby accurately predicting surge and preventing the occurrence of surge phenomena.

[0031] Based on the above technical solution, the present invention can be further improved as follows.

[0032] Furthermore, in step S3, the method for obtaining the surge point of the centrifugal compressor by partition is as follows:

[0033] S31. Establish partitions, divide the operating frequency range of the centrifugal compressor into G frequency intervals, 1≤m≤G, and divide the operating pressure ratio range into G pressure ratio intervals, with the G frequency intervals and G pressure ratio intervals corresponding one-to-one.

[0034] S32. Detect the surge point of the centrifugal compressor;

[0035] S33. Determine whether a surge point has already been recorded in the frequency range where the surge point is located. If so, remove a surge point in the frequency range and keep only one surge point in each frequency range. Otherwise, record the exhaust pressure, intake pressure and frequency of the surge point, and calculate the pressure ratio of the centrifugal compressor. Pressure ratio = exhaust pressure / intake pressure.

[0036] The beneficial effect of adopting the above-mentioned further solution is that the surge point can be monitored in real time during the operation of the centrifugal compressor, and the surge point used to establish the surge line can be obtained through the partition acquisition method, thereby realizing the online establishment function of the surge line.

[0037] Further, in step S2, the theoretical surge line equation is:

[0038] y=f(x) (1)

[0039] in:

[0040] y is the minimum operating frequency value;

[0041] x represents the pressure ratio.

[0042] Furthermore, in step S4, the equivalent distance D is calculated as follows:

[0043] Let the measured surge points be (x1, y1), (x2, y2), ..., (x m ,y m The minimum distance d from each point to the theoretical surge line y = f(x) is calculated using the bisection method. i ;

[0044] The formula for calculating the equivalent distance of the surge point is as shown in equation (2):

[0045] D i =λ j *d i (2)

[0046] Where λ is the correction coefficient, and the initial value λ0 = 1;

[0047] i represents the i-th surge point, 1≤i≤m;

[0048] j represents the j-th frequency interval, 1≤j≤G;

[0049] After a surge point is eliminated, the adjustment factor λ of the partition where the eliminated surge point is located is reduced. The formula for calculating the adjustment factor λ is as shown in equation (3):

[0050] λ jq =ελ j(q-1) (3)

[0051] Where ε is the decreasing step size;

[0052] q represents the number of surge points eliminated in the j-th frequency interval, where q≥1.

[0053] The beneficial effect of adopting the above-mentioned further scheme is that, since the actual surge line is not completely parallel to the theoretical surge line, and the extent of the offset is unknown, surge point data in areas where the actual surge line deviates significantly from the theoretical surge line may be frequently treated as outliers and removed. Outlier monitoring is performed by calculating the equivalent distance D. By comparison, the surge point corresponding to the maximum value of D is identified as an outlier. When the number of data points is greater than m, the outlier is removed. By setting a smaller step size ε, the problem of surge point data in areas deviating significantly from the theoretical surge line being frequently treated as outliers and removed is avoided. Through equivalent distance calculation, these outliers will eventually no longer be considered outliers. When the actual surge point does indeed deviate significantly from the theoretical value, after repeated removals several times, the D value for all areas will converge to the same level, and the outlier will no longer be detected as an outlier.

[0054] Furthermore, in step S31, the operating frequency of the centrifugal compressor is divided into G frequency intervals of equal width.

[0055] Furthermore, in step S7, when the number of recorded surge points is less than m, the zone where the surge point is located is protected by operating condition protection, and the protection method is as follows:

[0056] Let the pressure ratio at the surge point be K. i The frequency is f i The pressure ratio range where the surge point is located is (a, b], and the lower limit of the protection frequency range. t is the width of the frequency range; the upper limit of the protection frequency range A2 = A1 + t; when the actual operating pressure ratio of the centrifugal compressor is K p The protection frequency of the centrifugal compressor Where η represents the deviation.

[0057] Further, in step S33, it is determined whether a surge point has already been recorded in the frequency range where the surge point is located. If so, the previous surge point in the frequency range is removed, and the new surge point is retained.

[0058] The beneficial effect of adopting the above-mentioned further scheme is that, due to surge point protection, the compressor will not operate to the left of the surge point. Therefore, once a surge occurs again in the same zone, the newly appearing surge point must be to the right of the previous surge point. The surge point update logic is to replace the old surge point in the zone with the newly appearing surge point in the zone. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0060] Figure 1 This is a surge curve diagram for a centrifugal compressor.

[0061] Figure 2 This is a diagram showing the positional relationship between outliers and surge curves in this invention;

[0062] Figure 3 This is a flowchart of the method for online updating of surge lines and monitoring of abnormal values ​​in centrifugal compressors according to the present invention;

[0063] Figure 4 The flowchart of the centrifugal compressor surge monitoring method of the present invention is as follows: Detailed Implementation

[0064] The principles and features of the present invention are described below with reference to examples. The examples are only used to explain the present invention and are not intended to limit the scope of the present invention.

[0065] The inventors achieved online surge line establishment using a method for centrifugal compressors. However, outlier removal is a crucial and indispensable step in the surge line establishment process. Outliers, as the name suggests, are values ​​that significantly deviate from the majority of data points in the dataset. For regression analysis, outlier handling is even more critical because:

[0066] 1) When anomaly point 1 appears on the left, that is, when the anomaly value is to the left of the actual surge line, such as... Figure 2 As shown, the regression surge line equations will inevitably cause surge in the region C near this point, which will have an adverse effect on the stable operation of the equipment.

[0067] 2) When anomaly point 2 occurs on the right side, that is, when the anomaly value is to the right of the actual surge line, such as... Figure 2 As shown, the regression surge line equations will place the region D near this point in an overprotected surge state, which narrows the operating range of the equipment near this point and is not conducive to energy saving and consumption reduction.

[0068] Therefore, in order to obtain accurate results, we need to remove outliers so that we can ensure that our surge line can accurately reflect the actual operating conditions of the equipment.

[0069] like Figure 3 As shown, a method for online updating of surge lines and monitoring of outliers in a centrifugal compressor includes the following steps:

[0070] S1. Start the centrifugal compressor;

[0071] S2. Establish the theoretical surge line for the centrifugal compressor;

[0072] S3. Obtain the surge point of the centrifugal compressor by partition, including the following steps:

[0073] S31. Establish partitions, divide the operating frequency range of the centrifugal compressor into G frequency intervals, divide the operating pressure ratio range into G pressure ratio intervals, and the G frequency intervals and G pressure ratio intervals correspond one-to-one. Surge point data is obtained by partitioning; In this embodiment, the operating frequency range of the centrifugal compressor is divided into G frequency intervals of equal width.

[0074] For example, the theoretical maximum pressure ratios corresponding to the frequencies of 266.7Hz and 366.7Hz of a centrifugal compressor are 1.83 and 3.07, respectively. The pressure ratio range is divided into 10 zones: [1.83, 1.954], (1.954, 2.078], (2.078, 2.202], (2.202, 2.326], (2.326, 2.45], (2.45, 2.574], (2.574, 2.698], (2.698, 2.822], (2.822, 2.946], and (2.946, 3.07]. At the same time, the 266.7Hz to 366.7Hz range is divided into 10 frequency zones corresponding to the pressure ratio range, with each zone having a frequency range of 10Hz. Only one surge point is retained in each zone.

[0075] S32. Detect the surge point of the centrifugal compressor;

[0076] S33. Determine whether a surge point has already been recorded within the frequency range where the surge point is located. If so, remove one surge point from the frequency range and keep only one surge point in each frequency range. Otherwise, record the exhaust pressure, intake pressure, and frequency of the surge point, and calculate the pressure ratio of the centrifugal compressor. Pressure ratio = exhaust pressure / intake pressure.

[0077] More specifically, the discharge pressure, intake pressure, and frequency of the centrifugal compressor are continuously monitored during operation to determine whether the centrifugal compressor is experiencing surge. If no surge occurs, the monitoring continues. If surge occurs, it is determined whether a surge point has already been recorded within the frequency range where the surge point is located. If so, a surge point within that frequency range is removed, and only one surge point is retained in each frequency range. Otherwise, the discharge pressure, intake pressure, and frequency of the surge point are recorded.

[0078] S4. Calculate the equivalent distance D from each surge point to the theoretical surge line;

[0079] S5. Calculate the standard deviation σ of the equivalent distance of all surge points;

[0080] S6. Determine whether the standard deviation σ of the equivalent distance is greater than the threshold θ. If not, return to step S3 to continue monitoring; if yes, proceed to S7.

[0081] S7. Determine if the total number of recorded surge points is greater than m. If so, compare the equivalent distance D from the surge point to the theoretical surge line, remove the surge point with the largest equivalent distance D, and establish the surge line. If not, return to step S3 to continue monitoring.

[0082] More specifically,

[0083] In step S2, the theoretical surge line equation is:

[0084] y=f(x) (1)

[0085] Where: y is a variable, representing the minimum operating frequency value; x is a variable, representing the pressure ratio, i.e., pressure ratio = exhaust pressure / intake pressure;

[0086] In step S4, the equivalent distance D is calculated as follows:

[0087] Let the measured surge points be (x1, y1), (x2, y2), ..., (x m ,y m The minimum distance d from each point to the theoretical surge line y = f(x) is calculated using the bisection method. i ;

[0088] Since the actual surge line is not perfectly parallel to the theoretical surge line, and the extent of the offset is unknown, surge point data in areas where the actual surge line deviates significantly from the theoretical surge line may be frequently treated as outliers and removed. Therefore, this method uses the equivalent distance D to monitor outliers, and the calculation formula is shown in equation (2):

[0089] D i =λ j *d i (2)

[0090] λ is the correction coefficient, with an initial value of λ0 = 1;

[0091] i represents the i-th surge point, 1≤i≤m;

[0092] j represents the j-th frequency interval, 1≤j≤G;

[0093] By comparison, the surge point with the largest D value is identified as an outlier. When the number of data points is greater than m, the outlier is removed.

[0094] When a point is removed as an outlier, the λ of that region will decrease, as shown in equation (3): the step size is, that is, when a surge point is removed, the adjustment factor λ of the partition where the removed surge point is located is reduced:

[0095] λ jq =ελ j(q-1) (3)

[0096] Where ε is the decreasing step size;

[0097] q represents the number of surge points eliminated in the j-th frequency interval, where n≥1.

[0098] This method avoids the problem of surge point data in areas that deviate significantly from the theoretical surge line being frequently treated as outliers and removed. Through equivalent distance calculation, these data will eventually no longer be considered outliers. When actual surge points do indeed deviate significantly from the theoretical value, after several rounds of removal, the standard deviation (D) of all areas will converge, and they will no longer be monitored as outliers. Outliers are monitored periodically, and outlier monitoring stops when the standard deviation of D for all surge points falls below a threshold θ.

[0099] In step S7, when the number of recorded surge points is less than m, the zone containing the surge point is protected by operating condition protection. The protection method is as follows:

[0100] Let the pressure ratio at the surge point be K. i The frequency is f i The pressure ratio range where the surge point is located is (a, b], the protection frequency range is [A1, A2], and the lower limit of the protection frequency range is... t is the width of the frequency range; the upper limit of the protection frequency range A2 = A1 + t; when the actual operating pressure ratio of the centrifugal compressor is K p The protection frequency of the centrifugal compressor Where η represents the deviation. That is, when the actual working pressure ratio is K p The minimum operating frequency of a centrifugal compressor should not be lower than H, otherwise surge will easily occur.

[0101] For example: the pressure ratio at the surge point is K. i =2.01, frequency is f i=281.4Hz, the surge point is located in the pressure ratio range of (1.954, 2.078], the lower limit of the protection frequency range. The frequency range width is 10Hz; the upper limit of the protection frequency range A2 = A1 + 10; when the actual operating pressure ratio of the centrifugal compressor is K p The protection frequency of the centrifugal compressor Where η represents the deviation. That is, when the actual working pressure ratio is K p The minimum operating frequency of a centrifugal compressor should not be lower than H, otherwise surge will easily occur.

[0102] In step S7, the least squares method is used to perform nonlinear equation regression to obtain the surge line equation. The equation regression method is as follows:

[0103] Suppose that the nonlinear equation can be expressed by a polynomial, as in equation (4).

[0104] y = f(x) = a0 + a1x + a2x 2 …+a n x n (4)

[0105] Then, based on the principle of least squares, we find the coefficient a that minimizes the error E. The formula for calculating the error E is shown in equation (5):

[0106]

[0107] Where: y is the dependent variable, representing the minimum operating frequency value;

[0108] x is the independent variable, representing the pressure ratio;

[0109] f(x) is a function;

[0110] a0, a1, ... a n These are the polynomial coefficients;

[0111] n is the highest order of the polynomial;

[0112] i indicates which data point is being read;

[0113] The nonlinear regression problem is transformed into an extremum problem. Equation (5) is applied to a0, a1, ... a n Taking the partial derivative and setting it to 0, we obtain equation (6):

[0114]

[0115] Arrange equation (6) to obtain the matrix form, as shown in equation (7):

[0116]

[0117] Further simplified to equation (8):

[0118] MA = B (8)

[0119] in:

[0120]

[0121] At the same time, it can be found that M is a symmetric positive definite matrix, and according to the Cholesky decomposition method, equation (9) can be obtained:

[0122] LL T =M (9)

[0123] in:

[0124]

[0125] Among them l nn This represents the value in the nth row and nth column of matrix L;

[0126] From equations (8) and (9), we obtain equation (10):

[0127] LL T A = B (10)

[0128] Then find the inverse matrix L of matrix L. -1 Sum matrix L T The inverse matrix L T-1 Thus, we obtain equation (11).

[0129] A = L T-1 L -1 B (11)

[0130] Finally, the coefficients a0, a1, ... a are obtained. n The value of .

[0131] Determining the value of n: Since different equipment has different surge properties, the order of the optimal regression equation may also be different. Considering the PLC's computing power and the accuracy of Lreal data, this method sets the maximum value of n to 6. Therefore, nonlinear regression needs to obtain the optimal order solution from n = 1, 2, 3, 4, 5, 6.

[0132] When n = 1, 2, 3, 4, 5, 6, the corresponding coefficients a0, a1, ... a can be calculated respectively. n Values, and substitute the coefficients into Find the minimum value of f(n), and the value of n corresponding to a0, a1, ... a n The value is the coefficient of the final surge line, and n is the highest order of the surge line.

[0133] Using the above method, we can achieve online regression of the surge line of centrifugal compressors, which can not only accurately obtain the surge line, making the operating conditions of centrifugal compressors wider, but also save the investment of funds and time in factory testing.

[0134] like Figure 4 As shown, in step S32, the centrifugal compressor system further includes a frequency converter for controlling the operation of the centrifugal compressor and a pipeline network connected to the centrifugal compressor; the centrifugal compressor includes an inlet guide vane, the pipeline network is provided with a bypass valve, and also includes a frequency converter input power detection device. The method for detecting the surge point of the centrifugal compressor is as follows;

[0135] S321. Continuously measure the input power X of the frequency converter within a set period using an input power detection device. i Where i is a positive integer greater than or equal to 1, calculate the standard deviation s of the power change. If the absolute value of the standard deviation s within one cycle is less than or equal to the surge judgment setting value Q, then continue to detect the input power of the frequency converter; if the absolute value of the standard deviation s within one cycle is greater than the surge judgment setting value Q, then continue to execute S322.

[0136] S322. Determine if the issue is caused by changes in the active load. Changes in the active load include changes in the inverter frequency, the inlet guide vane opening, and the bypass valve opening. If not, execute S323; if so, continue execution.

[0137] S322.1 Determine if the change is caused by a change in the set frequency. If so, execute S321 after the frequency change response time TF. If the frequency is increasing, the frequency change response time TF is the main unit's acceleration response time, TF = Ta * Ha / Hmax. If the frequency is decreasing, the frequency change response time TF is the main unit's deceleration response time, TF = t + Td * Hd / Hmax, where Hmax is the inverter's maximum frequency, Ta is the acceleration time, Td is the deceleration time, Ha is the frequency increase, Hd is the frequency decrease, and the duration of the power feedback fluctuation caused by the frequency change is t. If the change is not caused by a change in the set frequency, continue executing S322.2.

[0138] S322.2 Determine if the issue is caused by a change in the inlet guide vane opening. If so, execute S321 after the inlet guide vane response time TR. The inlet guide vane response time TR = t1 + T1 * a / 100, where T1 is the full stroke time of the inlet guide vane, a is the change in the inlet guide vane opening, and the duration of the power feedback fluctuation caused by the change in the inlet guide vane opening is t1. If not, execute S322.3.

[0139] S322.3 is caused by a change in the bypass valve opening. S321 will be executed after the bypass valve response time Tb. The bypass valve response time Tb = t2 + T2 * b / 100, where T2 is the full stroke time of the bypass valve and b is the change in the bypass valve opening. The duration of the power feedback fluctuation caused by the change in the bypass valve opening is t2.

[0140] It should be noted that since changes in the frequency of the inverter, the opening of the inlet guide vane, and the opening of the bypass valve can also cause power feedback fluctuations, when the system experiences the above-mentioned active load changes, power feedback monitoring should be performed after a certain response time to determine whether surge has occurred, so as to avoid false alarms from the equipment.

[0141] S323. Determine whether the power fluctuation is caused by passive load changes: When the load increases, a correlation coefficient r ≥ 0.8 indicates a positive high correlation, while when the load decreases, a correlation coefficient r ≤ -0.8 indicates a negative high correlation. When a centrifugal compressor experiences surge, the inverter's input power will change periodically. If the absolute value of the power standard deviation s within one cycle is greater than the surge judgment setting value Q, and the correlation coefficient r first shows a negative high correlation and then a positive high correlation, then it is determined that the centrifugal compressor has experienced a surge. Record the inverter's frequency and the ratio of the centrifugal compressor's suction pressure to discharge pressure during the surge. If the absolute value of the standard deviation s is greater than the surge judgment setting value Q, and the correlation coefficient does not show a negative high correlation followed by a positive high correlation, then it is considered to be caused by passive load changes, and continue to execute S321.

[0142] During surge, the power characteristics of the inverter fluctuate periodically, with the periodicity and power standard deviation varying with the severity of the surge. Mild surges have longer surge cycles and smaller power standard deviations, while severe surges have shorter surge cycles and larger power standard deviations. Therefore, the surge intensity judgment in this invention is determined by the standard deviation and the surge cycle. A strong correlation coefficient that is initially negative and then positive indicates surge. If this pattern is not observed, it is considered to be caused by passive load. Passive load-induced surges may only show a positive or negative correlation coefficient, and the correlation coefficient may not reach 0.8. Therefore, the standard deviation will be greater than the set value. As long as the initial negative-then-positive pattern is not observed, it is considered to be caused by passive load changes.

[0143] Furthermore, when the load remains constant, it will not cause changes in the power standard deviation and correlation coefficient, thus avoiding false surge predictions. The speed of active load adjustment is determined by the setter and may be very fast, adjusting every few seconds or tens of seconds; however, passive load fluctuations in the system will not be as frequent.

[0144] The formula for calculating the standard deviation s is as follows:

[0145]

[0146] Where n is the number of samples, and the sample mean is: X i This is a power sample.

[0147] The formula for calculating the correlation coefficient r is as follows:

[0148]

[0149] Where, Y i For X i The power sample corresponds to the time sample, and -1≤r≤1.

[0150] The correlation coefficient ranges from -1 to +1. When r > 0, it indicates a positive correlation between the two variables, meaning that the larger the value of one variable, the larger the value of the other variable will be. When r < 0, it indicates a negative correlation between the two variables, meaning that the larger the value of one variable, the smaller the value of the other variable will be. The larger the absolute value of r, the stronger the correlation between the two variables. Generally, |r| > 0.95 is considered a significant correlation, |r| ≥ 0.8 is considered a high correlation, 0.5 ≤ |r| < 0.8 is considered a moderate correlation, 0.3 ≤ |r| < 0.5 is considered a low correlation, and |r| < 0.3 is considered a very weak correlation, or no correlation.

[0151] The standard deviation and correlation coefficient of the inverter input power change are used to determine whether a centrifugal compressor has experienced surge. However, since changes in active and passive loads can also cause changes in the power standard deviation and correlation coefficient, in order to avoid system misjudgment, the characteristics of active and passive load changes are analyzed and eliminated in the surge change judgment to ensure the accuracy of surge judgment.

[0152] This invention eliminates outliers during the surge line establishment process, creating an accurate surge line that is updated in real time. This allows for accurate surge prediction and prevention. Using this online update method, centrifugal compressors can operate more stably and have a wider operating range, improving efficiency and performance while extending their lifespan. Furthermore, since this online establishment method can be tested and verified during actual operation, it saves factories time and money on testing and experimentation, thus improving production efficiency.

[0153] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. 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 online updating of surge lines and monitoring of outliers in a centrifugal compressor, comprising a centrifugal compressor system, wherein the centrifugal compressor system includes a centrifugal compressor, characterized in that, The online update and outlier monitoring method for surge lines includes the following steps: S1. Start the centrifugal compressor; S2. Establish the theoretical surge line of the centrifugal compressor; S3. Obtain the surge point of the centrifugal compressor by partitioning; S4. Calculate the equivalent distance D from each surge point to the theoretical surge line; S5. Calculate the standard deviation σ of the equivalent distance of all surge points; S6. Determine whether the standard deviation σ of the equivalent distance is greater than the threshold θ. If not, return to step S3 to continue monitoring; if yes, proceed to S7. S7. Determine whether the total number of recorded surge points is greater than m. If so, compare the equivalent distance D from the surge point to the theoretical surge line, remove the surge point with the largest equivalent distance D, and establish the surge line. If not, return to step S3. In step S3, the method for obtaining the surge point of the centrifugal compressor by partition is as follows: S31. Establish partitions, divide the operating frequency range of the centrifugal compressor into G frequency intervals, 1≤m≤G, and divide the operating pressure ratio range into G pressure ratio intervals, with the G frequency intervals and G pressure ratio intervals corresponding one-to-one. S32. Detect the surge point of the centrifugal compressor; S33. Determine whether a surge point has been recorded in the frequency range where the surge point is located. If so, remove a surge point in the frequency range and keep only one surge point in each frequency range. Otherwise, record the exhaust pressure, intake pressure and frequency of the surge point, and calculate the pressure ratio of the centrifugal compressor. Pressure ratio = exhaust pressure / intake pressure.

2. The method for online updating of surge lines and monitoring of abnormal values ​​in centrifugal compressors according to claim 1, characterized in that, In step S2, the theoretical surge line equation is: （1） in: This is the minimum operating frequency value; This refers to the pressure ratio.

3. The method for online updating of surge lines and monitoring of abnormal values ​​in centrifugal compressors according to claim 2, characterized in that, In step S4, the equivalent distance D is calculated as follows: Let the measured surge point be... The bisection method was used to calculate the distance from each point to the theoretical surge line. minimum distance ; The equivalent distance of the surge point is: (2) in, For correction coefficients, initial value =1; Indicates the first A surge point, ; Indicates the first One frequency range, ; Once a surge point is removed, the adjustment factor for reducing the partition containing the removed surge point is: （3） in, To reduce the step size; The number of times surge points are removed within the j-th frequency interval. ≥1.

4. The method for online updating of surge lines and monitoring of abnormal values ​​in centrifugal compressors according to claim 1, characterized in that, In step S31, the operating frequency of the centrifugal compressor is divided into G frequency intervals of equal width.

5. The method for online updating of surge lines and monitoring of abnormal values ​​in centrifugal compressors according to claim 1, characterized in that, In step S7, when the number of recorded surge points is less than m, the partition where the surge point is located is protected by the operating point protection method as follows: Let the pressure ratio at the surge point be... , frequency is The pressure ratio range where the surge point is located is (a, b], and the lower limit of the protection frequency range. t is the width of the frequency range; Upper limit of protection frequency range When the actual operating pressure ratio of the centrifugal compressor is The protection frequency of the centrifugal compressor , where η represents the deviation.

6. The method for online updating of surge lines and monitoring of abnormal values ​​in centrifugal compressors according to claim 5, characterized in that, Step S33: Determine whether a surge point has already been recorded in the frequency range where the surge point is located. If so, remove the previous surge point in the frequency range and retain the new surge point.

Images