Real-time estimation method and system for pump performance
By constructing an analytical model using a generalized similarity law model and sparse data, the problem of insufficient accuracy and robustness of existing virtual sensing technologies in pump performance estimation is solved, enabling efficient and accurate pump performance estimation on resource-constrained equipment, and applicable to scenarios without flow sensors.
Patent Information
- Application Number
- CN202511493804.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-20
- Publication Date
- 2026-01-20
AI Technical Summary
Existing virtual sensing technologies suffer from insufficient accuracy and poor robustness when estimating pump performance. They are particularly inadequate for engineering requirements when operating over a wide range of conditions or deviating from the design conditions. Furthermore, existing methods require massive amounts of data and large storage resources, making real-time computation impossible on resource-constrained devices.
A generalized similarity law model is adopted, and an analytical model is constructed using sparsely distributed performance test data. The generalized scaling factor function k(r) is used to map the benchmark performance curve to the performance curve at any target frequency or speed. Combined with the optimization algorithm to identify the coefficients, real-time accurate estimation is achieved.
It achieves high-precision estimation across the entire frequency range, especially improving the accuracy of virtual flow measurement, reducing hardware costs, with fast calculation speed and low resource consumption, strong physical consistency of the model, and the ability to identify abnormal operating conditions and avoid absurd results.
Smart Images
Figure CN121365513A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of digital twinning and fluid machinery, and particularly relates to a real-time estimation method and system for pump performance. BACKGROUND
[0002] Variable frequency speed regulation water pumps are widely used in industry, agriculture and municipal water supply. Real-time acquisition of the running flow of the water pump is crucial for system optimization control and energy saving. However, in many application scenarios, due to reasons such as cost, installation conditions or maintenance difficulty, flow sensors are not installed, and therefore the development of virtual sensing technology without flow meters has become an important direction in the industry.
[0003] There are two major difficulties in existing virtual sensing technology. Firstly, the physical model based on the ideal similarity law has physical interpretability, but it ignores complex fluid dynamics phenomena such as Reynolds number effect and efficiency decay, resulting in that the estimation accuracy cannot meet the engineering requirements when the frequency or speed deviates from the design condition. Secondly, the method based on multi-dimensional lookup table (LUT) or pure "black box" data model (such as neural network) sacrifices the physical interpretability in order to improve the accuracy through brute force fitting. However, this not only requires a large amount of test data and huge device storage resources, but more critically, the model is essentially discrete and unstructured, and has no reliable extrapolation ability and poor robustness.
[0004] Therefore, there is an urgent need in the art for a next-generation virtual sensing technology that can not only inherit the structure and extrapolation potential of the physical model, but also accurately compensate for physical deviations through data-driven methods, and the model itself is an analytical expression and is light enough to facilitate real-time calculation on resource-constrained devices. SUMMARY
[0005] According to a first aspect of an embodiment of the present application, a real-time estimation method for pump performance is provided, comprising the following steps: Offline modeling stage: Obtain a sparse distribution performance test data set of a pump unit at at least two different operating frequencies or operating speeds, and the data set includes flow, head and power data of multiple operating points; Determine a reference performance curve equation based on performance data at a reference frequency or reference speed; Construct a generalized similarity law model, and map the reference performance curve equation to a performance curve equation at any target frequency or speed through a set of parameterized generalized scaling factor functions k(r) with normalized frequency ratio r as the independent variable; wherein r=F / F_base or r=n / n_base, F and n are the target operating frequency and speed respectively, and F_base and n_base are the reference operating frequency and speed respectively; The analytical form of the generalized scaling factor function k(r) and its coefficients are identified by an optimization algorithm using the performance test data set; The online estimation stage: The current operating frequency F_field or operating speed n_field of the pump set is obtained in real time, as well as a real-time operating parameter, which is the head H_field or power P_field measured in real time; The real-time frequency ratio r_field=F_field / F_base or r_field=n_field / n_base is calculated; The real-time frequency ratio r_field is substituted into the identified generalized scaling factor function k(r) to calculate the current scaling factor value in real time; Using the current scaling factor value, the algebraic operation of the reference performance curve equation and the generalized similarity law model is used to generate the accurate performance curve equation under the current working condition in real time; The real-time operating parameter is substituted into the accurate performance curve equation generated in real time to analytically solve at least one target estimated parameter.
[0006] Further, the mapping expression of the generalized similarity law model for the head-flow (H-Q) performance relationship is: Q_pred=Q_ref_H*k_HQ(r); H_pred=H_ref*[k_H(r)]^2; Where (Q_ref_H, H_ref) is a reference point on the reference H-Q curve, and k_HQ(r) and k_H(r) are a pair of generalized scaling factor functions associated with the H-Q relationship.
[0007] Further, the mapping expression of the generalized similarity law model for the power-flow (P-Q) performance relationship is: Q_pred=Q_ref_P*k_PQ(r); P_pred=P_ref*[k_P(r)]^3; Where (Q_ref_P, P_ref) is a reference point on the reference P-Q curve, and k_PQ(r) and k_P(r) are a pair of generalized scaling factor functions associated with the P-Q relationship.
[0008] Further, the generalized scaling factor function k(r) is parameterized by a polynomial function, and its formula is: k(r)=a_m*r^m+...+a_1*r+a_0, where m is the polynomial order, and a_m,...,a_0 are the coefficients identified by the optimization algorithm.
[0009] Further, when the real-time operating parameter is the head H_field, the target estimated parameter is the flow Q_field, which is obtained by substituting the head H_field into the real-time generated head-flow accurate performance curve equation.
[0010] Further, when the real-time operating parameter is the power P_field, the target estimated parameter includes the flow Q_field and the head H_field; the solving step includes: First, substitute the power P_field into the real-time generated power-flow accurate performance curve equation, and analytically solve the flow Q_field; Then, substitute the flow Q_field into the real-time generated head-flow accurate performance curve equation, and analytically calculate the head H_field.
[0011] Further, after the offline modeling stage, there is also an analysis step: verifying whether the power-flow curve generated by the generalized similarity law model has monotonicity in the concerned operating interval, and retaining the curve with monotonicity to ensure the uniqueness of the online inverse solution flow value.
[0012] Further, when the input real-time operating parameter has no intersection with the real-time generated accurate performance curve equation in physics, an no solution flag is output to indicate an abnormal operating condition.
[0013] According to the second aspect of the embodiment of the present application, a real-time estimation system of pump performance is provided for executing the real-time estimation method of pump performance of the first aspect, comprising: a data input interface for receiving an operating frequency or rotating speed signal, and a head or power signal from at least one physical sensor; a processing unit configured to execute the steps of the online estimation stage; a data output interface for outputting the estimated flow, head, and power signals; When only a power sensor is equipped, the system constitutes a virtual sensor system without pressure sensor, which can simultaneously output virtual flow and virtual head signals.
[0014] According to the third aspect of the embodiment of the present application, a computer readable medium with non-volatile program code executable by a processor is provided, which makes the processor run the real-time estimation method of pump performance of the first aspect.
[0015] The real-time estimation method and system of pump performance according to the embodiment of the present application have the following beneficial effects: High precision: through data learning to automatically correct the deviation of physical law, high-precision estimation is realized in the whole frequency range, and the accuracy of virtual flow measurement is greatly improved.
[0016] Low cost: only a few frequency data sets are needed to model. It supports "no pressure sensor" solution, which estimates flow and head only by motor power and frequency, greatly saving hardware cost.
[0017] Fast and resource-saving: online calculation is only simple algebraic operation without table lookup, which is extremely fast and can be directly embedded into low-end controller.
[0018] Physically consistent and more reliable: the model is rooted in physical laws, the output is reasonable, and it can automatically identify abnormal input to avoid absurd results, which is highly reliable.
[0019] It is to be understood that both the foregoing general description and the following detailed description are exemplary and intended to provide further explanation of the subject technology claimed. BRIEF DESCRIPTION OF DRAWINGS
[0020] Figure 1 A flow chart of a real-time estimation method of pump performance according to an embodiment of the present application. DETAILED DESCRIPTION
[0021] The preferred embodiments of the present application will be described in detail with reference to the accompanying drawings, which further clarify the present application.
[0022] First, a real-time estimation method of pump performance according to an embodiment of the present application will be described. Figure 1 The real-time estimation method of pump performance according to an embodiment of the present application is used for high-fidelity performance modeling and real-time flow estimation of pump units driven by a frequency converter without a flow sensor, based on deep fusion of physical priori and data-driven, which has a wide range of application scenarios.
[0023] Terminology explanation about operating frequency and operating speed: the core of the performance estimation method proposed in the present application is to establish an analytical transformation model driven by the normalized ratio r, i.e. r=F / F_base or r=n / n_base. Those skilled in the art know that for asynchronous motor pump units driven by a frequency converter, since the motor slip varies with load, the operating frequency (F) and the actual speed (n) are not strictly linearly related, so there is a difference in the numerical value between the frequency ratio and the speed ratio.
[0024] A key technical feature of the present application is that the parameters of the generalized scaling factor function k(r) are obtained by identifying pump unit performance data containing all system-level physical effects (including but not limited to motor slip, Reynolds number effect, efficiency decay, etc.). Therefore: When the frequency ratio (r=F / F_base) which is easy to measure is selected as the model input, the non-ideal relationship between frequency and actual performance (including the effect of slip) is automatically learned and fixed into the coefficients of the scaling factor function k(r) during the identification process.
[0025] When the speed ratio (r = n / n_base) can be directly obtained, the method is also directly applicable.
[0026] In summary, whether the frequency ratio or the speed ratio is used, the present application can establish an accurate mapping model through data-driven manner, which embodies the robustness of the present method and the convenience of practical engineering application.
[0027] As shown in Figure 1 The real-time estimation method for pump performance of the embodiment of the present application has the following steps: Offline modeling stage: Obtain a sparse distribution performance test data set of the pump unit at at least two different operating frequencies or operating speeds, the data set including flow rate, head and power data of multiple operating points; Determine a reference performance curve equation based on performance data at a reference frequency or reference speed; Construct a generalized similarity law model, map the reference performance curve equation to a performance curve equation at any target frequency or speed through a set of parameterized generalized scaling factor functions k(r) with normalized frequency ratio r as the independent variable; wherein r = F / F_base or r = n / n_base, F and n are the target operating frequency and speed respectively, F_base and n_base are the reference operating frequency and speed respectively; wherein the generalized scaling factor function is used to modify the fixed scaling relationship of the ideal similarity law; Use the performance test data set to identify the analytical form and coefficients of the generalized scaling factor function k(r) through an optimization algorithm; Online estimation stage: Real-time obtain the current operating frequency F_field or operating speed n_field of the pump unit, and a real-time operating parameter, the real-time measurement parameter being head H_field or power P_field; Calculate the real-time frequency ratio r_field = F_field / F_base or r_field = n_field / n_base; Substitute the real-time frequency ratio r_field into the identified generalized scaling factor function k(r) to real-time calculate the current scaling factor value; Use the current scaling factor value, combine the algebraic operation of the reference performance curve equation and the generalized similarity law model, and real-time analytically generate an accurate performance curve equation at the current operating point; Substitute the real-time operating parameter into the real-time generated accurate performance curve equation to analytically solve at least one target estimated parameter.
[0028] Further, in the present embodiment, the generalized similarity law model for the head-flow (H-Q) performance relationship is expressed as: Q_pred=Q_ref_H*k_HQ(r); H_pred=H_ref*[k_H(r)]^2; where (Q_ref_H, H_ref) is a reference point on the baseline H-Q curve, k_HQ(r) and k_H(r) are a pair of generalized scaling factor functions associated with the H-Q relationship.
[0029] Further, in the present embodiment, the generalized similarity law model for the power-flow (P-Q) performance relationship is expressed as: Q_pred=Q_ref_P*k_PQ(r); P_pred=P_ref*[k_P(r)]^3; where (Q_ref_P, P_ref) is a reference point on the baseline P-Q curve, k_PQ(r) and k_P(r) are a pair of generalized scaling factor functions associated with the P-Q relationship.
[0030] Further, in the present embodiment, the generalized scaling factor function k(r) is parameterized by a polynomial function, whose formula is k(r) = a_m*r^m +... + a_1*r + a_0, where m is the polynomial order, and a_m,..., a_0 are coefficients identified by an optimization algorithm.
[0031] Further, in the present embodiment, when the real-time operating parameter is head H_field, the target estimated parameter is flow Q_field, which is obtained by substituting the head H_field into the real-time generated head-flow accurate performance curve equation.
[0032] Further, in the present embodiment, when the real-time operating parameter is power P_field, the target estimated parameters include flow Q_field and head H_field; the solving steps include: First, substitute the power P_field into the real-time generated power-flow accurate performance curve equation to analytically solve for the flow Q_field; Then, substitute the flow Q_field into the real-time generated head-flow accurate performance curve equation to analytically calculate the head H_field.
[0033] Further, in the present embodiment, after the offline modeling stage, there is also an analysis step: verifying whether the power-flow curve generated by the generalized similarity law model has monotonicity within the concerned operating interval, and retaining the monotonic curve to ensure the uniqueness of the online inverse flow value.
[0034] Further, in the present embodiment, when the input real-time operating parameter and the real-time generated accurate performance curve equation have no intersection in physics, an output no solution flag is output to indicate an abnormal working condition.
[0035] A specific embodiment application is exemplified (taking the estimation of the flow rate of a pump as an example): 1. Obtain the reference model and sparse test data First, a reference frequency is selected, for example, F_base=50Hz. The complete performance curve at this frequency is obtained through conventional testing means, and is fitted into a mathematical equation as the reference performance manifold Γ_base. For example, the head-flow curve can be fitted into a quadratic polynomial: H_base=A*Q_base^2+B*Q_base+C, where A, B, and C are known coefficients. At the same time, at least one (i.e., N≥2 total frequency number) different from the reference frequency, its performance test data is obtained. The performance test data contains several working condition points (F_i, Q_i, H_i, P_i) sufficient to represent the performance characteristics at this speed.
[0036] 2. Define a generalized similarity law model We define a generalized similarity law model to describe how the performance curve changes from the reference frequency F_base to any target frequency F. This model modifies the ideal similarity law by introducing a set of learnable generalized scaling factor functions k(r). Taking the head and flow as an example: Q_pred=Q_ref*k_Q(r); H_pred=H_ref*[k_H(r)]^2; Where r=F / F_base is the normalized frequency ratio. The ideal similarity law corresponds to the special case of k_Q(r)=r and k_H(r)=r. The core of the present invention is to learn the true form of these two functions k_Q(r) and k_H(r).
[0037] 3. Parameterize the scaling factor function and identify A key insight is that although the underlying physical process is complex, its macroscopic effect on the scaling factor can often be approximated by a relatively smooth low-order function. Without loss of generality, we assume that k_H(r) can be efficiently parameterized by a second-order polynomial: k_H(r)=a*r^2+b*r+c; Where {a, b, c} are the polynomial coefficients to be identified.
[0038] Next, a loss function L is constructed to quantify the total error between the model's predicted values and all available experimental data points. For example, for head, the loss function is the sum of the square of the difference between the predicted and true values at all measured points: L({a,b,c})=Σ[H_pred(i)-H_exp(i)]^2; where H_exp(i) is the measured head at the i-th experimental point, and H_pred(i) is the model's predicted head. H_pred(i) is calculated as follows: First, the scaling factors k_Q(r_i) and k_H(r_i) are calculated based on the frequency F_i of the i-th experimental point. Next, the base performance curve equation Γ_base is analytically transformed into a performance curve prediction equation Γ_pred(i) at the current frequency F_i by algebraic substitution using these scaling factors. Finally, the measured flow rate Q_exp(i) of this experimental point is substituted into this prediction equation Γ_pred(i), and the predicted head H_pred(i) is calculated.
[0039] Using all the sparse test data obtained in Step 1, a standard numerical optimization method (or optimization algorithm) is used to identify the optimal polynomial coefficients {a, b, c} by minimizing the loss function L. The numerical optimization method can include but is not limited to gradient-based algorithms (such as gradient descent, conjugate gradient), or Newton-based algorithms, or coordinate rotation methods. The appropriate algorithm can be selected by the skilled person according to the requirements of computational efficiency and convergence. Through this step, the specific analytical form of the function k_H(r) (for example, k_H(r)=0.8*r^2+0.15*r+0.05) is completely fixed. Similarly, k_Q(r) can be identified.
[0040] 4. Online real-time estimation During device operation, the controller obtains the operating frequency F_field and operating head H_field in real time.
[0041] a) Calculate the real-time frequency ratio r_field=F_field / F_base.
[0042] b) Substitute r_field into the identified scaling factor function, for example, k_H(r_field)=0.8*r_field^2+0.15*r_field+0.05, to directly analytically calculate the current scaling factor value k_H_field.
[0043] c) Using the scaling factor value and the base model Γ_base, a real-time performance curve equation Γ_field under the current working condition is constructed by the generalized similarity law H = H_base * [k_H_field]^2, combined with H_base = A * Q_base^2 + B * Q_base + C.
[0044] d) The real-time measured operating head H_field is substituted into the real-time performance curve equation Γ_field, and the only unknown quantity is solved, that is, the final estimated flow Q_field.
[0045] The entire online estimation process only involves a few algebraic operations, and the calculation speed is extremely fast, with almost no storage requirements, which is very suitable for embedded systems.
[0046] To verify the superiority of the method of the present application compared with the prior art, and to show its high precision and comprehensive performance in practical engineering applications, the applicant has carried out a series of comparative tests.
[0047] 1. Experimental setup Experimental system: The data comes from a closed-loop test bench composed of a 50-185F type centrifugal pump, a frequency converter and high-precision sensors.
[0048] Data set: Performance data of the pump in a wide frequency range of 60Hz to 200Hz (frequency interval of 10Hz, total data of 15 frequency segments) are collected. Among them, the data of three sparse frequencies of 60Hz, 130Hz and 200Hz are selected as the training set, and the data of the remaining frequencies are selected as the test set.
[0049] 2. Comparison schemes and evaluation tasks Based on the above data set, the PiiF model of the present application and three benchmark models representing pure physics, advanced engineering practice and pure data-driven respectively are constructed and compared: the classic similarity law (IAL), the quadratic curve fitting based on coefficient interpolation (QF-CI), and the multivariate polynomial regression (MPR).
[0050] We evaluated all models in two key dimensions: (1) Core application: virtual flow sensing, i.e. solving flow Q from (F, H); (2) Basic ability: forward performance modeling, i.e. predicting head H from (F, Q).
[0051] To ensure that the evaluation results can reflect the application value of each model in practical engineering, this comparative test focuses on the preferred operating range (POR) of the pump unit, which is the key to ensure the long-term, efficient and safe operation of the equipment.
[0052] The evaluation index adopts root mean square error (RMSE) and mean absolute percentage error (MAPE).
[0053] 3. Experimental results The prediction accuracy results of each model on the test set are summarized in Table 1 below.
[0054] Table 1: Accuracy comparison of different models in virtual sensing and forward modeling tasks
[0055] Note 1: In the (F, H)→Q task, the asterisk (*) indicates that the PiiF model of the present application successfully identified and labeled 2 query points that had no real solution because the input (F, H) did not conform to the physical law, and these points were incorrectly estimated in other models.
[0056] Note 2: In the forward head prediction task, the MAPE accuracy of PiiF and MPR models is basically the same.
[0057] 4. Experimental conclusions The experimental data in Table 1 clearly reveals the comprehensive technical advantages of the PiiF model of the present application: Significant advantage in core application (virtual flow sensing): In the task of estimating flow from head, the MAPE error of the PiiF model of the present application is only 8.57%, which is significantly better than all comparison schemes, reducing by nearly 50% compared to the sub-optimal MPR model. More importantly, the PiiF model can automatically identify abnormal working condition points (2 in this case) that cannot be solved due to input parameters violating physical constraints, demonstrating its excellent physical consistency and robustness. This feature is not possessed by pure data-driven methods, fundamentally solving the problem of unreliable prediction in key inverse solving tasks.
[0058] Maintain top level in basic modeling capability: In the task of predicting head in the forward direction, the accuracy of the PiiF model is completely at the same level as the performance of the pure "black box" model (MPR), showing equal performance. This shows that the present application has not sacrificed the basic fitting accuracy while introducing strong physical constraints and interpretability.
[0059] Achieve unified breakthrough in precision and physical consistency: Existing technologies either perform well in forward fitting like MPR, but significantly decrease in performance in key inverse applications and lack physical checking capability, or have clear physical structure or engineering logic like IAL and QF-CI, but have serious accuracy problems. Only the PiiF model of the present application successfully combines the structural advantages of physical models and the precision advantages of data-driven methods, and is the only solution that performs top (S-Tier) in both forward modeling and inverse sensing dimensions.
[0060] In summary, this invention solves the problem of existing technologies being unable to balance accuracy, physical consistency, and application flexibility through an innovative analytical physical information framework. Its comprehensive technical advantages and great practical value are fully supported by experimental data.
[0061] The method described in this invention also provides a computationally highly efficient implementation for virtual sensing in scenarios where even pressure sensors are not installed. Its online estimation process is as follows: Online real-time acquisition: The controller acquires the operating frequency F_field and the input power P_field read from the frequency converter in real time.
[0062] Constructing real-time performance equations: Using the identified generalized scaling factor functions (e.g., k_P(r), k_PQ(r), k_H(r), k_HQ(r)) corresponding to the power-flow and head-flow relationships, respectively, and the baseline performance curves, two core performance equations at the current frequency F_field are generated in real-time and analytically through algebraic operations: The power-flow equation (Γ_P-Q): P=f_P(Q) describes how the power P changes with the flow rate Q at this frequency.
[0063] The head-flow equation (Γ_H-Q): H=f_H(Q) describes how the head H changes with the flow rate Q at this frequency.
[0064] Serial analytical solution: a) Step 1: Inverse flow solution. Substitute the real-time acquired power value P_field into the real-time generated power-flow equation P_field=f_P(Q). Since this equation is an analytical expression of Q (usually a polynomial), the unique, physically meaningful flow value Q_field can be analytically solved directly using the quadratic formula or simple numerical methods.
[0065] b) Second step: Forward modeling of head. Substitute the flow rate value Q_field obtained in the previous step into the real-time generated head-flow rate equation H=f_H(Q_field), and through a simple algebraic calculation, directly obtain the final estimated head H_field.
[0066] This serial analytical solution approach avoids the complex iterative solution of simultaneous equations, simplifying the entire computation process into two independent, deterministic analytical operations. This not only greatly improves computational speed and robustness but also enables this method to achieve high-performance, pressure-free virtual sensing on the most basic embedded controllers with extremely low resource consumption.
[0067] Applicability notes: It should be noted that the reliability of the method described in this embodiment is directly related to the functional characteristics of the power-flow curve (P-Q curve).
[0068] In a preferred embodiment, the method is particularly suitable for pump system whose power-flow curve is monotonic in the main operating range. For such systems, the power value and flow value have a one-to-one mapping relationship, thereby ensuring the uniqueness and accuracy of the inverse-solved flow Q_field.
[0069] Therefore, the method of the present application can include an offline analysis step, i.e. after identifying the performance model, the generated power-flow curve is analyzed to confirm its monotonicity in the target working range. The technical solution of this embodiment provides a very efficient and economical virtual sensor solution for pump systems with this monotonicity. For complex systems whose P-Q curve may have hump or flat non-monotonic behavior in some areas, it is recommended to use the technical solution described in other embodiments of the present application (such as head-based estimation) to ensure the best estimation performance.
[0070] According to the second aspect of the embodiment of the present application, a real-time estimation system of pump performance is provided for executing the real-time estimation method of pump performance of the first aspect, comprising: a data input interface for receiving the operating frequency or speed signal and the head or power signal from at least one physical sensor; a processing unit configured to execute the steps of the online estimation phase; a data output interface for outputting the estimated flow, head, and power signals; When only equipped with a power sensor, the system constitutes a virtual sensor system without pressure sensor, which can output virtual flow and virtual head signals simultaneously.
[0071] According to the third aspect of the embodiment of the present application, a computer readable medium having non-volatile program code executable by a processor is provided, and the program code causes the processor to run the real-time estimation method of pump performance of the first aspect.
[0072] Where the readable storage medium can be a computer-readable storage medium, also can be a communication medium. The communication medium includes any medium that facilitates the transfer of computer program from one place to another. The computer storage medium can be any available medium that can be accessed by a general or special purpose computer. For example, the readable storage medium is coupled to the processor, so that the processor can read information from the readable storage medium, and can write information to the readable storage medium. Of course, the readable storage medium can also be an integral part of the processor. The processor and the readable storage medium can be located in the application specific integrated circuit. In addition, the application specific integrated circuit can be located in the device. Of course, the processor and the readable storage medium can also exist as discrete components in the communication device. The readable storage medium can be read-only memory (ROM), random access memory (RAM), CD-ROM, magnetic tape, floppy disk and optical data storage device, etc. The present application also provides a program product, which includes execution instructions stored in a readable storage medium. The at least one processor of the device can read the execution instructions from the readable storage medium, and the at least one processor executes the execution instructions so that the device implements the real-time estimation method of pump performance provided by various embodiments described above. In the above embodiment of the device, it should be understood that the processor can be a central processing unit (English: Central Processing Unit, for short: CPU), but also can be other general-purpose processors, digital signal processors (English: Digital Signal Processor, for short: DSP), application specific integrated circuits (English: Application Specific Integrated Circuit, for short: ASIC) and the like. The general-purpose processor can be a microprocessor or the processor can be any conventional processor or the like. The steps of the method disclosed in the present application can be directly embodied as the execution of the hardware processor, or the combination of the hardware and software modules in the processor.
[0073] It should be noted that in the present specification, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or includes elements inherent to such process, method, article or device. Without more limitations, the elements defined by the statement "include" do not exclude the presence of other identical elements in the process, method, article or device including the elements.
[0074] While the application has been described in detail and with reference to specific preferred embodiments thereof, it will be apparent to one skilled in the art that various modifications and alternatives can be employed without departing from the spirit and scope of the application. Accordingly, the scope of the application should be determined by the appended claims and their equivalents.
Claims
1. A method of real-time estimation of pump performance, characterized in that, The method comprises the following steps: An offline modeling stage: Obtaining a sparse distributed performance test data set of a pump unit at at least two different operating frequencies or operating speeds, the data set comprising flow rate, head and power data of multiple operating points; Determining a reference performance curve equation based on performance data at a reference frequency or reference speed; Constructing a generalized similarity law model to map the reference performance curve equation to a performance curve equation at any target frequency or speed through a set of parameterized generalized scaling factor functions k(r) with normalized frequency ratio r as the independent variable; wherein r=F / F_base or r=n / n_base, F and n are the target operating frequency and speed respectively, and F_base and n_base are the reference operating frequency and speed respectively; Using the performance test data set, identifying the analytical form of the generalized scaling factor functions k(r) and their coefficients through an optimization algorithm; An online estimation stage: Real-time obtaining of the current operating frequency F_field or operating speed n_field of the pump unit, and a real-time operating parameter, which is the head H_field or power P_field; Calculating the real-time frequency ratio r_field=F_field / F_base or r_field=n_field / n_base; Substituting the real-time frequency ratio r_field into the identified generalized scaling factor functions k(r) to real-time calculate the current scaling factor value; Using the current scaling factor value, combining the reference performance curve equation and the algebraic operation of the generalized similarity law model to real-time analytically generate an accurate performance curve equation at the current operating point; Substituting the real-time operating parameter into the real-time generated accurate performance curve equation to analytically solve at least one target estimated parameter.
2. The method of real-time estimation of pump performance of claim 1, wherein, The mapping expression of the generalized similarity law model for the head-flow (H-Q) performance relationship is: Q_pred=Q_ref_H*k_HQ(r); H_pred=H_ref*[k_H(r)]^2; Where (Q_ref_H, H_ref) is a reference point on the reference H-Q curve, and k_HQ(r) and k_H(r) are a pair of generalized scaling factor functions associated with the H-Q relationship.
3. The method of real-time estimation of pump performance of claim 2, wherein, The mapping expression of the generalized similarity law model for the power-flow (P-Q) performance relationship is: Q_pred=Q_ref_P*k_PQ(r); P_pred=P_ref*[k_P(r)]^3; Where (Q_ref_P, P_ref) is a reference point on the reference P-Q curve, and k_PQ(r) and k_P(r) are a pair of generalized scaling factor functions associated with the P-Q relationship.
4. The method of real-time estimation of pump performance of claim 1, wherein, The generalized scaling factor functions k(r) are parameterized by a polynomial function, and the formula is: k(r)=a_m*r^m+...+a_1*r+a_0, where m is the polynomial order, and a_m,...,a_0 are the coefficients identified by the optimization algorithm.
5. The method of real-time estimation of pump performance of claim 2, wherein, When the real-time operating parameter is head H_field, the target estimated parameter is flow Q_field, which is obtained by substituting head H_field into the real-time generated head-flow precise performance curve equation.
6. The method of real-time estimation of pump performance of claim 3, wherein, When the real-time operating parameter is power P_field, the target estimated parameter includes flow Q_field and head H_field; the solving step includes: First, substitute power P_field into the real-time generated power-flow precise performance curve equation, and analytically solve for flow Q_field; Then, substitute flow Q_field into the real-time generated head-flow precise performance curve equation, and analytically calculate head H_field.
7. The method of real-time estimation of pump performance of claim 6, wherein, After the offline modeling phase, an analysis step is further included: verifying whether the power-flow curve generated by the generalized similarity law model has monotonicity within the concerned operating interval, and retaining the curve with monotonicity to ensure the uniqueness of the online inverse solution flow value.
8. The method of real-time estimation of pump performance of claim 1, wherein, When the input real-time operating parameter has no intersection with the real-time generated precise performance curve equation in physics, an no solution flag is output to indicate an abnormal operating condition.
9. A system for real-time estimation of pump performance for performing the method of real-time estimation of pump performance according to any one of claims 1 to 8, characterized in that Comprise: a data input interface for receiving operating frequency or rotating speed signals, and head or power signals from at least one physical sensor; a processing unit configured to perform the steps of the online estimation phase; a data output interface for outputting estimated flow, head, and power signals; When only equipped with a power sensor, the system constitutes a virtual sensor system without pressure sensor, which can simultaneously output virtual flow and virtual head signals.
10. A computer readable medium having non-transitory program code executable by a processor, the program code comprising instructions for: The program code causes the processor to run the real-time estimation method of pump performance according to any one of claims 1-8. The program code causes the processor to run the real-time estimation method of pump performance according to any one of claims 1-8.