A method and system for analyzing load shedding data of pumped storage power stations based on signal decomposition and reconstruction

Abstract

This invention discloses a method and system for analyzing load shedding data of pumped storage power stations based on signal decomposition and reconstruction. First, a one-dimensional numerical simulation method is used to simulate the unit's load shedding transient process, which has boundary conditions essentially consistent with the measured signal. The permutation entropy value of the simulated signal is calculated. Next, variational mode decomposition is performed on the measured signal, and component reconstruction is performed using mutual information as a criterion to reduce the permutation entropy value. Then, the reconstructed signal undergoes fully adaptive noise-complete ensemble empirical mode decomposition processing, and the components are superimposed to obtain experimental data consistent with the permutation entropy of the simulated signal. The system includes a data acquisition module, a data processing module, and a data output module. This invention can filter out noise from complex measured data and extract accurate water hammer pressure and pressure pulsation, which is fundamental for analyzing the actual characteristics of the power station's transient process and improving the accuracy of transient process inversion and prediction.

Description

Technical Field

[0001] This invention relates to the field of water conservancy and hydropower engineering technology, and in particular to a method and system for analyzing load shedding data of pumped storage power stations based on signal decomposition and reconstruction. Background Technology

[0002] Pumped storage power stations play a crucial role in peak shaving and valley filling within the power grid. With the integration of intermittent energy sources such as wind and solar power, the operating conditions of pumped storage units change more frequently. During operating condition transitions, the water-transfer power generation system undergoes a transient process, and the load shedding transient process is one of the more dangerous ones. Due to the S-characteristics of reversible units, significant water hammer pressure is generated within the flow channel, accompanied by severe pressure pulsations. Accurate analysis and prediction of the transient process of pumped storage power stations are of great significance for the safe operation of the power station.

[0003] For the study of the characteristics of the transient process in hydropower stations, scholars both domestically and internationally often employ one-dimensional numerical simulation and three-dimensional numerical simulation using turbulence models. However, due to differences in boundary conditions, turbine characteristics, hydraulic parameters at various cross-sections of the water conveyance pipeline compared to actual engineering conditions, and calculation errors, discrepancies arise between the calculated results and measured data for the transient process. For the commonly used one-dimensional transient process simulation, the influence of pressure pulsation cannot be considered. Therefore, engineering design needs to account for calculation errors and pressure pulsation corrections. However, different unit manufacturers use different correction standards, and there is currently a lack of sufficient evidence for the correction of pressure pulsation.

[0004] Before a pumped-storage power station is officially put into operation, a field load shedding test is generally required to assess the safety of the water system and the quality and performance of electromechanical equipment such as the generator and governor. The measured data has significant research value. During the load shedding test, to avoid affecting the generator, long connecting pipes are often used to measure the data. However, long pressure measuring pipes are prone to resonance with the generator's flow components under steady-state or unsteady-state operation, and the acquired signals inevitably contain a large amount of noise, making it difficult to accurately obtain the extreme values ​​of water hammer pressure. The measured pressure data during the transient process can be roughly divided into water hammer pressure, pressure pulsation, and noise. How to filter out noise from the complex measured data and extract accurate water hammer pressure and pressure pulsation is fundamental to analyzing the actual characteristics of the power station's transient process and improving the accuracy of transient process inversion and prediction.

[0005] In summary, when a pump-turbine unit experiences load shedding, designing an accurate method to extract pressure pulsations and determine whether the water hammer pressure of the pump-turbine meets the requirements for safe operation has become a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a method and system for analyzing load shedding data of pumped storage power stations based on signal decomposition and reconstruction, so as to determine whether the water hammer pressure of the pump turbine is within the safe threshold range, and effectively protect the safety of the system during the load shedding process of the turbine.

[0007] To address the aforementioned technical problems, this invention provides a method for analyzing load shedding data of pumped storage power stations based on signal decomposition and reconstruction, comprising the following steps:

[0008] Step 1: Using a one-dimensional simulation method, obtain a simulation signal with boundary conditions that are basically consistent with the measured pressure signal during load shedding, and calculate the entropy value E of the simulation signal arrangement. 仿真 ;

[0009] Step 2: Obtain the measured pressure signal x(t) of the water pump turbine during load shedding and use it as the input information for VMD decomposition;

[0010] Step 3: Determine the number of VMD decomposition modes using the center frequency method, perform VMD decomposition on the measured pressure signal of load shedding, and obtain several intrinsic mode functions (IMFs);

[0011] Step 4: Calculate the mutual information value M of each intrinsic mode function. k Then normalize the mutual information value τ. k The weights for reconstructing the signal based on the eigenmode functions are used to obtain the reconstructed signal J(t);

[0012] Step 5: Perform CEEMDAN decomposition on the reconstructed signal to obtain several components C, and calculate their permutation entropy values.

[0013] Step 6: Superimpose the components after CEEMDAN decomposition to obtain the measured secondary processed signal y, which is consistent with the arrangement entropy of the simulated signal. n (t);

[0014] Step 7: The difference between the measured signal and the measured secondary processed signal is the pressure pulsation signal.

[0015] Preferably, in step 1, a one-dimensional simulation method is used, namely, the fluid mass conservation equation and motion equation are discretized based on the method of characteristics, and the entire system is solved by combining the boundary conditions of the reservoir, surge tank, and unit.

[0016] Preferably, in step 3, the number of VMD decomposition modes is determined by the center frequency method. First, the range of values ​​for the number of VMD modes is given, and then the VMD algorithm is executed sequentially for pre-decomposition to obtain the center frequency of each mode. The number of modes with the largest center frequency interval is the optimal solution. Based on the obtained optimal parameter values, VMD decomposition is performed on the load shedding measured pressure signal to obtain several intrinsic mode functions (IMFs).

[0017] Preferably, in step 4, the expression for using the normalized mutual information value as the weight of the intrinsic mode function to reconstruct the signal is:

[0018] τ k =M k / max(M k )

[0019]

[0020] In the formula, τ k M is the normalized mutual information value; k The mutual information values ​​between each intrinsic mode component and the measured signal are: J(t) = u(t) / (t - u) = u(t - u ... k (t) represents the intrinsic mode components obtained after VMD decomposition.

[0021] Preferably, in step 6, the components after CEEMDAN decomposition are superimposed to obtain the expression of the measured secondary processed signal that is consistent with the arrangement entropy of the simulated signal:

[0022]

[0023] y n (t)=J K (t)+nδC N-K-1

[0024] |E y -E 仿真 |≤ΔE

[0025] In the formula, C represents several components obtained from the CEEMDAN decomposition; calculate the permutation entropy value of each signal until the permutation entropy value of the signal is greater than the permutation entropy value of the simulated signal, and obtain k-1 = K; J k (t) represents the intermediate processed signal obtained after each superposition; J K (t) represents the initial signal at which C is first superimposed; R represents the residual after CEEMDAN decomposition; N represents the total number of decomposition layers excluding the residual; y n (t) represents the secondary processed measured pressure signal obtained from each superposition; δ is the step size; n is the number of superpositions; E y y after each superposition n The permutation entropy value of (t); E 仿真 ΔE represents the entropy value of the simulated signal arrangement; ΔE represents the entropy error of the arrangement.

[0026] Accordingly, a pumped storage power station load shedding data analysis system based on signal decomposition and reconstruction includes: a data acquisition module, a data processing module, and a data output module; the data acquisition module acquires a simulation signal with boundary conditions basically consistent with the measured pressure signal during load shedding, and calculates the permutation entropy value E of the simulation signal.仿真 The system acquires the measured pressure signal x(t) of the pump turbine under load shedding and uses it as input information for VMD decomposition. The data processing module determines the number of VMD decomposition modes using the center frequency method, performs VMD decomposition on the measured pressure signal under load shedding, obtains several intrinsic mode functions (IMFs), and calculates the mutual information value M of each IMF. k Then normalize the mutual information value τ. k The weights of the reconstructed signal are used to obtain the reconstructed signal J(t). CEEMDAN decomposition is then performed on the reconstructed signal to obtain several components C, and their permutation entropy values ​​are calculated. These components are then superimposed to obtain the measured secondary-processed signal y, which matches the permutation entropy of the simulated signal. n (t); The data output module outputs a pressure pulsation signal, which is the difference between the measured signal and the measured secondary processed signal.

[0027] The beneficial effects of this invention are as follows: This invention proposes a joint processing method combining variational mode decomposition and fully adaptive noise-complete set empirical mode decomposition. First, variational mode decomposition is performed on the signal, and component reconstruction is performed using mutual information as a criterion to reduce the permutation entropy value. Then, fully adaptive noise-complete set empirical mode decomposition is performed on the reconstructed signal, and the components are superimposed to obtain experimental data consistent with the permutation entropy of the simulated signal. Engineering case studies demonstrate that the processing method proposed in this paper can decompose the measured pressure signal at the volute inlet relatively quickly and accurately. Simultaneously, the use of mutual information improves the accuracy of component reconstruction using correlation coefficients, providing a new method for the accurate extraction and analysis of pressure pulsations and ensuring the safe operation of hydropower units. Attached Figure Description

[0028] Figure 1 This is a flowchart of the method for extracting pressure pulsations in a water pump turbine according to the present invention.

[0029] Figure 2 The figures show the measured and simulated inlet pressure changes of the volute casing under 100% load in this embodiment of the invention.

[0030] Figure 3 This is a VMD decomposition result diagram of the measured volute inlet pressure signal in an embodiment of the present invention.

[0031] Figure 4 This is the initial processing curve of the volute inlet pressure change after VMD decomposition and mutual information superposition in this embodiment of the invention.

[0032] Figure 5 This is a spectrum diagram of the inlet pressure of the volute casing measured under 100% load in an embodiment of the present invention.

[0033] Figure 6This is a preliminary processing pressure spectrum diagram of the volute inlet in an embodiment of the present invention.

[0034] Figure 7 This is a diagram showing the reconstructed measured signal CEEMDAN decomposition result in an embodiment of the present invention.

[0035] Figure 8 This is a comparison chart of the final isentropic signal and the simulated volute inlet pressure curve in the embodiment of the present invention.

[0036] Figure 9 This is a comparison chart of the difference between the measured data and the simulation data before and after processing in an embodiment of the present invention.

[0037] Figure 10 It is the number of eigenmode functions synthesized in the isentropic process in the embodiments of the present invention.

[0038] Figure 11 This is a time-domain diagram of the pressure pulsation at the volute inlet in an embodiment of the present invention.

[0039] Figure 12 This is a spectrum diagram of the pressure pulsation at the inlet of the volute in an embodiment of the present invention. Detailed Implementation

[0040] like Figure 1 As shown, a method for analyzing load shedding data of pumped storage power stations based on signal decomposition and reconstruction includes the following steps:

[0041] Step 1: Using a one-dimensional simulation method, obtain a simulation signal with boundary conditions that are basically consistent with the measured pressure signal during load shedding, and calculate the entropy value E of the simulation signal arrangement. 仿真 ;

[0042] Step 2: Obtain the measured pressure signal x(t) of the water pump turbine during load shedding and use it as the input information for VMD decomposition;

[0043] Step 3: Determine the number of VMD decomposition modes using the center frequency method, perform VMD decomposition on the measured pressure signal of load shedding, and obtain several intrinsic mode functions (IMFs);

[0044] Step 4: Calculate the mutual information value M of each intrinsic mode function. k Then normalize the mutual information value τ. k The weights for reconstructing the signal based on the eigenmode functions are used to obtain the reconstructed signal J(t);

[0045] Specifically, the expression for using the normalized mutual information value as the weight of the intrinsic mode function to reconstruct the signal is as follows:

[0046] τ k =M k / max(M k )

[0047]

[0048] In the formula, τ k M is the normalized mutual information value; k The mutual information values ​​between each intrinsic mode component and the measured signal are: J(t) = u(t) / (t - u) = u(t - u ... k (t) represents the intrinsic mode components obtained after VMD decomposition.

[0049] Step 5: Perform CEEMDAN decomposition on the reconstructed signal J(t) to obtain several components C, and calculate their permutation entropy value;

[0050] Step 6: Superimpose the components after CEEMDAN decomposition to obtain the measured secondary processed signal y, which is consistent with the arrangement entropy of the simulated signal. n (t);

[0051] The specific expression is:

[0052]

[0053] y n (t)=J K (t)+nδC N-K-1

[0054] |E y -E 仿真 |≤ΔE

[0055] In the formula, C represents several components obtained from the CEEMDAN decomposition; calculate the permutation entropy value of each signal until the permutation entropy value of the signal is greater than the permutation entropy value of the simulated signal, and obtain k-1 = K; J k (t) represents the intermediate processed signal obtained after each superposition; J K (t) represents the initial signal at which C is first superimposed; R represents the residual after CEEMDAN decomposition; N represents the total number of decomposition layers excluding the residual; y n (t) represents the secondary processed measured pressure signal obtained from each superposition; δ is the step size; n is the number of superpositions; E y y after each superposition n The permutation entropy value of (t); E 仿真 ΔE represents the entropy value of the simulated signal arrangement; ΔE represents the entropy error of the arrangement.

[0056] Step 7: The difference between the measured signal and the measured secondary processed signal is the pressure pulsation signal.

[0057] Figure 2The figures show the measured and simulated inlet pressure changes of the spiral casing under 100% load in this embodiment of the invention. This pumped storage power station adopts a one-tunnel, two-turbine layout. The turbine has a rated power of 357MW, a rated head of 710m, a rated speed of 500r / min, 10 runner blades, a rotational frequency of 8.33Hz, and a rated flow rate of 56.85m³ / min. 3 / s.

[0058] Figure 3 This is a VMD decomposition result diagram of the measured volute inlet pressure signal in this embodiment of the invention. The K value for VMD decomposition was selected as 8 based on the center frequency method, and the decomposition result is as follows: Figure 3 As shown. The mutual information M and normalized mutual information τ between the intrinsic mode components (IMF) of each order and the original volute inlet signal were calculated (Table 1). The calculation results are as follows:

[0059] Table 1 shows the calculated mutual information M and normalized mutual information τ between the intrinsic mode components (IMF) of each order and the original volute inlet signal.

[0060]

[0061] VMD decomposes the original signal into eigenmode components arranged from low frequency to high frequency. As can be seen from the mutual information, the dependence of the components on the original signal decreases as the order increases. Based on this information, the mutual information is normalized and multiplied and superimposed to reconstruct a new volute inlet pressure signal J(t).

[0062] Figure 4 This is the initial processing curve of the volute inlet pressure change after VMD decomposition and mutual information superposition in this embodiment of the invention. Figure 5 This is a spectrum diagram of the inlet pressure of the volute casing measured under 100% load in an embodiment of the present invention. Figure 6 This is a preliminary processing pressure spectrum diagram of the volute inlet in an embodiment of the present invention. (Comparison) Figure 5 and Figure 6 It can be seen that, compared with the unprocessed pressure, the measured inlet pressure of the volute after VMD decomposition and mutual information superposition shows that the high-frequency components are basically separated, but some low-frequency pressure pulsations and noise are still mixed in. This indicates that preliminary processing can achieve preliminary extraction of time-averaged pressure, improving the speed and accuracy of subsequent secondary processing.

[0063] Figure 7 This is a diagram showing the CEEMDAN decomposition results of the reconstructed measured signal in this embodiment of the invention. Table 2 shows the measured, reconstructed, and simulated volute inlet pressures and the entropy values ​​of the arrangement of each order of intrinsic mode functions. The reconstructed volute inlet pressure signal is subjected to CEEMDAN decomposition, and the resulting intrinsic mode function is C, as shown below. Figure 7As shown. The permutation entropy of the measured, reconstructed, and simulated volute inlet pressures and the intrinsic mode functions decomposed by CEEMDAN were calculated (Table 2). The permutation entropy calculation results are as follows:

[0064] Table 2 shows the permutation entropy of the measured, reconstructed, and simulated volute inlet pressures and the intrinsic mode functions derived from CEEMDAN decomposition.

[0065]

[0066] Continued table

[0067]

[0068] As shown in Table 3, the reconstructed signal arrangement entropy decreases, achieving a preliminary extraction effect. Furthermore, with the increase of the decomposition order, its eigenmode functions generally show a decreasing trend, indicating a decrease in complexity and information content. Next, through step 6, a secondary processed volute inlet pressure signal consistent with the simulated signal arrangement entropy is obtained. The difference between this signal and the unprocessed measured signal is the pressure pulsation signal. Figure 8 This is a comparison chart of the final isentropic signal and the simulated volute inlet pressure curve in this embodiment of the invention. The maximum inlet pressure of the volute after processing, the correlation coefficient, and the number of iterations were obtained using the method of this invention, as shown in Table 3.

[0069] Table 3. Maximum pressure at the volute inlet, correlation coefficient, and number of iterations.

[0070]

[0071] As shown in Table 3, after processing by the method of the present invention, the maximum pressure at the volute inlet decreased from 1116.16 m to 1047.43 m compared with the initial measured data. When the simulated volute inlet pressure change was compared with the pressure extracted from the measured signal, the relative deviation decreased from 8.73% to 2.03%, and the correlation coefficient increased from 0.8973 to 0.9128.

[0072] Figure 9 This is a comparison chart of the difference between measured data and simulation data before and after processing in this embodiment of the invention. It shows that the fluctuation of the difference between the processed volute inlet pressure data and the simulation data is smaller than that before processing, indicating that the processed measured data is closer to the simulation data, further explaining the reason for the improvement in the correlation coefficient after processing.

[0073] Figure 10This refers to the number of intrinsic mode functions synthesized during the isentropic process in this embodiment of the invention. As the number of synthesized intrinsic mode functions increases, the signal arrangement entropy gradually approaches the simulation signal arrangement entropy value, indicating that establishing the isentropic principle between the measured signal and the simulation signal is feasible. The correlation coefficient between the processed signal and the simulation signal does not reach 1, mainly because the characteristic curve and waterway boundary conditions used in the simulation are not completely consistent.

[0074] Figure 11 This is a time-domain diagram of the volute inlet pressure pulsation in an embodiment of the present invention. Table 4 shows the components of the pressure pulsation in the mixed-flow pump turbine:

[0075] Table 4. Components of Pressure Pulsation in Mixed-Flow Pump-Turbine Types

[0076]

[0077] from Figure 12 As can be seen, there is a low-frequency, low-amplitude pressure pulsation of 6.34 Hz, approximately 0.76 times the rotational frequency, corresponding to the rotational frequency multiple range for rotating stall in Table 4. Simultaneously, there is a pressure pulsation with a frequency of 169.1 Hz, approximately twice the blade rotational frequency, corresponding to the rotational frequency multiple range for dynamic-static interference in Table 4. This indicates that the pressure pulsation amplitude is rapidly increasing due to enhanced dynamic-static interference between the turbine impeller and guide vanes. In summary, during the pump-turbine load shedding process, as the rotational speed rapidly increases, blockage may occur in the runner flow channel, leading to a rapid decrease in flow rate. The low-frequency pressure pulsation at the volute inlet is mainly caused by dynamic-static interference and rotating stall within the runner.

Claims

1. A method for analyzing load shedding data of pumped storage power stations based on signal decomposition and reconstruction, characterized in that, Includes the following steps: Step 1: Using a one-dimensional simulation method, obtain a simulation signal with boundary conditions that are basically consistent with the measured pressure signal during load shedding, and calculate the entropy value E of the simulation signal arrangement. 仿真 ; Step 2: Obtain the measured pressure signal x(t) of the water pump turbine during load shedding and use it as the input information for VMD decomposition; Step 3: Determine the number of VMD decomposition modes using the center frequency method, perform VMD decomposition on the measured pressure signal of load shedding, and obtain several intrinsic mode functions (IMFs); Step 4: Calculate the mutual information value M of each intrinsic mode function. k Then normalize the mutual information value τ. k The weights for reconstructing the signal based on the eigenmode functions are used to obtain the reconstructed signal J(t); Step 5: Perform CEEMDAN decomposition on the reconstructed signal to obtain several components C, and calculate their permutation entropy values. Step 6: Superimpose the components after CEEMDAN decomposition to obtain the measured secondary processed signal y, which is consistent with the arrangement entropy of the simulated signal. n (t); Step 7: The difference between the measured signal and the measured secondary processed signal is the pressure pulsation signal.

2. The method for analyzing load shedding data of pumped storage power stations based on signal decomposition and reconstruction as described in claim 1, characterized in that, In step 1, a one-dimensional simulation method is used, namely the method of characteristics, to discretize the fluid mass conservation equation and the equation of motion, and then solve the entire system by combining the boundary conditions of the reservoir, the surge tank, and the unit.

3. The method for analyzing load shedding data of pumped storage power stations based on signal decomposition and reconstruction as described in claim 1, characterized in that, In step 3, the number of VMD decomposition modes is determined by the center frequency method. First, the range of values ​​for the number of VMD modes is given, and then the VMD algorithm is executed sequentially for pre-decomposition to obtain the center frequency of each mode. The number of modes with the largest center frequency interval is the optimal solution. Based on the obtained optimal parameter values, VMD decomposition is performed on the load shedding measured pressure signal to obtain several intrinsic mode functions (IMFs).

4. The method for analyzing load shedding data of pumped storage power stations based on signal decomposition and reconstruction as described in claim 1, characterized in that, In step 4, the normalized mutual information value is used as the weight of the intrinsic mode function to reconstruct the signal, and the expression is as follows: t k =M k / max(M k ) In the formula, τ k M is the normalized mutual information value; k The mutual information values ​​between each intrinsic mode component and the measured signal are: J(t) = u(t) / (t - u) = u(t - u ... k (t) represents the intrinsic mode components obtained after VMD decomposition.

5. The method for analyzing load shedding data of pumped storage power stations based on signal decomposition and reconstruction as described in claim 1, characterized in that, In step 6, the components after CEEMDAN decomposition are superimposed to obtain the expression for the measured secondary processed signal that matches the arrangement entropy of the simulated signal: y n (t)=J K (t)+nδC N-K-1 In the formula, C represents several components obtained from the CEEMDAN decomposition; calculate the permutation entropy value of each signal until the permutation entropy value of the signal is greater than the permutation entropy value of the simulated signal, and obtain k-1 = K; J k (t) represents the intermediate processed signal obtained after each superposition; J K (t) represents the initial signal at which C is first superimposed; R represents the residual after CEEMDAN decomposition; N represents the total number of decomposition layers excluding the residual; y n (t) represents the secondary processed measured pressure signal obtained from each superposition; δ is the step size; n is the number of superpositions; E y y after each superposition n The permutation entropy value of (t); E 仿真 ΔE represents the entropy value of the simulated signal arrangement; ΔE represents the entropy error of the arrangement.

6. A load shedding data analysis system for pumped storage power stations based on signal decomposition and reconstruction, characterized in that, include: Data acquisition module, data processing module, and data output module; The data acquisition module acquires a simulation signal with boundary conditions that are basically consistent with the measured pressure signal during load shedding, and calculates the entropy value E of the simulation signal arrangement. 仿真 The system acquires the measured pressure signal x(t) of the pump turbine under load shedding and uses it as input information for VMD decomposition. The data processing module determines the number of VMD decomposition modes using the center frequency method, performs VMD decomposition on the measured pressure signal under load shedding, obtains several intrinsic mode functions (IMFs), and calculates the mutual information value M of each IMF. k Then normalize the mutual information value τ. k The weights of the reconstructed signal are used to obtain the reconstructed signal J(t). CEEMDAN decomposition is then performed on the reconstructed signal to obtain several components C, and their permutation entropy values ​​are calculated. These components are then superimposed to obtain the measured secondary-processed signal y, which matches the permutation entropy of the simulated signal. n (t); The data output module outputs a pressure pulsation signal, which is the difference between the measured signal and the measured secondary processed signal.

Images