Hybrid driving and cascaded error attribution method for multi-subsystem parameter identification of hydroelectric generating units

Abstract

The application discloses a kind of fusion driving and cascading error attribution hydroelectric generating set multi-subsystem parameter identification method, comprising: construct the high-fidelity simulation model of four cascading subsystems of local control unit LCU, speed regulator, electro-hydraulic servo system, hydroelectric unit;Design dynamic fusion coefficient with iteration attenuation, measured signal and model simulation signal are fused as the input of each subsystem, realize the transition from data-driven to model consistency-driven;Each subsystem output error is decomposed into upstream propagation error and its intrinsic error, and the optimization objective function with intrinsic error as the core is reconstructed;Improved particle swarm optimization algorithm, based on intrinsic error dynamic distribution search weight and introduce propagation error suppression factor, realize multi-subsystem parameter collaborative optimization.The application can effectively suppress cascading error accumulation, improve the physical interpretability and identification accuracy of parameter, and is suitable for hydroelectric generating set digital twin and online control scene.

Description

Technical Field

[0001] This invention relates to the field of power system automation, and specifically to a method for identifying parameters of multiple subsystems in hydropower units that integrates drive and cascade error attribution. Background Technology

[0002] With the advancement of the "dual carbon" goals, intermittent renewable energy sources such as wind power and photovoltaics are being integrated into the power grid on a large scale. Coupled with the rapid development of long-distance ultra-high-voltage transmission networks, the power system faces severe challenges in balancing supply and demand and maintaining frequency stability. Hydropower units, with their advantages of fast response, flexible regulation, and large single-unit capacity, have become an indispensable regulating resource for power grid peak shaving, frequency regulation, and emergency backup. However, the hydropower unit regulation system is a complex dynamic system encompassing mechanical, hydraulic, electrical, and hydraulic multi-physics fields. Under complex operating conditions, it exhibits strong nonlinearity, strong coupling, and multi-timescale characteristics, making its high-precision modeling and parameter identification significantly more difficult than that of conventional thermal power units.

[0003] In the field of hydropower unit modeling, the system structure can usually be determined based on physical mechanisms (such as Newton's laws of motion and fluid dynamics equations). However, key parameters (such as the governor integral time constant, the relay response time constant, and the water flow inertial time constant) are affected by factors such as equipment wear, installation deviations, and operating environment, making it difficult to obtain them directly through theoretical calculations or factory calibration. Parameter identification must rely on field measurement data. In early research, least squares and its variants were widely used due to their simple principles and high computational efficiency. These methods construct an error function between the model output and the measured data, and iteratively solve for the optimal parameter solution using the gradient descent approach. However, for highly nonlinear objects like hydropower turbine regulation systems, the error function generally exhibits multi-peak characteristics, and field measurement data often contains non-differentiable factors such as sensor noise, electromagnetic interference, and even signal loss. This makes gradient-based methods prone to getting trapped in local optima, and the identification results are highly sensitive to the choice of initial values, making it difficult to meet engineering requirements in terms of stability.

[0004] To overcome the limitations of gradient-based methods, metaheuristic algorithms such as Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) have been gradually introduced into the field of hydropower unit parameter identification since 2010. These algorithms, by simulating the collaborative search behavior of biological populations, can explore the global parameter space without relying on gradient information, effectively solving nonlinear and multi-peak optimization problems, and achieving good results under simple operating conditions such as single-unit no-load and stable load operation. However, existing research has largely focused on improving the algorithms themselves, with insufficient attention paid to the multi-cascaded structural characteristics of hydropower unit regulation systems—the turbine regulation system is essentially a cascaded closed-loop system composed of four functionally defined subsystems: the Local Control Unit (LCU), the governor, the electro-hydraulic servo system, and the hydropower generation unit, with signals transmitted sequentially at each level. These subsystems are tightly physically coupled.

[0005] Existing parameter identification methods suffer from two typical defects due to neglecting this structural characteristic: First, the segmented independent identification method divides the system into four independent subsystems and independently identifies the parameters of each subsystem using the measured data of the corresponding measurement points. Although this reduces the complexity of the problem, it does not consider the signal transmission relationship between the subsystems, causing the modeling error of the upstream subsystem to propagate and amplify step by step in the subsequent cascading process, resulting in a significant deviation between the overall model output and the measured data. Second, the overall black box identification method treats the entire system as a single black box and only uses the terminal output error as the optimization target. Although this avoids error accumulation, the lack of physical constraints of intermediate subsystems leads to a large number of equivalent solutions in the parameter space, and the identification results often deviate from the physical reality, resulting in poor model generalization ability.

[0006] More importantly, the output deviation of downstream subsystems in a multi-cascaded system is essentially a superposition of "upstream input error propagation" and "self-parameter mismatch". If the output errors of each subsystem are directly accumulated to construct the objective function, the upstream error will be repeatedly included in the downstream, artificially amplifying the weight of the upstream error. This misleads the optimization algorithm to forcibly adjust the downstream parameters to compensate for the upstream error, ultimately resulting in an identification result that deviates significantly from the true physical parameters.

[0007] In summary, existing hydropower unit parameter identification technologies generally suffer from problems such as error accumulation, weak physical interpretability of parameters, and optimization direction deviation due to insufficient exploitation of the signal propagation structure characteristics of multi-cascade systems. These limitations make it difficult to meet the application requirements of high-fidelity models in engineering scenarios such as digital twins and online control. How to integrate measured data constraints with the internal signal transmission mechanism of the system during the identification process, and construct a collaborative identification method that balances local accuracy and overall consistency, has become a key technical bottleneck that urgently needs to be overcome. Summary of the Invention

[0008] The technical problem to be solved by the present invention is to provide a method for identifying multi-subsystem parameters of hydropower units that integrates drive and cascade error attribution, which can achieve high-precision and high-stability identification of multi-subsystem parameters.

[0009] To solve the above-mentioned technical problems, the technical solution adopted by the present invention includes: A simulation model is constructed comprising four subsystems: a local control unit (LCU), a speed governor, an electro-hydraulic servo system, and a hydroelectric power generation unit. These four subsystems are cascaded and connected in a closed loop: the LCU subsystem receives power commands and feedback signals, and outputs PWM pulse control signals to the speed governor subsystem; the speed governor subsystem integrates the PWM pulse control signals and outputs guide vane opening commands to the electro-hydraulic servo system subsystem; the electro-hydraulic servo system subsystem drives the guide vanes and outputs the actual opening signal to the hydroelectric power generation unit subsystem; the hydroelectric power generation unit subsystem outputs the unit's active power and feeds it back to the LCU subsystem, forming a closed loop; the LCU subsystem outputs PWM pulse control signals based on power deviation, denoted as the structure function. The speed controller subsystem uses an integral element to describe its dynamic characteristics, denoted as the structure function. The electro-hydraulic servo system subsystem uses a first-order inertial element to describe hydraulic hysteresis, denoted as the structure function. The hydropower generation unit subsystem employs a nonlinear turbine model and a high-order transfer function to simulate the elastic water hammer effect, denoted as the structure function. The structural definition based on the simulation model includes the vector of parameters to be identified for the physical parameters of each subsystem. The simulation model can output analog signals of each subsystem. Set a dynamic fusion coefficient that decays with the number of iterations. , where k is the iteration number. In the first iteration, the actual measured data set of the hydropower unit under the preset operating conditions is obtained as the input of the simulation model. In subsequent iterations, the simulated signals generated by each subsystem of the simulation model in the previous iteration are weighted and fused with the key signals of each subsystem obtained from the field measurement to generate the fused input signal of each subsystem in the current iteration. In each iteration, the generated input signal is input to the simulation model to perform forward signal propagation, and the total error of each subsystem output is calculated. The total error is decomposed into propagation error caused by upstream input deviation and intrinsic error caused only by current subsystem parameter deviation. The upstream error interference is removed to obtain the intrinsic error that only reflects its own parameter mismatch. The objective function is constructed with minimizing the intrinsic error as the core, and an improved particle swarm optimization algorithm is used to iteratively update the parameter vector to be identified. This includes: calculating the magnitude of the intrinsic error of each subsystem, and constructing a normalized adaptive weight based on the magnitude to scale the particle velocity piecewise, so that the subsystem with large error has a stronger search capability; calculating the ratio of the intrinsic error to the propagation error of each subsystem to construct a suppression factor, and multiplying the suppression factor by the parameter update amount to suppress the parameter update of the subsystem when the ratio approaches 0. When the convergence condition is met, the current parameter vector to be identified is output as the final parameter identification result.

[0010] As a further improvement to the method of the present invention, the parameter vector to be identified includes: The local control unit (LCU) subsystem includes the pulse period T and minimum pulse width. Maximum pulse width and proportionality coefficient The pulse width Adaptive adjustment based on power deviation:

[0011] In the formula, Due to power difference, The sign function is defined as follows: it takes the value 1 when the independent variable is greater than 0, and takes the value -1 when the independent variable is less than or equal to 0. The speed controller subsystem includes an opening given integral time constant. ; The electro-hydraulic servo system subsystem includes the relay response time constant Ty; The hydropower generation unit subsystem includes the water flow inertial time constant. and the time constant of the phase of water hammer waves .

[0012] As a further improvement to the method of the present invention, the structure function , , , The specific content is as follows: Including the control signal u(t) output by the LCU: ; Where M is the PWM output amplitude, ΔP is the difference between the actual unit active power and the target value, T is the pulse period, and k is the pulse number. For symbolic functions, their calculation rules are as follows: ; The dynamic characteristics are described using the transfer function of the integral element: ; Hydraulic hysteresis is described using the transfer function of a first-order inertial element: ; The elastic water hammer effect is simulated using a nonlinear turbine model and a higher-order transfer function. The nonlinear turbine model is as follows: ; ; Where M is the turbine torque, Q is the flow rate, α is the guide vane opening, n is the rotational speed, and H is the head; To address the elastic water hammer effect in the water diversion system, a higher-order transfer function is used to simulate the propagation and reflection characteristics of pressure waves in the pipeline. Its expression is: .

[0013] As a further improvement to the method of the present invention, the dynamic fusion coefficient The calculation formula is: ; Where k is the number of iterations. λ is the preset initial fusion coefficient, and λ is the preset attenuation coefficient.

[0014] As a further improvement to the method of the present invention, the calculation formula for the fused input signal is as follows: ; in For any subsystem in the simulation model, the fused input signal is... These are the key signals obtained from on-site measurements of this subsystem. This refers to the analog signal obtained by the subsystem in the previous iteration. If this is the first iteration, then .

[0015] As a further improvement to the method of the present invention, the propagation error is calculated as follows: ; Where i represents the i-th subsystem in the corresponding simulation model. This is the structure function corresponding to this subsystem. This is the vector of parameters to be identified for this subsystem. and These represent the key signals obtained from on-site measurements of the previous subsystem and the simulated signals obtained in the previous iteration, respectively.

[0016] As a further improvement to the method of the present invention, the objective function is constructed as follows: ; Where J is the intrinsic error Minimize the objective function: ; These are the weighting coefficients for each subsystem; For structural consistency constraints:

[0017] Where the summation index i represents the discrete time step, and Let and represent the output of the simulation model and the state variables of the actual system at time i, respectively. and Then, corresponding to the state variables of the previous moment, this formula forces the dynamic response structure of the model to be consistent with the real system by calculating the sum of the squares of the relative rates of change of the simulation and the actual deviations between adjacent moments.

[0018] As a further improvement to the method of the present invention, the improved particle swarm optimization algorithm is used to iteratively update the parameter vector to be identified. In each particle swarm iteration, only one complete cascaded signal forward propagation process is executed, and the objective function value is calculated synchronously without internal nested iteration.

[0019] The present invention also provides a multi-subsystem parameter identification system for hydropower units with fusion drive and cascade error attribution, comprising a processor and a computer-readable storage medium interconnected thereto, wherein the computer-readable storage medium stores a computer program, and the computer program is executed by the processor to implement the steps of the above-described multi-subsystem parameter identification method for hydropower units with fusion drive and cascade error attribution.

[0020] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for identifying parameters of a multi-subsystem hydropower unit based on fusion drive and cascade error attribution.

[0021] Compared with the prior art, the advantages of the present invention are as follows: This invention effectively solves the error accumulation problem in the identification of multi-cascaded systems by constructing a dynamic fusion mechanism of "measurement-simulation". Through the "propagation-intrinsic" error attribution decoupling strategy, it cuts off the interference of upstream errors on downstream parameter updates, avoiding repeated penalties and optimization direction shifts. By designing an adaptive particle swarm optimization algorithm based on intrinsic error feedback, it realizes the dynamic tilting of computing resources towards high-error subsystems and introduces a propagation error suppression factor to prevent overcorrection. Thus, while ensuring the physical interpretability of parameters, it significantly improves identification accuracy, convergence speed and engineering practicality. Attached Figure Description

[0022] Figure 1 A schematic diagram of the signal transmission process according to an embodiment of the present invention. Detailed Implementation

[0023] The present invention will be further described below with reference to the accompanying drawings and specific preferred embodiments, but this does not limit the scope of protection of the present invention.

[0024] The technical solution adopted in this embodiment includes: A simulation model is constructed comprising four subsystems: a local control unit (LCU), a speed governor, an electro-hydraulic servo system, and a hydroelectric power generation unit. These four subsystems are cascaded and connected in a closed loop: the LCU subsystem receives power commands and feedback signals, and outputs PWM pulse control signals to the speed governor subsystem; the speed governor subsystem integrates the PWM pulse control signals and outputs guide vane opening commands to the electro-hydraulic servo system subsystem; the electro-hydraulic servo system subsystem drives the guide vanes and outputs the actual opening signal to the hydroelectric power generation unit subsystem; the hydroelectric power generation unit subsystem outputs the unit's active power and feeds it back to the LCU subsystem, forming a closed loop; the LCU subsystem outputs PWM pulse control signals based on power deviation, denoted as the structure function. The speed controller subsystem uses an integral element to describe its dynamic characteristics, denoted as the structure function. The electro-hydraulic servo system subsystem uses a first-order inertial element to describe hydraulic hysteresis, denoted as the structure function. The hydropower generation unit subsystem employs a nonlinear turbine model and a high-order transfer function to simulate the elastic water hammer effect, denoted as the structure function. The structural definition based on the simulation model includes the vector of parameters to be identified for the physical parameters of each subsystem. The simulation model can output analog signals of each subsystem. Set a dynamic fusion coefficient that decays with the number of iterations. , where k is the iteration number. In the first iteration, the actual measured data set of the hydropower unit under the preset operating conditions is obtained as the input of the simulation model. In subsequent iterations, the simulated signals generated by each subsystem of the simulation model in the previous iteration are weighted and fused with the key signals of each subsystem obtained from the field measurement to generate the fused input signal of each subsystem in the current iteration. In each iteration, the generated input signal is input to the simulation model to perform forward signal propagation, and the total error of each subsystem output is calculated. The total error is decomposed into propagation error caused by upstream input deviation and intrinsic error caused only by current subsystem parameter deviation. The upstream error interference is removed to obtain the intrinsic error that only reflects its own parameter mismatch. The objective function is constructed with minimizing the intrinsic error as the core, and an improved particle swarm optimization algorithm is used to iteratively update the parameter vector to be identified. This includes: calculating the magnitude of the intrinsic error of each subsystem, and constructing a normalized adaptive weight based on the magnitude to scale the particle velocity piecewise, so that the subsystem with large error has a stronger search capability; calculating the ratio of the intrinsic error to the propagation error of each subsystem to construct a suppression factor, and multiplying the suppression factor by the parameter update amount to suppress the parameter update of the subsystem when the ratio approaches 0. When the convergence condition is met, the current parameter vector to be identified is output as the final parameter identification result.

[0025] The hydropower unit regulation system is a complex nonlinear system composed of multiple functional sub-modules with clearly defined physical meanings tightly coupled together. Constructing a high-fidelity simulation model is fundamental for parameter identification and experimentation, and its accuracy and applicability directly determine the reliability of subsequent experimental results. This embodiment first constructs a nonlinear simulation model based on mechanistic modeling, which can accurately reflect the characteristics of the hydropower unit in dynamic processes such as participating in automatic generation control (AGC) of the power grid. This model focuses on simulating the control logic of the power signal being directly fed back to the local control unit (LCU) in the AGC opening mode under grid-connected conditions. Its core components include four key parts: LCU, governor, electro-hydraulic servo system, and hydropower generation unit.

[0026] In AGC (Automatic Generation Control) mode, the LCU (Low-Cooler Unit) acts as the control command generator, receiving the deviation signal between the actual active power of the generator unit and the target power value. After processing, it sends control commands to the speed controller in the form of pulse width modulation (PWM) pulses. Let ΔP be the difference between the actual active power of the generator unit and the target value (unit: MW), then the control signal u(t) output by the LCU is defined as: ; Where M is the PWM output amplitude, ΔP is the difference between the actual unit active power and the target value, T is the pulse period, and k is the pulse number. For symbolic functions, their calculation rules are as follows: ; Pulse width The power deviation is adaptively adjusted and constrained by minimum and maximum pulse width thresholds. The specific calculation logic is as follows:

[0027] In the formula, Due to power difference, The sign function is defined as follows: it takes the value 1 when the independent variable is greater than 0, and takes the value -1 when the independent variable is less than or equal to 0.

[0028] The speed controller receives the PWM pulse signal from the LCU and converts it into a continuous control quantity through an integrator. Its dynamic characteristics in AGC opening mode are described by the following transfer function: ; In the formula, The integral time constant is given for the opening degree.

[0029] The electro-hydraulic servo system converts the electrical signal output by the speed governor into mechanical displacement to drive the guide vane. Considering the dynamic hysteresis characteristics of the hydraulic servo mechanism, its transfer function is expressed as: ; Where Ty is the reaction time constant of the relay.

[0030] To accurately capture nonlinear dynamics such as water hammer effect, a nonlinear turbine model is adopted, where the turbine torque M and flow rate Q are expressed as functions of guide vane opening α, rotational speed n, and head H: ; ; Where M is the turbine torque, Q is the flow rate, α is the guide vane opening, n is the rotational speed, and H is the head; To address the elastic water hammer effect in the water diversion system, a higher-order transfer function is used to simulate the propagation and reflection characteristics of pressure waves in the pipeline. Its expression is: .

[0031] in, It is the inertial time constant of the water flow. It is the time constant of the water hammer wave direction.

[0032] Based on the detailed modeling of the four core components—LCU, governor, electro-hydraulic servo system, and hydropower generation unit—this method constructs a simulation model that can highly reproduce the physical operating laws and nonlinear dynamic characteristics of the hydropower unit's regulation system. This high-fidelity model not only accurately depicts the entire response mechanism from power deviation sensing to guide vane mechanical action and power change, but also provides a reliable digital twin foundation for parameter identification of multi-stage closed-loop systems in subsequent chapters. This makes it possible to reproduce the system's dynamic behavior under real-world conditions and solve for key unknown parameters in a simulation environment.

[0033] Based on the above simulation model, the hydropower unit regulation system can be abstracted as a closed-loop structure consisting of four core subsystems cascaded together: the local control unit (LCU), the governor, the electro-hydraulic servo system, and the hydropower generation unit. The specific signal transmission flow of these subsystems is as follows: Figure 1 As shown, these four subsystems are denoted as subsystems A, B, C, and D, respectively. The parameter vector to be identified includes: The local control unit (LCU) subsystem includes the pulse period T and minimum pulse width. Maximum pulse width and proportionality coefficient The pulse width is adaptively adjusted according to the power deviation.

[0034] In the formula, Due to power difference, The sign function is defined as follows: it takes the value 1 when the independent variable is greater than 0, and takes the value -1 when the independent variable is less than or equal to 0. The speed controller subsystem includes an opening given integral time constant. ; The electro-hydraulic servo system subsystem includes the relay response time constant Ty; The hydropower generation unit subsystem includes the water flow inertial time constant. and the time constant of the phase of water hammer waves .

[0035] Due to the significant cascaded signal transmission and feedback coupling characteristics of the system, the output of each subsystem depends not only on its own parameters but also on the output of the previous subsystem or the feedback signal of the next subsystem. Given a parameter vector... Under these conditions, the dynamic response of the system is determined by the following cascade relationship: ; In the formula, These are the structure functions of the corresponding subsystems. The time series set output by the model simulation is denoted as... Correspondingly, the set of key signal time series of each subsystem obtained through on-site measurements is denoted as... The core objective of parameter identification is to find the optimal parameter vector given the system structure function and measured data. This enables the model simulation output With actual system response Minimize the deviation between them.

[0036] To quantify the fit between simulation and measured data, an evaluation function based on the integral of the absolute value of time error (ITAE) is constructed. This serves as the objective function for parameter identification. The function accumulates the model output throughout the dynamic process using an integral form. Compared with measured data The absolute error is expressed mathematically as follows: ; In the formula, This represents the time series of measured data. This indicates that the input variables and parameter vector are... The time series output by the time model, The length of the observation time window is denoted as . A smaller objective function value indicates that the identified parameters more accurately reflect the true physical characteristics of the system, and the higher the simulation accuracy of the model. In summary, the parameter identification problem of the hydropower unit regulation system is formalized as a multidimensional nonlinear optimization problem constrained by physical boundaries, i.e., in Under the premise of satisfying the actual value constraints, find the way to make By obtaining the parameter combination that achieves the global minimum, the dynamic characteristics of a multi-level closed-loop system can be accurately reproduced.

[0037] In the parameter identification process of multi-cascaded structures, the various sub-modules of the system form coupling relationships through the step-by-step transmission of signals. To effectively incorporate this structural information during parameter optimization while avoiding the instability caused by directly constructing a strongly coupled closed loop, a progressive coupling identification mechanism based on the fusion of real and analog signals is constructed.

[0038] In hydropower unit regulation systems, subsystems form significant structural coupling relationships through the cascading transmission of signals. The primary subsystem is influenced not only by internal feedback signals but also by external regulatory inputs. Therefore, during parameter identification, it is necessary to characterize both the effects of external inputs and the internal signal transmission mechanisms within a unified framework, thereby improving the model's ability to describe the dynamic characteristics of the actual system.

[0039] Therefore, an external excitation signal (U), i.e., command power, is introduced into the identification model. This signal is determined by the unit's operating conditions or regulation commands and is a known input quantity, which does not participate in the fusion and optimization calculations during the parameter identification process. Based on this, the system retains the feedback signal generated by the output of the final-stage subsystem, and constructs a unified input by fusing real data with the model's calculation results.

[0040] Specifically, for the feedback signal (D), its fusion form under the (k)th optimization iteration is defined as: ; in, This is the measured signal. The simulation results were obtained from the previous round of propagation. The fusion coefficient changes with iteration.

[0041] Based on this, the output of the primary subsystem (A) can be expressed as: ; in, These are the parameters to be identified for subsystem A.

[0042] For subsequent subsystems, signal transmission continues according to a cascaded structure, with a fusion mechanism between real and analog signals introduced at each stage. The specific process is as follows: ; ; ; ; ; The final system output is: ; Through the above-described step-by-step propagation process, the input signal completes an orderly transmission between each subsystem. The analog signal not only participates in error calculation as an output, but also participates in the internal propagation of the system as part of the input of the subsequent subsystem, thereby realizing the explicit introduction of structural coupling relationship.

[0043] To achieve a smooth transition in the identification process from being driven by real measurement data to being driven by structural simulation data, the fusion coefficients are... The scheduling design that changes with iteration can be expressed as: ; in, The maximum number of iterations, and These are preset parameters.

[0044] In the initial optimization phase, The values ​​are relatively large, and the inputs to each subsystem mainly consist of real observation data, thus ensuring that the response of a single subsystem closely approximates the actual system behavior; as iteration progresses, As the ratio of the calculation results within the model to the signal transmission gradually decreases, the system parameters gradually adapt to the overall dynamic characteristics under the cascaded structure.

[0045] Through the above mechanism, a unified modeling of external input drive and internal feedback propagation is achieved without explicitly solving the closed-loop steady state. This enables real data and model information to form a dynamic synergy during the identification process, thereby effectively improving the consistency and accuracy of multi-subsystem parameter identification while ensuring optimization stability.

[0046] Under the aforementioned structural coupling identification mechanism, each subsystem forms a unified dynamic response through step-by-step signal transmission. However, since the output of the preceding subsystem serves as the input for subsequent subsystems in the calculation, its error will continue to propagate during the cascading process. This results in the output deviation of the subsequent subsystem containing both its own modeling error and the superposition of errors from the preceding subsystem. In this case, directly accumulating and optimizing the output errors of each subsystem will cause the same error to be repeatedly counted at different levels, leading to error amplification and optimization direction shift.

[0047] To address the aforementioned issues, the error propagation process in a multi-cascaded system is modeled and analyzed. The output error of each subsystem is uniformly represented as: ; in, This represents the total error of the i-th subsystem. This error actually consists of two parts: one part comes from the subsystem's own parameter deviation, and the other part comes from the propagation effect of the upstream input error.

[0048] To distinguish the sources of error, the concept of "ideal input" is introduced. This means replacing the input of the current subsystem with the actual upstream signal while keeping the current subsystem parameters unchanged, thus constructing a propagation error that only reflects the influence of input deviation. For the i-th subsystem, its propagation error is defined as: ; in, and These represent the actual output and simulated output of the previous subsystem, respectively.

[0049] Based on the above definition, the total error of the subsystem can be decomposed into the sum of the propagation error and the intrinsic error, thus obtaining the expression for the intrinsic error: ; This formula shows that the intrinsic error characterizes the degree of deviation between the subsystem's own model and the real system after eliminating the influence of input bias.

[0050] Based on this, the optimization objective function for parameter identification is reconstructed, transforming the traditional "total error minimization" into "intrinsic error minimization": ; in, These are the weighting coefficients for each subsystem.

[0051] Through the above objective function, each subsystem is only responsible for its own modeling error during the optimization process, without repeatedly penalizing the deviation caused by the propagation of errors from the previous stage, thereby effectively improving the accuracy of the parameter update direction and the physical consistency of the identification results.

[0052] Furthermore, to enhance the model's ability to characterize the overall dynamic behavior of the system, a structural consistency constraint term is introduced into the objective function. This constraint measures the degree of consistency in the dynamic relationships between adjacent subsystems during the cascading process, and is defined as: ; Where the summation index i represents the discrete time step, and Let and represent the output of the simulation model and the state variables of the actual system at time i, respectively. and Then, corresponding to the state variables of the previous moment, this formula forces the dynamic response structure of the model to be consistent with the real system by calculating the sum of the squares of the relative rates of change of the simulation and the actual deviations between adjacent moments.

[0053] This reflects the degree of matching between the system's gain characteristics or response trends, which helps to avoid structural distortion problems caused by relying solely on point value errors.

[0054] Finally, the comprehensive optimization objective function is constructed as follows: ; in, This is a weighting factor.

[0055] In the actual optimization process, combined with the aforementioned real-simulation fusion mechanism, the inputs of each subsystem exhibit different reliability characteristics at different iteration stages. As the proportion of signals within the model gradually increases in cascade propagation, the error propagation effect gradually intensifies. However, this method achieves effective control over the source of error by explicitly separating propagation error from intrinsic error, thereby fully utilizing structural coupling information while avoiding the adverse effects of error cascade amplification.

[0056] In summary, the proposed error attribution decoupling method starts from the error generation mechanism and reconstructs the optimization objective function, transforming the parameter identification process of multi-subsystem from "error superposition-driven" to "error attribution-driven", thereby achieving stable collaborative optimization in structurally coupled systems.

[0057] Because the constructed multi-cascade system inherently possesses significant nonlinear and strongly coupled physical characteristics, and the objective function exhibits a complex nonconvex structure under the influence of the error attribution mechanism, traditional gradient-based information optimization methods are difficult to apply directly. Therefore, this paper employs the particle swarm optimization (PSO) method to perform a global search of parameters, and introduces an improved strategy oriented towards structural coupling characteristics, enabling the optimization process to effectively utilize the error attribution information within the system.

[0058] Let the particle position be represented as a complete parameter vector: ; In the (t)th iteration, the velocity and position updates of the (i)th particle still follow the standard PSO form: ; ; Based on this, a structure-aware improvement mechanism based on error attribution is introduced.

[0059] First, based on the intrinsic error, calculate the error contribution of each subsystem: ; Based on this, normalized weights are constructed: ; This weight reflects the importance of the current identification error of each subsystem. During the parameter update process, the particle velocity is weighted piecewise according to the subsystem parameter blocks: ; in, This indicates a segmented scaling operation based on parameter blocks. This mechanism enables subsystems with larger errors to have stronger search capabilities in the parameter space, thereby achieving adaptive resource allocation.

[0060] Furthermore, to suppress the interference of error propagation on parameter updates, a propagation error suppression factor is introduced. For the (i)th subsystem, it is defined as follows: ; And correct the corresponding parameter update amount: ; When the error of a certain subsystem mainly originates from upstream propagation (i.e., When the value is relatively large, The error will decrease, thus suppressing unnecessary adjustments to the subsystem parameters; conversely, when the error mainly originates from its own modeling bias, its update magnitude will be increased. Through this mechanism, "only the parts that need optimization" are optimized, improving the stability and physical rationality of parameter identification.

[0061] For fitness evaluation, the traditional static input-output mapping is no longer used; instead, calculations are performed based on the structurally coupled signal propagation mechanism described in Section 5.3. In each PSO iteration, for each particle, only one complete cascaded signal propagation process is executed (i.e., one round of signal transmission is completed along the system structure), and the objective function value defined in Section 5.4 is calculated accordingly. ; It is important to emphasize that this propagation process does not involve internal convergence iterations, but rather proceeds synchronously with the PSO iterations. This "single-round propagation—global update" approach ensures that the parameter search process remains dynamically aligned with the system structure, thereby avoiding the increased computational complexity associated with traditional two-level iterations.

[0062] Furthermore, an iterative process-based fusion scheduling mechanism is introduced. Let the fusion coefficient of the t-th iteration be... It changes monotonically with the number of iterations. In the early stages of optimization, The system response is more dependent on real measurement data, thus ensuring that each subsystem possesses reasonable local dynamic characteristics; as iteration progresses, As the number of steps decreases, the system response relies more on signals propagated within the model, gradually strengthening the overall structural consistency during the optimization process. This mechanism allows the algorithm to naturally transition from "segmented calibration" to "overall collaborative optimization."

[0063] In summary, this embodiment, within the standard particle swarm optimization framework, achieves a deep integration of the optimization algorithm with the structural characteristics of multi-cascaded systems by introducing an error contribution weighting mechanism, a propagation error suppression factor, and a fitness evaluation method driven by structural coupling. This significantly improves the accuracy and stability of parameter identification while ensuring computational efficiency.

[0064] This embodiment also includes a multi-subsystem parameter identification system for hydropower units with fusion drive and cascade error attribution, comprising a processor and a computer-readable storage medium interconnected with each other. The computer-readable storage medium stores a computer program, which is executed by the processor to implement the steps of the above-described multi-subsystem parameter identification method for hydropower units with fusion drive and cascade error attribution.

[0065] This embodiment also includes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method for identifying parameters of a multi-subsystem hydropower unit based on fusion drive and cascade error attribution.

[0066] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-readable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process. Figure 1 One or more processes and / or boxes Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0067] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for identifying parameters of multiple subsystems in a hydropower unit that integrates drive and cascade error attribution, characterized in that, include: A simulation model is constructed comprising four subsystems: a local control unit (LCU), a speed governor, an electro-hydraulic servo system, and a hydroelectric power generation unit. These four subsystems are cascaded and connected in a closed loop: the LCU subsystem receives power commands and feedback signals, and outputs PWM pulse control signals to the speed governor subsystem; the speed governor subsystem integrates the PWM pulse control signals and outputs guide vane opening commands to the electro-hydraulic servo system subsystem; the electro-hydraulic servo system subsystem drives the guide vanes and outputs the actual opening signal to the hydroelectric power generation unit subsystem; the hydroelectric power generation unit subsystem outputs the unit's active power and feeds it back to the LCU subsystem, forming a closed loop; the LCU subsystem outputs PWM pulse control signals based on power deviation, denoted as the structure function. The speed controller subsystem uses an integral element to describe its dynamic characteristics, denoted as the structure function. The electro-hydraulic servo system subsystem uses a first-order inertial element to describe hydraulic hysteresis, denoted as the structure function. The hydropower generation unit subsystem employs a nonlinear turbine model and a high-order transfer function to simulate the elastic water hammer effect, denoted as the structure function. ; The structural definition based on the simulation model includes the unidentified parameter vectors of the physical parameters of each subsystem. The simulation model can output analog signals of each subsystem. Set a dynamic fusion coefficient that decays with the number of iterations. , where k is the iteration number. In the first iteration, the actual measured data set of the hydropower unit under the preset operating conditions is obtained as the input of the simulation model. In subsequent iterations, the simulated signals generated by each subsystem of the simulation model in the previous iteration are weighted and fused with the key signals of each subsystem obtained from the field measurement to generate the fused input signal of each subsystem in the current iteration. In each iteration, the generated input signal is input to the simulation model to perform forward signal propagation, and the total error of each subsystem output is calculated. The total error is decomposed into propagation error caused by upstream input deviation and intrinsic error caused only by current subsystem parameter deviation. The upstream error interference is removed to obtain the intrinsic error that only reflects its own parameter mismatch. The objective function is constructed with minimizing the intrinsic error as the core, and an improved particle swarm optimization algorithm is used to iteratively update the parameter vector to be identified. This includes: calculating the magnitude of the intrinsic error of each subsystem, and constructing a normalized adaptive weight based on the magnitude to scale the particle velocity piecewise, so that the subsystem with large error has a stronger search capability; calculating the ratio of the intrinsic error to the propagation error of each subsystem to construct a suppression factor, and multiplying the suppression factor by the parameter update amount to suppress the parameter update of the subsystem when the ratio approaches 0. When the convergence condition is met, the current parameter vector to be identified is output as the final parameter identification result.

2. The method for identifying multi-subsystem parameters of a hydropower unit based on fusion drive and cascade error attribution as described in claim 1, characterized in that, The parameter vector to be identified includes: The local control unit (LCU) subsystem includes the pulse period T and minimum pulse width. Maximum pulse width and proportionality coefficient The pulse width Adaptive adjustment based on power deviation: In the formula, Due to power difference, The sign function is defined as follows: it takes the value 1 when the independent variable is greater than 0, and takes the value -1 when the independent variable is less than or equal to 0. The speed controller subsystem includes an opening given integral time constant. ; The electro-hydraulic servo system subsystem includes the relay response time constant Ty; The hydropower generation unit subsystem includes the water flow inertial time constant. and the time constant of the phase of water hammer waves .

3. The method for identifying multi-subsystem parameters of a hydropower unit based on fusion drive and cascade error attribution as described in claim 2, characterized in that, The structure function , , , The specific content is as follows: Including the control signal u(t) output by the LCU: ； Where M is the PWM output amplitude, ΔP is the difference between the actual unit active power and the target value, T is the pulse period, and k is the pulse number. For symbolic functions, their calculation rules are as follows: ； The dynamic characteristics are described using the transfer function of the integral element: ； Hydraulic hysteresis is described using the transfer function of a first-order inertial element: ； The elastic water hammer effect is simulated using a nonlinear turbine model and a higher-order transfer function. The nonlinear turbine model is as follows: ； ； Where M is the turbine torque, Q is the flow rate, α is the guide vane opening, n is the rotational speed, and H is the head; To address the elastic water hammer effect in the water diversion system, a higher-order transfer function is used to simulate the propagation and reflection characteristics of pressure waves in the pipeline. Its expression is: 。 4. The method for identifying multi-subsystem parameters of a hydropower unit based on fusion drive and cascade error attribution as described in claim 3, characterized in that, The dynamic fusion coefficient The calculation formula is: ； Where k is the number of iterations. λ is the preset initial fusion coefficient, and λ is the preset attenuation coefficient.

5. The method for identifying multi-subsystem parameters of a hydropower unit based on fusion drive and cascade error attribution as described in claim 4, characterized in that, The formula for calculating the fused input signal is: ； in For any subsystem in the simulation model, the fused input signal is... These are the key signals obtained from on-site measurements of this subsystem. This refers to the analog signal obtained by the subsystem in the previous iteration. If this is the first iteration, then .

6. The method for identifying multi-subsystem parameters of a hydropower unit based on fusion drive and cascade error attribution as described in claim 5, characterized in that, The propagation error is calculated as follows: ； Where i represents the i-th subsystem in the corresponding simulation model. This is the structure function corresponding to the subsystem. This is the vector of parameters to be identified for this subsystem. and These represent the key signals obtained from on-site measurements of the previous subsystem and the simulated signals obtained in the previous iteration, respectively.

7. The method for identifying multi-subsystem parameters of a hydropower unit based on fusion drive and cascade error attribution as described in claim 6, characterized in that, The objective function is constructed as follows: ； Where J is the intrinsic error Minimize the objective function: ； These are the weighting coefficients for each subsystem; For structural consistency constraints: Where the summation index i represents the discrete time step, and Let and represent the output of the simulation model and the state variables of the actual system at time i, respectively. and The formula, which corresponds to the state variable at the previous moment, calculates the sum of squares of the relative rates of change of the simulation and the actual deviations at adjacent moments, thus forcing the dynamic response structure of the model to remain consistent with the actual system.

8. The method for identifying multi-subsystem parameters of a hydropower unit based on fusion drive and cascade error attribution as described in claim 1, characterized in that, The improved particle swarm optimization algorithm is used to iteratively update the parameter vector to be identified. In each particle swarm iteration, only one complete cascaded signal forward propagation process is executed, and the objective function value is calculated synchronously without internal nested iterations.

9. A multi-subsystem parameter identification system for hydropower units that integrates drive and cascade error attribution, characterized in that, The device includes an interconnected processor and a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program that is executed by the processor to implement the steps of the hydropower unit multi-subsystem parameter identification method according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for identifying parameters of a multi-subsystem hydropower unit based on fusion drive and cascade error attribution as described in any one of claims 1 to 8.

Images