Generator excitation regulation and control optimization system based on model prediction

Abstract

The invention provides a generator excitation regulation and control optimization system based on model prediction, and relates to the technical field of data processing, and the system comprises a data acquisition module which is used for collecting the operation parameters of a generator; the state estimation module is used for calculating a state variable of the generator excitation system and constructing a state space model of the generator excitation system; the model prediction control module is used for calculating excitation current prediction values at multiple moments in the future by adopting a rolling horizon optimization method, and performing dynamic correction based on a prediction error correction mechanism; the optimization calculation module is used for constructing a target optimization function which comprises a generator voltage stability target item and an excitation current adjustment smoothness target item, and calculating an optimal excitation current adjustment amount by adopting a constraint optimization method; the excitation adjusting module is used for controlling a generator excitation system so as to adjust excitation current; according to the invention, the autonomy and accuracy of the generator excitation regulation and control optimization system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a generator excitation control optimization system based on model prediction. Background Art

[0002] In the prior art, the excitation control of the generator is usually adjusted by a method based on traditional PID (proportional-integral-differential) control or fuzzy control. Specifically, the PID control method measures the deviation between the generator terminal voltage and the set value and calculates the corresponding adjustment amount to adjust the excitation current to maintain the stability of the system. The fuzzy control method is based on expert experience and rules, and adjusts the excitation current through a fuzzy reasoning system to adapt to different operating conditions. These methods can improve the dynamic response capability of the generator and improve the power quality of the power grid to a certain extent. In addition, in recent years, some studies have introduced neural networks or genetic algorithms to optimize the excitation control parameters to improve the adaptability and regulation accuracy of the system. However, these methods still have certain limitations and are difficult to meet the optimization needs in complex power grid environments.

[0003] In practical applications, existing excitation control methods may have problems with dynamic response lag when there is a grid failure or severe load fluctuations. For example, in the scenario of large-scale new energy grid connection, due to the uncertainty of the output of renewable energy sources such as wind power and photovoltaics, the voltage fluctuation of the grid is large. Traditional excitation control methods are difficult to adapt to these changes in real time, resulting in a lag in the voltage regulation of the generator. In the process of connecting a hydropower station to the main grid, if the water speed changes suddenly in a short period of time, the traditional PID control may not be able to adjust quickly due to fixed parameters, resulting in overshoot or undervoltage of the generator voltage, thereby affecting the stability of the entire grid. In addition, fuzzy control based on expert experience may have unreasonable rule settings when dealing with complex working conditions, which reduces the accuracy of excitation regulation and further increases the uncertainty of grid operation. Summary of the invention

[0004] The purpose of the present invention is to provide a generator excitation control optimization system based on model prediction, aiming to solve the problems mentioned in the background technology.

[0005] In order to solve the above technical problems, the technical solution of the present invention is as follows:

[0006] A generator excitation control optimization system based on model prediction, the system comprising:

[0007] The data acquisition module is used to collect the operating parameters of the generator, including the generator terminal voltage, stator current, rotor current, speed and grid frequency, and filter the operating parameters to obtain pre-processed data;

[0008] The state estimation module is used to calculate the state variables of the generator excitation system according to the preprocessed data, including voltage deviation, excitation current change rate, flux change rate and speed deviation, and to construct the state space model of the generator excitation system according to the state variables;

[0009] The model predictive control module is used to calculate the predicted values ​​of the excitation current at multiple moments in the future according to the state space model by using the rolling time domain optimization method to obtain the excitation current prediction sequence, and dynamically correct the excitation current prediction sequence based on the prediction error correction mechanism to obtain the excitation current prediction correction sequence;

[0010] The optimization calculation module is used to construct the target optimization function according to the excitation current prediction correction sequence, including the generator voltage stability target item and the excitation current regulation smoothness target item, and calculate the optimal excitation current adjustment amount by using the constrained optimization method to obtain the optimal excitation control signal;

[0011] The excitation regulation module is used to control the generator excitation system according to the optimal excitation control signal to adjust the excitation current.

[0012] Preferably, the state estimation module comprises:

[0013] A data correction unit, used to suppress noise on preprocessed data based on a dynamic adaptive filtering method, and calculate correction data in combination with historical operation data;

[0014] A state variable calculation unit is used to calculate the state variables of the excitation system according to the correction data, including voltage deviation, excitation current change rate, flux change rate and speed deviation, and to construct an initial estimate of the state space model based on a state variable dynamic fitting method;

[0015] The error compensation unit is used to calculate the state prediction error according to the initial estimated value of the state space model, and dynamically adjust the state variables using an incremental correction strategy to obtain a corrected state space model.

[0016] Preferably, the model predictive control module comprises:

[0017] A future time domain prediction unit is used to calculate the state variables of the excitation system at multiple future moments according to the modified state space model, and to calculate the predicted values ​​of the excitation current at multiple future moments using a sliding window strategy to obtain an excitation current prediction sequence;

[0018] The prediction error correction unit is used to calculate the excitation current prediction error term based on the prediction error cumulative trend analysis method, and use a dynamic error correction mechanism to adjust the excitation current prediction sequence to obtain the excitation current prediction correction sequence.

[0019] Preferably, the optimization calculation module includes:

[0020] An objective function construction unit is used to construct an optimization objective function of the excitation system according to the excitation current prediction correction sequence, wherein the optimization objective function includes a generator voltage stability objective item and an excitation current regulation smoothness objective item, and calculates an initial weight value of the optimization objective function based on a weighted cumulative error analysis method;

[0021] The target weight adjustment unit is used to adjust the initial weight value of the optimization target function according to the dynamic load characteristics of the power grid and generate an optimization weight parameter set;

[0022] The constrained optimization calculation unit is used to calculate the optimal excitation current adjustment amount using a constrained optimization solution method according to the optimization objective function and the optimization weight parameter set.

[0023] Preferably, the excitation system state variables, including voltage deviation, excitation current change rate, flux change rate and speed deviation, are calculated according to the correction data, and the initial estimate of the state space model is constructed based on the state variable dynamic fitting method, including:

[0024] According to the correction data, the voltage deviation, the excitation current change rate, the flux change rate and the speed deviation are calculated, and the variable weight adaptive normalization method is used for different state variables to obtain the normalized state variables;

[0025] According to the normalized state variables, the initial state fitting model is constructed by using the nonlinear multi-order curve fitting method;

[0026] According to the initial state fitting model, the time trend term of the state variable is calculated, and the initial estimate of the state space model is constructed in combination with the historical state data.

[0027] Preferably, the state prediction error is calculated according to the initial estimated value of the state space model, and the state variable is dynamically adjusted by using an incremental correction strategy to obtain a corrected state space model, including:

[0028] According to the initial estimate of the state space model, the state prediction error is calculated based on the actual state variables and the predicted state variables at the current moment;

[0029] According to the state prediction error, the state increment correction term is calculated by using the dynamic error compensation method;

[0030] According to the state increment correction term, the initial estimation value of the state space model is dynamically adjusted to obtain a corrected state space model.

[0031] Preferably, the method of calculating the excitation system state variables at multiple future moments according to the modified state space model, and calculating the excitation current prediction values ​​at multiple future moments using a sliding window strategy to obtain the excitation current prediction sequence includes:

[0032] According to the modified state space model and based on the recursive state update method, the state variables of the excitation system at multiple moments in the future are calculated;

[0033] According to the state variables of the excitation system at multiple future moments, a sliding window strategy is adopted to calculate the predicted values ​​of the excitation current at multiple future moments;

[0034] According to the predicted values ​​of the excitation current at multiple future moments, an excitation current prediction sequence is constructed.

[0035] Preferably, the method of calculating the excitation current prediction error term based on the prediction error cumulative trend analysis method and adjusting the excitation current prediction sequence using a dynamic error correction mechanism to obtain the excitation current prediction correction sequence includes:

[0036] According to the excitation current prediction sequence, based on the historical prediction error data, the excitation current prediction error term is calculated;

[0037] According to the excitation current prediction error term, a dynamic error correction mechanism is adopted to calculate the excitation current adjustment correction value;

[0038] According to the excitation current adjustment correction value, the error compensation is performed on the excitation current prediction sequence to obtain the excitation current prediction correction sequence.

[0039] Preferably, the optimization objective function of the excitation system is constructed according to the excitation current prediction correction sequence, the optimization objective function includes the generator voltage stability objective term and the excitation current regulation smoothness objective term, and the initial weight value of the optimization objective function is calculated based on the weighted cumulative error analysis method, including:

[0040] According to the excitation current prediction correction sequence, the generator voltage deviation and the excitation current regulation change rate are calculated, and an optimization objective function is constructed, wherein the optimization objective function includes a generator voltage stability objective item and an excitation current regulation smoothness objective item; ,in, To optimize the objective function, is the prediction time step, and To optimize the weight coefficient of the objective function, For the The voltage deviation at the moment, The set voltage indicates the target voltage that the generator should maintain. For the The actual voltage at the moment, For the The rate of change of the excitation current at time For the The excitation current at time For the The excitation current at the moment;

[0041] According to the historical operation data, the weighted cumulative error analysis method is used to calculate the initial weight value of the optimization objective function;

[0042] According to the initial weight value of the optimization objective function, the optimization objective constraint relationship is established.

[0043] Preferably, the method of calculating the optimal excitation current adjustment amount by using a constrained optimization solution method according to the optimization objective function and the optimization weight parameter set includes:

[0044] According to the optimization objective function and the optimization weight parameter set, an optimal excitation current adjustment calculation model is constructed, wherein the calculation model includes excitation current adjustment constraint conditions;

[0045] According to the dynamic working range of the excitation system, the excitation current change rate threshold and the voltage stability threshold are set, and the calculation model is solved based on the constraint optimization solution method to obtain the optimal excitation current adjustment amount;

[0046] The calculation formula of the optimal excitation current adjustment amount is: ,in, is the optimal excitation current adjustment, is the excitation current adjustment, is an operator, which indicates the variable value that makes the objective function achieve the minimum value;

[0047] Constraints ( ) ,in, and are the minimum and maximum values ​​of the excitation current, respectively. is the upper limit of the excitation current change rate;

[0048] An optimal excitation control signal is generated according to the optimal excitation current adjustment amount.

[0049] The above solution of the present invention includes at least the following beneficial effects:

[0050] This system adopts an excitation control optimization method based on model prediction, and through the collaborative work of multiple functional modules, it significantly improves the accuracy, real-time performance and stability of excitation regulation.

[0051] The system first uses a data acquisition module to collect the operating parameters of the generator in real time, including the generator terminal voltage, stator current, rotor current, speed and grid frequency, and filters the operating parameters. Compared with the traditional control method that relies on a single variable (such as voltage deviation) for control, this system can integrate multiple key variables, improve the perception of the grid state, ensure the accuracy of data input, and thus provide more stable input for subsequent control calculations.

[0052] Based on data acquisition, the state estimation module calculates the state of the generator excitation system and constructs a state space model. Traditional methods usually use fixed control parameters or control rules based on experience, which makes it difficult to accurately describe the dynamic characteristics of the excitation system under different working conditions. This system calculates multiple key variables such as voltage deviation, excitation current change rate, flux change rate and speed deviation, and combines them with the state space model to establish a more accurate description of the system's dynamic characteristics, so that the control strategy can be adjusted at any time to adapt to changes in the power grid and avoid the adjustment lag caused by fixed parameters.

[0053] Based on the state space model, the model predictive control module uses the rolling time domain optimization method to calculate the predicted values ​​of the excitation current at multiple moments in the future and generate an excitation current prediction sequence. Unlike the existing technology that relies on historical data for control, this system can predict the excitation demand at future moments and make adjustments in advance, avoiding the problem of excitation regulation lag caused by sudden changes in the power grid state. In addition, this system also adopts a prediction error correction mechanism to dynamically adjust the excitation current prediction value, reduce the cumulative impact of the prediction error, and make the excitation regulation more accurate.

[0054] In terms of optimization calculation, the optimization calculation module constructs the optimization objective function, which includes the generator voltage stability objective term and the excitation current regulation smoothness objective term. Compared with the traditional PID control that only focuses on voltage stability, this system considers the smoothness of the excitation current change during the optimization process to avoid large fluctuations in the excitation current, thereby reducing the impact on the stability of the power grid. At the same time, this system adopts a constrained optimization method to ensure that the optimal excitation current adjustment is within a reasonable range, avoid the excitation current exceeding the equipment carrying capacity or the problem of regulation overshoot, and improve the robustness of the control strategy.

[0055] Finally, the excitation regulation module controls the generator excitation system according to the optimal excitation control signal, so that the excitation current can quickly respond to grid changes and ensure voltage stability. When wind power, photovoltaic and other new energy sources are connected to the grid, the traditional excitation control method may cause large voltage fluctuations due to the uncertainty of output, and the regulation system is difficult to adapt in real time. This system can achieve more accurate and faster excitation regulation through model prediction and optimization calculation, reduce the impact of new energy grid connection on the grid, and improve the overall stability of the system.

[0056] In summary, this system overcomes the shortcomings of existing excitation control methods through the coordinated work of data acquisition, state estimation, predictive control, optimization calculation and excitation regulation, and significantly improves the accuracy, real-time and stability of excitation regulation. It is particularly suitable for applications with large load fluctuations, frequent grid connection of new energy and complex power grid environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is an architecture diagram of a generator excitation control optimization system based on model prediction provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0058] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0059] like Figure 1 As shown, an embodiment of the present invention proposes a generator excitation control optimization system based on model prediction, and the system includes:

[0060] The data acquisition module is used to collect the operating parameters of the generator, including the generator terminal voltage, stator current, rotor current, speed and grid frequency, and filter the operating parameters to obtain pre-processed data;

[0061] The state estimation module is used to calculate the state variables of the generator excitation system according to the preprocessed data, including voltage deviation, excitation current change rate, flux change rate and speed deviation, and to construct the state space model of the generator excitation system according to the state variables;

[0062] The model predictive control module is used to calculate the predicted values ​​of the excitation current at multiple moments in the future according to the state space model by using the rolling time domain optimization method to obtain the excitation current prediction sequence, and dynamically correct the excitation current prediction sequence based on the prediction error correction mechanism to obtain the excitation current prediction correction sequence;

[0063] The optimization calculation module is used to construct the target optimization function according to the excitation current prediction correction sequence, including the generator voltage stability target item and the excitation current regulation smoothness target item, and calculate the optimal excitation current adjustment amount by using the constrained optimization method to obtain the optimal excitation control signal;

[0064] The excitation regulation module is used to control the generator excitation system according to the optimal excitation control signal to adjust the excitation current.

[0065] In the embodiment of the present invention, during the excitation control process of the generator, the traditional excitation control method is difficult to meet the dynamic adjustment requirements of the complex power grid environment, especially when the load changes drastically or the power grid fails, the conventional fixed parameter control strategy may cause the voltage fluctuation to intensify and affect the stability of the power grid. The generator excitation control optimization system based on model prediction realizes the precise optimization and control of the excitation current through the coordinated work of multiple modules such as data acquisition, state estimation, model prediction, optimization calculation and excitation regulation, thereby improving the dynamic response capability and operation stability of the power system.

[0066] During the operation of the excitation system, the data acquisition module collects the operating parameters of the generator in real time, including the generator terminal voltage, stator current, rotor current, speed, and grid frequency. These data may contain measurement errors and noise in the original acquisition state, so they need to be filtered to obtain pre-processed data to ensure the accuracy of subsequent calculations. Through effective data processing, the system's ability to perceive the actual grid state can be improved, providing reliable input for subsequent state estimation.

[0067] The state estimation module calculates the key state variables of the generator excitation system based on the preprocessed data, including voltage deviation, excitation current change rate, flux change rate and speed deviation. These variables can accurately characterize the operating state of the excitation system, and build a state space model of the system through the state space modeling method. This model is used to describe the dynamic behavior of the excitation system under different control inputs, providing a mathematical basis for subsequent prediction and optimization calculations.

[0068] Based on the state space model, the model predictive control module uses the rolling time domain optimization method to calculate the predicted values ​​of the excitation current at multiple moments in the future to form an excitation current prediction sequence. Traditional excitation control methods are difficult to adapt to the rapid changes in the operating status of the power grid, while this module can predict future excitation regulation requirements based on the real-time status of the system and improve adaptability to power grid disturbances. In addition, the module also integrates a prediction error correction mechanism that can dynamically adjust the excitation current prediction sequence and reduce the impact of prediction error accumulation on system control accuracy.

[0069] The optimization calculation module constructs an optimization objective function based on the excitation current prediction correction sequence, which includes the generator voltage stability objective term and the excitation current regulation smoothness objective term. By using the constrained optimization method to calculate the optimal excitation current adjustment, it ensures that the excitation system can maintain the stability of the generator terminal voltage and avoid drastic fluctuations in the excitation current, thereby improving the accuracy and robustness of the regulation.

[0070] The excitation regulation module finally performs real-time adjustment of the excitation current according to the optimal excitation control signal, so that the system can operate stably under different load conditions and grid conditions. Compared with traditional methods, this system can respond to grid disturbances faster, reduce voltage fluctuations, improve power quality, and reduce the control error of the excitation system, making the generator operation more efficient and safer.

[0071] In a preferred embodiment of the present invention, the state estimation module includes:

[0072] A data correction unit, used to suppress noise on preprocessed data based on a dynamic adaptive filtering method, and calculate correction data in combination with historical operation data;

[0073] A state variable calculation unit is used to calculate the state variables of the excitation system according to the correction data, including voltage deviation, excitation current change rate, flux change rate and speed deviation, and to construct an initial estimate of the state space model based on a state variable dynamic fitting method;

[0074] The error compensation unit is used to calculate the state prediction error according to the initial estimated value of the state space model, and dynamically adjust the state variables using an incremental correction strategy to obtain a corrected state space model.

[0075] In the embodiment of the present invention, during the excitation control process, due to the influence of measurement errors, data noise and environmental changes, directly using the collected operating parameters may lead to a decrease in control accuracy and affect the regulation effect of the excitation system. In order to solve this problem, the state estimation module improves the accuracy of state estimation through the coordinated work of multiple units such as data correction, state variable calculation and error compensation, thereby optimizing the subsequent excitation control calculation.

[0076] The data correction unit uses a dynamic adaptive filtering method to suppress noise in preprocessed data and performs correction in combination with historical operating data. The power grid operating environment is complex, and traditional static filtering methods may cause signal distortion when dealing with sudden load changes or short-term power grid disturbances. The dynamic adaptive filtering method can adaptively adjust the filtering parameters according to the data characteristics to ensure that the extracted data is more representative. Correction in combination with historical operating data can eliminate long-term drift errors and make state estimation more reliable.

[0077] The state variable calculation unit calculates the key state variables of the excitation system, including voltage deviation, excitation current change rate, flux change rate and speed deviation, based on data correction. The initial estimate of the state space model is constructed through the dynamic fitting method. This estimate can provide a more accurate initial state and provide more precise input for subsequent model predictive control.

[0078] The error compensation unit further optimizes the state estimation results. Since there may be errors in the calculation of the state space model, this unit uses an incremental correction strategy to dynamically adjust the state variables, so that the state estimation results can be continuously optimized during actual operation, ensuring that the control calculation of the excitation system is more accurate.

[0079] The implementation of the data correction unit mainly includes three stages: noise suppression, outlier detection and data compensation.

[0080] 1. Noise suppression:

[0081] The operating data of the excitation system mainly include the generator terminal voltage, stator current, rotor current, speed and grid frequency. These data may be subject to various interferences, such as:

[0082] Electromagnetic noise: interference from electrical equipment or high-frequency signals around the generator.

[0083] Measurement error: Insufficient sensor accuracy or aging leads to data deviation.

[0084] Signal jitter: Due to sampling frequency limitations, the data may have high-frequency fluctuations.

[0085] To address the noise problem, the data correction unit uses a dynamic adaptive filtering method, which automatically adjusts the filtering parameters according to the fluctuation of the data, so that the system can effectively filter out high-frequency noise without weakening the changing characteristics of the actual signal. For example:

[0086] When the power grid is stable, the filtering parameters are small, maintaining the original characteristics of the data to facilitate the detection of small changes.

[0087] When the power grid fluctuates greatly, the filtering parameters are automatically increased to reduce noise interference and improve signal quality.

[0088] 2. Outlier Detection:

[0089] Outliers are usually caused by sensor failures, communication errors, or sudden changes in the environment, such as:

[0090] Sudden voltage or current jumps far beyond the normal physical range.

[0091] The data fluctuates dramatically in a short period of time, but there is no corresponding adjustment of the load or excitation system.

[0092] The data correction unit performs real-time detection through a sliding window method, that is, analyzing historical data over a period of time and calculating the deviation between the data point and the historical data. If the deviation exceeds the set threshold, the data point is considered to be an outlier and marked for processing.

[0093] 3. Data compensation:

[0094] When abnormal data is detected, compensation is required to ensure the continuity and accuracy of the data. The compensation method usually adopts:

[0095] Interpolation method: If data is missing or there are few outliers, use the previous and next data to interpolate and supplement reasonable values.

[0096] Historical data fitting: If the anomaly lasts for a long time, a trend model is established using historical normal data, and reasonable values ​​are predicted based on the trend model for replacement.

[0097] Multi-sensor fusion: If the system has redundant sensors, such as multiple measurement points, abnormal data can be corrected by taking a weighted average of the different sensors.

[0098] Through the above correction steps, the data correction unit can effectively reduce the measurement error and noise impact, improve data quality, and make subsequent state estimation and control calculation more accurate.

[0099] In a preferred embodiment of the present invention, the model predictive control module includes:

[0100] A future time domain prediction unit is used to calculate the state variables of the excitation system at multiple future moments according to the modified state space model, and to calculate the predicted values ​​of the excitation current at multiple future moments using a sliding window strategy to obtain an excitation current prediction sequence;

[0101] The prediction error correction unit is used to calculate the excitation current prediction error term based on the prediction error cumulative trend analysis method, and use a dynamic error correction mechanism to adjust the excitation current prediction sequence to obtain the excitation current prediction correction sequence.

[0102] In the embodiment of the present invention, in the model predictive control, accurately predicting the state of the excitation system at a future time is crucial to optimizing the control strategy. The system improves the accuracy of the excitation current prediction through future time domain prediction and error correction mechanism, so that the excitation regulation can more accurately adapt to the changes in the power grid state.

[0103] The future time domain prediction unit calculates the state variables of the excitation system at multiple future moments based on the modified state space model, and uses a sliding window strategy to calculate the predicted values ​​of the excitation current at multiple future moments. The sliding window strategy can update the prediction data at each time step, avoiding the lag effect caused by the fixed window length, making the prediction results more accurate. Compared with the traditional fixed parameter prediction method, this method is more flexible in dealing with grid load changes, can adjust the prediction range in time, and improve the accuracy of excitation current calculation.

[0104] The prediction error correction unit further optimizes the excitation current prediction value based on the future time domain prediction. The prediction error cumulative trend analysis method is used to calculate the excitation current prediction error term, and the excitation current prediction sequence is adjusted through the dynamic error correction mechanism. Traditional methods often rely on fixed error correction parameters and are difficult to adapt to dynamically changing systems. This method can adaptively adjust the correction parameters to gradually converge the prediction error and improve the accuracy of excitation control.

[0105] In a preferred embodiment of the present invention, the optimization calculation module includes:

[0106] An objective function construction unit is used to construct an optimization objective function of the excitation system according to the excitation current prediction correction sequence, wherein the optimization objective function includes a generator voltage stability objective item and an excitation current regulation smoothness objective item, and calculates an initial weight value of the optimization objective function based on a weighted cumulative error analysis method;

[0107] The target weight adjustment unit is used to adjust the initial weight value of the optimization target function according to the dynamic load characteristics of the power grid and generate an optimization weight parameter set;

[0108] The constrained optimization calculation unit is used to calculate the optimal excitation current adjustment amount using a constrained optimization solution method according to the optimization objective function and the optimization weight parameter set.

[0109] In the embodiment of the present invention, during the excitation system optimization calculation process, the construction of the optimization objective function and the dynamic adjustment of the optimization parameters play a key role in the stable operation of the system. The optimization calculation module uses units such as objective function construction, target weight adjustment, and constraint optimization calculation to ensure that the excitation current adjustment can not only ensure the stability of the generator voltage, but also reduce unnecessary excitation current fluctuations, thereby improving the overall performance of the system.

[0110] The objective function construction unit establishes the optimization objective function based on the excitation current prediction correction sequence, including the generator voltage stability objective item and the excitation current regulation smoothness objective item. The voltage stability objective item ensures the minimum voltage fluctuation at the generator terminal, while the excitation current regulation smoothness objective item constrains the variation range of the excitation current to prevent system oscillation caused by frequent adjustments.

[0111] The target weight adjustment unit adjusts the initial weight value of the optimization objective function according to the dynamic load characteristics of the power grid and generates a set of optimization weight parameters. The load characteristics of the power grid are time-varying, and the requirements for excitation regulation are different under different load levels. By dynamically adjusting the weight of the objective function, the optimization calculation can be made more in line with actual operation requirements.

[0112] The constrained optimization calculation unit uses the constrained optimization solution method to calculate the optimal excitation current adjustment amount based on the target weight adjustment, ensuring that the optimization result makes the excitation current change smooth and reasonable while meeting the stability of the power grid, thereby improving the overall control effect.

[0113] In a preferred embodiment of the present invention, the excitation system state variables, including voltage deviation, excitation current change rate, flux change rate and speed deviation, are calculated based on the correction data, and the initial estimate of the state space model is constructed based on the state variable dynamic fitting method, including:

[0114] According to the correction data, the voltage deviation, the excitation current change rate, the flux change rate and the speed deviation are calculated, and the variable weight adaptive normalization method is used for different state variables to obtain the normalized state variables;

[0115] According to the normalized state variables, the initial state fitting model is constructed by using the nonlinear multi-order curve fitting method;

[0116] According to the initial state fitting model, the time trend term of the state variable is calculated, and the initial estimate of the state space model is constructed in combination with the historical state data.

[0117] In the embodiment of the present invention, during the generator excitation control process, the accuracy of state estimation has a decisive influence on the control effect of the entire system. Traditional state estimation methods have poor adaptability to system parameters and are easily affected by external interference factors, resulting in large errors in calculation results. This embodiment makes the state estimation results more accurate through steps such as data correction, state variable calculation and model construction, providing high-quality input for subsequent excitation current prediction and optimization.

[0118] During the data correction stage, the operating parameters obtained by the system are often affected by noise interference or measurement errors. If they are directly used for calculation, the system adjustment accuracy may be reduced. Therefore, a dynamic adaptive filtering method is used to suppress noise for parameters of different time scales, and error compensation is performed in combination with historical data. This method can effectively distinguish the real signal and noise components in the data and improve the credibility of the correction data.

[0119] In the process of state variable calculation, the voltage deviation, excitation current change rate, flux change rate and speed deviation are calculated based on the corrected data. These variables can accurately characterize the operating state of the generator excitation system. In the calculation process, the variable weight adaptive normalization method is used to ensure the consistency of the numerical scales between different variables, thereby avoiding the problem of weight imbalance caused by different numerical scales in subsequent calculations. This method can automatically adjust the influence weight of the variables under different operating conditions, making the state variable calculation more stable and reliable.

[0120] In the model construction stage, a nonlinear multi-order curve fitting method is used to establish an initial state fitting model based on the historical data of state variables. Traditional linear fitting methods are often difficult to accurately describe the dynamic changes of state variables when dealing with complex power grid operating environments, while multi-order curve fitting can better capture the nonlinear characteristics of the system and improve the accuracy of the state space model. Finally, based on the initial state fitting model, the time change trend of the state variables is calculated in combination with the historical state data, and the initial estimate of the state space model is constructed, making the state estimation result more accurate and providing reliable input for subsequent model predictive control.

[0121] The method of constructing an initial state fitting model based on the normalized state variables and using a nonlinear multi-order curve fitting method specifically includes:

[0122] 1. Fitting method selection:

[0123] The nonlinear multi-order curve fitting method usually adopts the following steps:

[0124] Data acquisition: Collect state variable data over a period of time, such as voltage deviation, excitation current change rate, flux change rate and speed deviation.

[0125] Model selection: Select the appropriate curve type according to the changing trend of the data, for example:

[0126] Quadratic curve fitting: Suitable for slowly changing state variables, such as excitation current in steady-state operation.

[0127] Cubic or higher order curve fitting: suitable for fast changing state variables, such as voltage deviation when load changes suddenly.

[0128] Calculate fitting parameters: Use historical data to calculate the best fitting curve with the minimum error.

[0129] Data verification: Compare the fitted model with the actual data to ensure that the fitting accuracy meets the requirements.

[0130] 2. Application of fitting results:

[0131] The fitted state variable model can be used to:

[0132] Short-term prediction: determine the changing trend of state variables in the future in advance, so that the control system can be adjusted more accurately.

[0133] Anomaly detection: If the actual data deviates significantly from the fitted model, it may indicate that there is an anomaly in the system, such as sensor failure or abnormal fluctuations in the power grid.

[0134] Optimization calculation input: serves as the basic data for subsequent optimization calculations to improve the accuracy of optimization calculations.

[0135] The fitting model according to the initial state, calculating the time variation trend item of the state variable, and constructing the initial estimation value of the state space model in combination with the historical state data specifically includes:

[0136] 1. Calculation of time-varying trend of state variables:

[0137] The time variation trend of the state variable reflects the rate of change of the system at different times, that is:

[0138] Changing trend of voltage deviation: reflects the response speed of generator voltage to excitation current adjustment.

[0139] Trend of excitation current change rate: used to evaluate the dynamic adjustment capability of the excitation system.

[0140] Trend of flux change rate: reflects the rate of change of the generator rotor magnetic field.

[0141] Methods for calculating trends include:

[0142] Sliding window analysis: Analyze the rate of change of state variables within a certain time window to avoid the impact of short-term mutations.

[0143] Trend fitting: Through curve fitting methods such as multi-order curve fitting, the growth or decay trend of state variables is analyzed, so that the system can predict state changes in advance.

[0144] Weighted cumulative analysis: Combines historical data and gives higher weight to recent changes to improve the timeliness of trend calculations.

[0145] 2. Construction of initial estimates of the state space model:

[0146] After obtaining the time-varying trend of the state variables, it is necessary to combine the historical state data to construct the initial estimate of the state space model. This process includes:

[0147] Extracting initial state based on historical data: By analyzing historical operating data, the steady-state parameters of the system are extracted as the reference point of the state space model.

[0148] Use trend information to adjust initial estimates: If trend analysis indicates that the state variables are changing rapidly, the initial estimates are adjusted to accommodate the current system state.

[0149] Dynamic update: As the system runs, the initial estimate of the state space model needs to be continuously updated to reflect the latest state changes.

[0150] Through the above method, the system can construct a more accurate state space model, making subsequent control calculations and optimization adjustments more accurate, thereby improving the efficiency and stability of excitation regulation.

[0151] In a preferred embodiment of the present invention, the state prediction error is calculated according to the initial estimation value of the state space model, and the state variable is dynamically adjusted by using an incremental correction strategy to obtain a corrected state space model, including:

[0152] According to the initial estimate of the state space model, the state prediction error is calculated based on the actual state variables and the predicted state variables at the current moment;

[0153] According to the state prediction error, the state increment correction term is calculated by using the dynamic error compensation method;

[0154] According to the state increment correction term, the initial estimation value of the state space model is dynamically adjusted to obtain a corrected state space model.

[0155] In the embodiment of the present invention, in the excitation control system, the accuracy of state estimation is crucial to the control effect. However, due to the complexity of the power grid operation state, the traditional fixed parameter model is difficult to adapt to the dynamically changing environment, resulting in deviations in state estimation. This embodiment improves the accuracy of state estimation through state prediction error calculation and dynamic correction strategy, so that the control system can better adapt to power grid changes.

[0156] First, the state prediction error is calculated based on the actual state variables and predicted state variables at the current moment. Due to the nonlinear characteristics of the excitation system, it is difficult to completely eliminate the prediction error even with a high-precision state modeling method. Therefore, the error calculation process not only considers the deviation value at the current moment, but also analyzes the historical error trend to determine whether the error has a cumulative effect or a periodic change trend. This error analysis method can more comprehensively reflect the uncertainty of state estimation, thereby providing a basis for subsequent corrections.

[0157] In the error correction stage, a dynamic error compensation method is used to calculate the state increment correction term according to the error change trend. Compared with the traditional fixed increment correction method, this method can dynamically adjust the correction amplitude under different working conditions, making the correction result more accurate. When the error is small, the correction amount can be small to avoid unnecessary parameter disturbance; when the error is large, the system automatically increases the correction strength to improve the convergence speed of state estimation.

[0158] Finally, based on the state increment correction term, the initial estimate of the state space model is dynamically adjusted to obtain the corrected state space model. This model can reflect the latest operating status of the excitation system in real time, improve the accuracy of state estimation, and provide a reliable basis for subsequent excitation current prediction and optimization calculation. Compared with traditional methods, this technology not only improves the accuracy of state estimation, but also reduces the impact on system stability, allowing the excitation control system to operate more efficiently and stably.

[0159] The step of calculating the state prediction error based on the initial estimated value of the state space model and the actual state variables and the predicted state variables at the current moment specifically includes:

[0160] 1. Data preparation:

[0161] Actual state variables at the current moment: These data come from the real-time operating parameters of the generator, including voltage, excitation current, speed, etc. After being processed by the data correction unit, they can more accurately reflect the current state of the generator.

[0162] Predicted state variables: This is calculated based on the initial estimate of the state space model, that is, assuming that the system develops according to the trend calculated by the model, the state it should theoretically reach.

[0163] 2. Calculate the state prediction error:

[0164] For example, if the state space model predicts that the voltage deviation at a certain moment in the future should gradually decrease, but the actual measured data shows that the voltage deviation increases instead of decreases, it indicates that there is an error in the state prediction and correction is needed.

[0165] In the process of calculating the state prediction error, in addition to the simple comparison between the current state and the predicted value, the following methods can also be combined:

[0166] Time series analysis: Consider the error trend over a period of time to determine whether the error is a short-term fluctuation or a long-term accumulation.

[0167] Error distribution analysis: Determine the distribution characteristics of the error. If the error continues to lean in a certain direction, it may indicate that the model has systematic deviations and needs to be adjusted overall.

[0168] Dynamic weight adjustment: Different weights are assigned to error calculations under different operating conditions. For example, when the load changes suddenly, the short-term error may be large, but the impact on the overall system is small, so a lower weight can be assigned, while in steady-state operation, a small change in the error may affect the control accuracy, so a higher weight should be assigned.

[0169] By calculating the state prediction error in the above manner, the system can more accurately evaluate the accuracy of the model prediction and provide a basis for subsequent error compensation and dynamic correction.

[0170] The state increment correction term is calculated by using a dynamic error compensation method according to the state prediction error, specifically including:

[0171] 1. Error classification:

[0172] Random error: short-term error fluctuations are large, but there is no obvious bias overall, which may be caused by measurement errors or short-term external interference. For this kind of error, the system can use a low-amplitude correction to avoid excessive adjustment affecting system stability.

[0173] Systematic error: The error continues to accumulate in a certain direction, indicating that there is an offset in the state space model, for example, the model parameters do not accurately reflect the actual characteristics of the system. In this case, the system needs to increase the correction efforts to eliminate the impact of the error on the control strategy as soon as possible.

[0174] Sudden error: A short-term error caused by a sudden change in the external environment, such as a sudden change in load, a short-term disturbance in the power grid, etc. This error may be temporary, so it is necessary to determine whether it needs to be corrected immediately or wait for the system to recover on its own.

[0175] 2. Error trend analysis:

[0176] Calculate the rate of change of the error at different time points. If the error continues to increase, it means that the correction effort needs to be increased; if the error is decreasing, it means that the system has a certain degree of self-adjustment ability and the correction effort can be appropriately reduced.

[0177] 3. Calculation of state increment correction item:

[0178] According to the type and trend of the error, the appropriate correction amount is calculated and the state space model is adjusted. For example, when the error is small, the correction term can be set to a smaller increment to avoid unnecessary adjustments in the system; when the error is large and continues to grow, the amplitude of the correction term can be appropriately increased to speed up the error convergence.

[0179] The method of dynamically adjusting the initial estimated value of the state space model according to the state increment correction term to obtain the corrected state space model specifically includes:

[0180] 1. Adjust the weight of state variables:

[0181] Under certain operating conditions, some state variables may have a greater impact on the system than other variables. For example, during the process of renewable energy grid connection, voltage fluctuations may be more obvious, so it is necessary to increase the weight of voltage deviation in the model calculation to ensure that the control strategy pays more attention to voltage stability.

[0182] 2. Adjust model parameters:

[0183] By analyzing historical data, the key parameters of the state space model are updated, such as adjusting the model's time constant, gain coefficient, etc., to make it more consistent with the current system state. For example, in the case of rapid load changes, the sensitivity of the system response can be increased to improve control accuracy.

[0184] 3. Dynamically modify the state transfer relationship:

[0185] Traditional state space models are usually based on a fixed state transfer matrix, but under different grid operating conditions, the dynamic characteristics of the system may change. For example, when operating at low load, the adjustment response of the excitation system is slow, while when operating at high load, the response speed may be faster. Therefore, it is necessary to dynamically adjust the state transfer relationship based on the results of the error analysis so that the model can better predict the future system state.

[0186] 4. Update the state space model in real time:

[0187] During operation, the system continuously monitors error changes and periodically updates the state space model so that it can adapt to changes in the grid environment and improve the overall stability of the control system.

[0188] Through the above method, the system can automatically adjust the model parameters in a complex power grid environment to make the state estimation more accurate, thereby improving the regulation capability of the excitation system, reducing the lag of the excitation current adjustment, improving the stability of voltage regulation, and enhancing the generator's adaptability to load fluctuations.

[0189] In a preferred embodiment of the present invention, the state variables of the excitation system at multiple future moments are calculated according to the modified state space model, and the predicted values ​​of the excitation current at multiple future moments are calculated using a sliding window strategy to obtain the excitation current prediction sequence, including:

[0190] According to the modified state space model and based on the recursive state update method, the state variables of the excitation system at multiple moments in the future are calculated;

[0191] in, , For the The state variables of the excitation system at time , For the The state variable vector of the excitation system at time , is the state transfer matrix, which describes the time evolution relationship of the excitation system state. is the control input matrix, describing the effect of excitation current adjustment on the state variables, For the The excitation current control input at the moment;

[0192] According to the state variables of the excitation system at multiple future moments, a sliding window strategy is adopted to calculate the predicted values ​​of the excitation current at multiple future moments;

[0193] in, , For the The predicted value of the excitation current at time is the output matrix, describing the state variables The influence on the excitation current is is the control input influence matrix, describing the excitation current control input Direct effect on the excitation current;

[0194] According to the predicted values ​​of the excitation current at multiple future moments, an excitation current prediction sequence is constructed.

[0195] In the embodiment of the present invention, in the control process of the excitation system, predicting the system state at the future moment is crucial to optimizing the control scheme. Traditional excitation control methods are often based on rule control with fixed parameters, which is difficult to adapt to dynamic changes in complex power grid environments in real time, resulting in lagging excitation current adjustment or excessive adjustment, thereby affecting the stability of the system. This embodiment uses future time domain prediction and sliding window calculation methods to make the excitation current prediction more accurate and improve the flexibility of excitation control.

[0196] In the future time domain prediction stage, the state variables of the excitation system at multiple moments in the future are calculated based on the revised state space model. Through the recursive state update method, the system can use the current state information to predict the future operation trend of the excitation system. The traditional static prediction method can only perform simple extrapolation based on the current state, which is difficult to adapt to the complex power grid dynamics. The recursive state update method can update the predicted value in each time step, making the prediction result more accurate.

[0197] When calculating the predicted value of the excitation current at multiple moments in the future, a sliding window strategy is used, which can dynamically adjust the predicted data at multiple moments in the future to adapt to different control requirements. Compared with the fixed window prediction method, the sliding window strategy can more flexibly adjust the prediction interval length, so that the system can quickly adjust the prediction range of the excitation current when facing sudden load changes or fault disturbances, thereby improving the reliability of the prediction results.

[0198] Finally, an excitation current prediction sequence is constructed, which provides a basis for subsequent optimization calculations and control signal generation. Compared with traditional methods, this technology can more accurately predict the future trend of excitation current changes, making excitation control more precise, reducing the problem of excessive or insufficient excitation current adjustment caused by prediction errors, and improving the stability and adaptability of the system.

[0199] Among them, the matrix , , , Used to describe the dynamic characteristics of a system, they can be obtained through theoretical modeling or system identification.

[0200] Theoretical modeling is to build a mathematical model based on the physical characteristics of the excitation system to derive the matrix , , , The excitation system is mainly composed of the excitation winding, armature winding, voltage regulator, etc. Its core function is to control the excitation current of the generator to maintain a stable generator terminal voltage.

[0201] The dynamic behavior of the excitation system can usually be mathematically related by the law of electromagnetic induction, Kirchhoff's circuit law, etc. For example:

[0202] The generator terminal voltage is mainly affected by the rotor flux, which is determined by the excitation current.

[0203] The change of excitation current is affected by the inductance, resistance and input excitation voltage of the excitation winding.

[0204] The voltage regulator is responsible for adjusting the excitation current to maintain the generator terminal voltage at the set value.

[0205] The mathematical model of the excitation system can be divided into state variables such as voltage deviation, excitation current change rate, etc., control input such as excitation current adjustment, and system output such as generator terminal voltage.

[0206] In order to convert the excitation system into state space expression, it is necessary to define the state variables and control input U:

[0207] State variables X: generator terminal voltage (indicating dynamic voltage changes), excitation current (indicating flux changes), speed deviation (affecting excitation current regulation);

[0208] Control input U: excitation voltage (signal generated by the excitation regulator), compensation of the voltage regulator

[0209] After determining these variables, the matrix , It is used to describe how the state variables evolve over time, while the matrix , Used to describe how state variables affect output.

[0210] Based on the electromagnetic equations of the generator and excitation system, the state equation can be established. For example:

[0211] The relationship between the voltage and current of the excitation winding is determined by the inductance, resistance and input voltage.

[0212] The generator terminal voltage is affected by the flux and rotor current.

[0213] After converting the electromagnetic equations into a discrete time system, we can get the matrix , , , .

[0214] Practical application: If the system parameters, such as inductance, resistance, and time constant are known, the matrix can be directly calculated using the standard mathematical model of the excitation system , , , This method is suitable for excitation systems with detailed physical parameters, such as those available under laboratory test conditions.

[0215] When the system parameters cannot be obtained directly, the system identification method can be used to estimate the matrix through experimental data. , , , This method is suitable for scenarios where accurate modeling is not possible or the excitation system parameters are greatly affected by the environment.

[0216] First, the following data needs to be collected in the actual operating environment of the excitation system:

[0217] Input data U: control signal applied to the excitation system (excitation current adjustment amount);

[0218] Output data Y: record the real-time changes of voltage and current at the generator terminal;

[0219] Status data X: record excitation current, voltage deviation, flux change, etc.;

[0220] Data collection can be completed through power monitoring systems or data acquisition equipment and stored as time series data.

[0221] Based on the input-output data of the excitation system, the matrix can be estimated using the following method , , , :

[0222] The least squares method is suitable for linear systems and fits the state equation by minimizing the prediction error. It requires a large amount of experimental data and is suitable for modeling large-scale excitation systems.

[0223] The recursive least squares method is applicable to the situation where the excitation system parameters change with time, and the matrix parameters can be updated dynamically.

[0224] Kalman filtering is suitable for the identification of systems containing noise by recursively estimating the system state and correcting the matrix parameters.

[0225] Deep learning methods, such as LSTM neural networks, are suitable for complex nonlinear excitation systems but require a large amount of data for training.

[0226] After obtaining the input U and output Y of the excitation system, the linear regression method can be used to calculate the matrix , , , :

[0227] Establish an input-output mapping relationship and find out the relationship between the control input U and the state variable X.

[0228] Apply the least squares method to fit the state space equation and obtain the best estimate matrix , , , .

[0229] In a preferred embodiment of the present invention, the prediction error cumulative trend analysis method is used to calculate the excitation current prediction error term, and the excitation current prediction sequence is adjusted using a dynamic error correction mechanism to obtain an excitation current prediction correction sequence, including:

[0230] According to the excitation current prediction sequence, based on the historical prediction error data, the excitation current prediction error term is calculated;

[0231] According to the excitation current prediction error term, a dynamic error correction mechanism is adopted to calculate the excitation current adjustment correction value;

[0232] in, , For the The excitation current adjustment correction value at the moment, is the error correction coefficient, is the error compensation term, For the The actual excitation current value measured at all times;

[0233] According to the excitation current adjustment correction value, the error compensation is performed on the excitation current prediction sequence to obtain the excitation current prediction correction sequence.

[0234] In the embodiment of the present invention, in the predictive control process of the excitation system, due to the uncertainty of the grid operation state, there may be errors in the prediction of the excitation current. Traditional methods usually use fixed error correction parameters for adjustment, which is difficult to effectively suppress the accumulation of prediction errors, thereby affecting the control accuracy. This embodiment makes the prediction of the excitation current more accurate and improves the stability of the excitation control through prediction error calculation and dynamic error correction mechanism.

[0235] First, the excitation current prediction error term is calculated based on the historical prediction error data. Since the excitation system is affected by many factors during actual operation, including load fluctuations, system parameter drift, etc., there is a certain deviation between the predicted value and the actual value of the excitation current. Therefore, by collecting the prediction error data at multiple times and using the error accumulation trend analysis method, the error change pattern can be identified, thereby determining whether the error is a random error, a systematic deviation, or a periodic error.

[0236] In the error correction stage, a dynamic error correction mechanism is used to calculate the excitation current adjustment correction value based on the cumulative trend of the error. Compared with the traditional fixed correction method, this method can maintain a smaller correction amount when the error is small to avoid over-adjustment of the system; when the error is large, the system can adaptively increase the correction amplitude to speed up the error convergence and improve the accuracy of the excitation current prediction.

[0237] Finally, according to the excitation current adjustment correction value, the error compensation of the excitation current prediction sequence is performed to obtain a more accurate excitation current prediction correction sequence. Compared with the traditional method, this technology can adaptively adjust the excitation current prediction value, so that the excitation control system can maintain a high control accuracy under different load conditions, and improve the overall stability and robustness of the system.

[0238] In a preferred embodiment of the present invention, the optimization objective function of the excitation system is constructed according to the excitation current prediction correction sequence, the optimization objective function includes the generator voltage stability objective term and the excitation current regulation smoothness objective term, and the initial weight value of the optimization objective function is calculated based on the weighted cumulative error analysis method, including:

[0239] According to the excitation current prediction correction sequence, the generator voltage deviation and the excitation current regulation change rate are calculated, and an optimization objective function is constructed, wherein the optimization objective function includes a generator voltage stability objective item and an excitation current regulation smoothness objective item; ,in, To optimize the objective function, is the prediction time step, and To optimize the weight coefficient of the objective function, For the The voltage deviation at the moment, The set voltage indicates the target voltage that the generator should maintain. For the The actual voltage at the moment, For the The rate of change of the excitation current at time For the The excitation current at time For the The excitation current at the moment;

[0240] According to the historical operation data, the weighted cumulative error analysis method is used to calculate the initial weight value of the optimization objective function;

[0241] According to the initial weight value of the optimization objective function, the optimization objective constraint relationship is established.

[0242] In the embodiment of the present invention, in the optimization calculation process of the excitation system, the reasonable construction of the optimization objective function directly determines the accuracy and stability of the control strategy. In order to ensure that the excitation system can maintain the stability of the generator terminal voltage and avoid large fluctuations in the excitation current, this embodiment constructs an optimization objective function and calculates its initial weight value based on historical operation data, so that the optimization calculation can more accurately meet the power grid operation requirements.

[0243] In the optimization objective function construction stage, based on the excitation current prediction correction sequence, the generator voltage deviation and the excitation current regulation change rate are calculated and used as the core parameters of the optimization objective. Traditional excitation control methods often focus on a single objective, such as optimizing voltage stability only, while ignoring the smoothness of excitation current adjustment, which may cause the excitation system to produce large current fluctuations during the adjustment process, thereby affecting system stability. The present invention introduces the excitation current regulation smoothness objective item in the optimization objective, so that the optimization calculation can suppress the large changes in excitation current while ensuring voltage stability, thereby improving the smoothness and robustness of control.

[0244] In the weight calculation stage, in order to ensure that the optimization objective function has a reasonable weight distribution under different operating conditions, the weighted cumulative error analysis method is used to calculate the initial weight value of the optimization objective function. Due to the influence of grid load changes and voltage disturbances, voltage deviation and excitation current adjustment requirements may have different importance in different time periods. The traditional fixed weight optimization method is difficult to adapt to the dynamic changes of the power grid, while the weighted cumulative error analysis method dynamically calculates the initial weight values ​​of the voltage stability objective item and the excitation current regulation objective item by analyzing historical error data, so that the optimization objective function is more in line with actual needs during optimization calculation.

[0245] Finally, based on the calculated initial weight value, the optimization objective constraint relationship is established to ensure that the optimization calculation results can improve the regulation accuracy of the excitation system while meeting the stability of the power grid operation. Compared with the traditional method, the construction of this optimization objective function not only improves the accuracy of voltage control, but also makes the excitation current regulation smoother, reduces the power grid fluctuations caused by frequent adjustments, and improves the overall operation stability.

[0246] In a preferred embodiment of the present invention, adjusting the initial weight value of the optimization objective function according to the dynamic load characteristics of the power grid to generate an optimization weight parameter set includes:

[0247] Calculate the optimization target weight adjustment coefficient according to the current grid load level, load change rate and system operation stability;

[0248] According to the initial weight value of the optimization objective function and the optimization objective weight adjustment coefficient, the optimization objective weight is adaptively adjusted to generate an optimization weight parameter set;

[0249] The optimized weight parameter set includes a weight coefficient and ,in, , in, is the initial weight value of the generator voltage stability objective term, is the time window for cumulative calculation, For the The actual voltage at the moment, is the absolute value of the voltage error, indicating the voltage deviation of the generator. is the attenuation coefficient, is the weight adjustment coefficient, is the active power change rate, which indicates the rate of change of the grid load; , in, is the initial weight value of the excitation current adjustment smoothness objective term, For the The excitation current at time For the The excitation current at time is the excitation current variation amplitude, is the attenuation coefficient, is the weight adjustment factor, is the reactive power change rate, indicating the fluctuation of reactive power in the power grid;

[0250] According to the optimization weight parameter set, the optimization priority of the optimization objective function under different operating conditions is dynamically adjusted.

[0251] In the embodiment of the present invention, during the optimization calculation process, although the initial weight value of the optimization objective function can be obtained by calculating the historical error data, due to the dynamic change characteristics of the load level and operation status of the power grid, the fixed initial weight value may not fully meet the system requirements during the actual operation process. In order to further improve the adaptability of the optimization calculation, this embodiment generates an optimization weight parameter set by adjusting the initial weight value of the optimization objective function, so that the optimization calculation result can better adapt to the dynamic characteristics of the power grid.

[0252] In the weight adjustment stage, the optimization target weight adjustment coefficient is first calculated based on the current grid load level, load change rate and system operation stability. The change rate of the grid load directly affects the regulation demand of the excitation current. For example, in the case of a large load mutation, the importance of voltage stability will be much higher than the smoothness of the excitation current regulation. When the grid is running smoothly, the smoothness of the excitation current regulation is more critical. Therefore, by real-time monitoring of the load level and change rate and calculating the optimization target weight adjustment coefficient, it can ensure that the weights of different target items in the optimization calculation process are reasonably distributed.

[0253] In the process of generating the optimization weight parameter set, the initial weight value of the optimization objective function is adaptively adjusted in combination with the calculated optimization target weight adjustment coefficient to form a new optimization weight parameter set. This method avoids the limitations of the traditional fixed weight calculation method and enables the optimization calculation to be more flexible to meet the needs of different working conditions. For example, in the operation stage with large load fluctuations, the voltage stability target weight will be automatically increased to ensure that the voltage fluctuation will not affect the stability of the power grid, while in the operation stage with small load changes, the smoothness weight of the excitation current regulation will be appropriately increased to reduce the impact of frequent adjustments on system equipment.

[0254] Finally, based on the optimization weight parameter set, the optimization priority of the optimization objective function under different operating conditions is dynamically adjusted, so that the excitation control strategy can more accurately adapt to the changes in the grid operation state. Compared with the traditional method, this solution adaptively adjusts the weight of the optimization target, so that the optimization calculation can obtain better regulation effect under different grid operation conditions, and improve the response speed and stability of the excitation system.

[0255] In a preferred embodiment of the present invention, the method of calculating the optimal excitation current adjustment amount by using a constrained optimization solution method according to the optimization objective function and the optimization weight parameter set includes:

[0256] According to the optimization objective function and the optimization weight parameter set, an optimal excitation current adjustment calculation model is constructed, wherein the calculation model includes excitation current adjustment constraint conditions;

[0257] According to the dynamic working range of the excitation system, the excitation current change rate threshold and the voltage stability threshold are set, and the calculation model is solved based on the constraint optimization solution method to obtain the optimal excitation current adjustment amount;

[0258] The calculation formula of the optimal excitation current adjustment amount is: ,in, is the optimal excitation current adjustment, is the excitation current adjustment, is an operator, which indicates the variable value that makes the objective function achieve the minimum value;

[0259] Constraints ( ) ,in, and are the minimum and maximum values ​​of the excitation current, respectively. is the upper limit of the excitation current change rate;

[0260] An optimal excitation control signal is generated according to the optimal excitation current adjustment amount.

[0261] In the embodiment of the present invention, in the optimization calculation of the excitation system, only dynamically adjusting the weight of the optimization target is not enough to ensure the global optimality of the optimization result. In the actual execution of the optimization calculation, it is also necessary to combine the constraints of the excitation system and use the constraint optimization solution method to calculate the optimal excitation current adjustment amount to ensure that the calculation result of the final control signal not only meets the optimization target, but also can operate within the physical constraint range to prevent abnormal changes in the excitation current from affecting the system.

[0262] In the optimization solution stage, firstly, the optimal excitation current adjustment calculation model is constructed according to the optimization objective function and the optimization weight parameter set, and the excitation current adjustment constraint conditions are introduced into the calculation model. The adjustment range of the excitation current is usually limited by the physical characteristics of the equipment, such as the maximum load current of the excitation winding, the voltage adjustment range of the generator, etc. Therefore, in the optimization calculation process, it is necessary to reasonably constrain the excitation current adjustment amount to ensure that the calculation results can operate within the physical safety range.

[0263] Subsequently, based on the dynamic working range of the excitation system, the excitation current change rate and voltage stability threshold are set, and the calculation model is solved based on the constrained optimization solution method to obtain the optimal excitation current adjustment amount. Traditional optimization methods often use fixed thresholds for calculations and cannot dynamically adapt to changes in the system's operating status. This solution dynamically constrains the adjustment range of the excitation current by combining the system's real-time operating data, making the optimization results more in line with actual operating requirements. For example, when the load changes suddenly in a short period of time, the optimization calculation will allow the excitation current adjustment range to increase appropriately to ensure the stability of the grid voltage. When the grid is operating stably, the excitation current adjustment range will be automatically reduced to reduce unnecessary power loss.

[0264] Finally, according to the calculated optimal excitation current adjustment, the optimal excitation control signal is generated and sent to the excitation regulation module to achieve dynamic optimization and control of the excitation current. Compared with the traditional method, this solution not only improves the accuracy of excitation current calculation through the optimization solution method, but also introduces dynamic constraints, so that the optimization results can better meet the operating constraints of the excitation system, improving the safety and stability of system operation.

[0265] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. Generator excitation control optimization system based on model prediction, characterized in that: The system comprises: The data acquisition module is used to collect the operating parameters of the generator, including the generator terminal voltage, stator current, rotor current, speed and grid frequency, and filter the operating parameters to obtain pre-processed data; The state estimation module is used to calculate the state variables of the generator excitation system according to the preprocessed data, including voltage deviation, excitation current change rate, flux change rate and speed deviation, and to construct the state space model of the generator excitation system according to the state variables; The model predictive control module is used to calculate the predicted values ​​of the excitation current at multiple moments in the future according to the state space model by using the rolling time domain optimization method to obtain the excitation current prediction sequence, and dynamically correct the excitation current prediction sequence based on the prediction error correction mechanism to obtain the excitation current prediction correction sequence; The optimization calculation module is used to construct the target optimization function according to the excitation current prediction correction sequence, including the generator voltage stability target item and the excitation current regulation smoothness target item, and calculate the optimal excitation current adjustment amount by using the constrained optimization method to obtain the optimal excitation control signal; The excitation regulation module is used to control the generator excitation system according to the optimal excitation control signal to adjust the excitation current.

2. The generator excitation control optimization system based on model prediction according to claim 1 is characterized in that: The state estimation module comprises: A data correction unit, used to suppress noise on preprocessed data based on a dynamic adaptive filtering method, and calculate correction data in combination with historical operation data; A state variable calculation unit is used to calculate the state variables of the excitation system according to the correction data, including voltage deviation, excitation current change rate, flux change rate and speed deviation, and to construct an initial estimate of the state space model based on a state variable dynamic fitting method; The error compensation unit is used to calculate the state prediction error according to the initial estimated value of the state space model, and dynamically adjust the state variables using an incremental correction strategy to obtain a corrected state space model.

3. The generator excitation control optimization system based on model prediction according to claim 2 is characterized in that: The model predictive control module includes: A future time domain prediction unit is used to calculate the state variables of the excitation system at multiple future moments according to the modified state space model, and to calculate the predicted values ​​of the excitation current at multiple future moments using a sliding window strategy to obtain an excitation current prediction sequence; The prediction error correction unit is used to calculate the excitation current prediction error term based on the prediction error cumulative trend analysis method, and use a dynamic error correction mechanism to adjust the excitation current prediction sequence to obtain the excitation current prediction correction sequence.

4. The generator excitation control optimization system based on model prediction according to claim 3 is characterized in that: The optimization calculation module comprises: An objective function construction unit is used to construct an optimization objective function of the excitation system according to the excitation current prediction correction sequence, wherein the optimization objective function includes a generator voltage stability objective item and an excitation current regulation smoothness objective item, and calculates an initial weight value of the optimization objective function based on a weighted cumulative error analysis method; The target weight adjustment unit is used to adjust the initial weight value of the optimization target function according to the dynamic load characteristics of the power grid and generate an optimization weight parameter set; The constrained optimization calculation unit is used to calculate the optimal excitation current adjustment amount using a constrained optimization solution method according to the optimization objective function and the optimization weight parameter set.

5. The generator excitation control optimization system based on model prediction according to claim 4 is characterized in that: The excitation system state variables are calculated according to the correction data, including voltage deviation, excitation current change rate, flux change rate and speed deviation, and the initial estimate of the state space model is constructed based on the state variable dynamic fitting method, including: According to the correction data, the voltage deviation, the excitation current change rate, the flux change rate and the speed deviation are calculated, and the variable weight adaptive normalization method is used for different state variables to obtain the normalized state variables; According to the normalized state variables, the initial state fitting model is constructed by using the nonlinear multi-order curve fitting method; According to the initial state fitting model, the time trend term of the state variable is calculated, and the initial estimate of the state space model is constructed in combination with the historical state data.

6. The generator excitation control optimization system based on model prediction according to claim 5 is characterized in that: The state prediction error is calculated according to the initial estimated value of the state space model, and the state variable is dynamically adjusted by using an incremental correction strategy to obtain a corrected state space model, including: According to the initial estimate of the state space model, the state prediction error is calculated based on the actual state variables and the predicted state variables at the current moment; According to the state prediction error, the state increment correction term is calculated by using the dynamic error compensation method; According to the state increment correction term, the initial estimation value of the state space model is dynamically adjusted to obtain a corrected state space model.

7. The generator excitation control optimization system based on model prediction according to claim 6 is characterized in that: The method of calculating the excitation system state variables at multiple future moments according to the modified state space model and calculating the excitation current prediction values ​​at multiple future moments using a sliding window strategy to obtain an excitation current prediction sequence includes: According to the modified state space model and based on the recursive state update method, the state variables of the excitation system at multiple moments in the future are calculated; According to the state variables of the excitation system at multiple future moments, a sliding window strategy is adopted to calculate the predicted values ​​of the excitation current at multiple future moments; According to the predicted values ​​of the excitation current at multiple future moments, an excitation current prediction sequence is constructed.

8. The generator excitation control optimization system based on model prediction according to claim 7 is characterized in that: The method for analyzing the cumulative trend of prediction errors is used to calculate the excitation current prediction error term, and a dynamic error correction mechanism is used to adjust the excitation current prediction sequence to obtain the excitation current prediction correction sequence, including: According to the excitation current prediction sequence, based on the historical prediction error data, the excitation current prediction error term is calculated; According to the excitation current prediction error term, a dynamic error correction mechanism is adopted to calculate the excitation current adjustment correction value; According to the excitation current adjustment correction value, the error compensation is performed on the excitation current prediction sequence to obtain the excitation current prediction correction sequence.

9. The generator excitation control optimization system based on model prediction according to claim 8, characterized in that: According to the excitation current prediction correction sequence, an optimization objective function of the excitation system is constructed, the optimization objective function includes a generator voltage stability objective item and an excitation current regulation smoothness objective item, and an initial weight value of the optimization objective function is calculated based on a weighted cumulative error analysis method, including: According to the excitation current prediction correction sequence, the generator voltage deviation and the excitation current regulation change rate are calculated, and an optimization objective function is constructed, wherein the optimization objective function includes a generator voltage stability objective item and an excitation current regulation smoothness objective item; ,in, To optimize the objective function, is the prediction time step, and To optimize the weight coefficient of the objective function, For the The voltage deviation at the moment, The set voltage indicates the target voltage that the generator should maintain. For the The actual voltage at the moment, For the The rate of change of the excitation current at time For the The excitation current at time For the The excitation current at the moment; According to the historical operation data, the weighted cumulative error analysis method is used to calculate the initial weight value of the optimization objective function; According to the initial weight value of the optimization objective function, the optimization objective constraint relationship is established.

10. The generator excitation control optimization system based on model prediction according to claim 9, characterized in that: The method of calculating the optimal excitation current adjustment amount by using a constrained optimization solution method according to the optimization objective function and the optimization weight parameter set includes: According to the optimization objective function and the optimization weight parameter set, an optimal excitation current adjustment calculation model is constructed, wherein the calculation model includes excitation current adjustment constraint conditions; According to the dynamic working range of the excitation system, the excitation current change rate threshold and the voltage stability threshold are set, and the calculation model is solved based on the constraint optimization solution method to obtain the optimal excitation current adjustment amount; The calculation formula of the optimal excitation current adjustment amount is: ,in, is the optimal excitation current adjustment, is the excitation current adjustment, is an operator, which indicates the variable value that makes the objective function achieve the minimum value; And meet ,in, and are the minimum and maximum values ​​of the excitation current, respectively. is the upper limit of the excitation current change rate; An optimal excitation control signal is generated according to the optimal excitation current adjustment amount.

Images