Channel time delay estimation method, device and equipment for boiler-turbine coordination system of thermal power generating unit

Abstract

The application discloses a kind of boiler coordination system channel time delay estimation method, device and equipment of thermal power unit, and it relates to industrial control system technical field.The method is in the operation process of thermal power unit, the time sequence sample corresponding to the input variable and output variable of boiler coordination system is collected;Zero-phase filtering and difference processing are carried out to the time sequence sample of each variable respectively, and stationary difference signal sample is obtained;After resampling to difference signal sample, the conditional mutual information value under different lag order of each input variable and each output variable is calculated;For each input variable and each output variable between channel, the lag order corresponding to the maximum conditional mutual information value is determined as the time delay estimation value of the channel, and the time delay matrix is comprehensively obtained.The application quantifies the causal contribution of input to output by conditional mutual information value, eliminates false correlation interference, avoids delay distortion when deep peak shaving, and improves the response accuracy of thermal power unit to automatic generation control instruction under smart grid.

Description

Technical Field

[0001] This invention relates to the field of industrial control system technology, specifically to a method, apparatus, and equipment for estimating channel delay in the boiler-turbine coordination system of a thermal power unit. Background Technology

[0002] With the rapid advancement of smart grid construction and the large-scale grid connection of new energy sources such as wind power and photovoltaics, thermal power units need to frequently participate in deep peak shaving of the power grid, and load commands exhibit characteristics of wide-range and rapid changes. The core of a thermal power unit is the boiler-turbine coordination system (boiler and turbine), which is a multiple-input multiple-output (MIMO) system with large delays and strong coupling. Its actuators include the coal feeder, feedwater control valve, and turbine high-pressure control valve, and the controlled variables include the main steam pressure, the enthalpy at the steam-water separator outlet, and the unit power. Among them, the turbine side responds extremely quickly (on the order of seconds), while the boiler side has huge physical inertia and transmission delays. From the change in coal feed command to the response of main steam pressure, there is a delay of tens to hundreds of seconds.

[0003] In the field of industrial process control, establishing high-precision dynamic mathematical models is fundamental to achieving predictive and optimal control. When using Model Predictive Control (MPC) to optimize boiler-turbine coordination, the aforementioned time delay parameters are the core basis for the time axis registration of the internal predictive model. If there is a deviation in time delay identification, it will cause the MPC predictive output to be misaligned with the actual physical state of the unit on the time axis, resulting in pressure overshoot or severe oscillations, and in severe cases, triggering unit protection tripping. Therefore, accurately detecting the channel time delay between each actuator and each sensor is crucial to ensuring the stability and robustness of MPC control.

[0004] In existing time delay estimation methods, model-driven methods (such as model structure search and parametric joint estimation) estimate time delay as a model parameter. The effectiveness of these methods is highly dependent on the matching degree of the model structure; model mismatch can lead to systematic biases, and they also have high algorithm complexity and computational overhead. Data-driven methods (such as cross-correlation analysis, information theory analysis, and geometric and time-domain feature analysis) directly utilize the statistical characteristics of input and output signals for identification, making it difficult to distinguish between the coupling relationship between signals and the actual physical lag. Under nonlinear strong interference or closed-loop conditions, they are prone to time delay estimation distortion. Therefore, existing time delay estimation methods are insufficient to meet the engineering requirements of accurate time delay parameter identification by MPC under deep peak-shaving conditions in thermal power units. Summary of the Invention

[0005] The technical problem to be solved by the present invention is how to solve the time delay distortion problem of existing time delay estimation methods under deep peak shaving conditions of power units. The purpose is to provide a time delay estimation method, device and equipment for the boiler-generator coordination system of thermal power units, which solves the above-mentioned problem.

[0006] This invention is achieved through the following technical solution:

[0007] In a first aspect, the present invention provides a method for estimating channel delay in the boiler-turbine coordination system of a thermal power unit, comprising:

[0008] During the operation of a thermal power unit, time-series samples corresponding to the input and output variables of the boiler-turbine coordination system are collected; the input variables are the commands issued by the actuators, and the output variables are the sensor measurements.

[0009] Zero-phase filtering and differential processing are performed on the time-series samples of each variable to obtain stable differential signal samples;

[0010] The differential signal samples are resampled to obtain resampled signal samples;

[0011] Based on the resampled signal samples, calculate the conditional mutual information value between each input variable and each output variable at different lag orders;

[0012] For each channel between an input variable and an output variable, the lag order corresponding to the maximum conditional mutual information value is determined as the time delay estimate for that channel. The time delay estimates of all channels are then combined to obtain the time delay matrix.

[0013] Optionally, the input variables include coal feed rate command, water flow rate command, and turbine control valve opening command; the output variables include main steam pressure, steam-water separator outlet enthalpy, and unit power.

[0014] Optionally, the step of performing zero-phase filtering and differential processing on the time-series samples of each variable to obtain stationary differential signal samples includes:

[0015] Symmetric mean filtering is applied to the time series samples of each variable to obtain the filtered time series samples of each variable.

[0016] The filtered time series samples of each variable are simultaneously differentially processed to obtain stable differential signal samples.

[0017] Optionally, resampling the differential signal samples to obtain resampled signal samples includes:

[0018] The differential signal samples are mapped to Euclidean space and normalized. The K-nearest neighbor algorithm is used to estimate the local probability density of each sample point.

[0019] Calculate the inverse probability weight of each sample point based on the local probability density of each sample point;

[0020] According to the distribution of the inverse probability weights, the differential signal samples are subjected to weighted random sampling without replacement to obtain resampled signal samples.

[0021] Optionally, the formula for calculating the local probability density is as follows:

[0022]

[0023] in, For sample points The local probability density; ; This represents the total sample size. The specified number of neighboring nodes; for The volume of a unit hypersphere in 3D space; For sample points Its first The Euclidean distance between the nearest neighbors.

[0024] Optionally, the formula for calculating the inverse probability weight is as follows:

[0025]

[0026] in, For sample points Inverse probability weights; For sample points The local probability density; Sample points The local probability density; This represents the total sample size.

[0027] Optionally, the formula for calculating the conditional mutual information value is as follows:

[0028]

[0029] in, for Output variables at the sampling time; for Output variables at the sampling time; The time lag is the time delay. for Input variables at the sampling time; for Input variables at the sampling time; For variables The joint probability density; For a given under conditions The conditional probability density; For a given under conditions The conditional probability density; Indicates that given historical information Under the conditions, and The conditional mutual information values ​​between them.

[0030] Optionally, after obtaining the delay matrix by combining the delay estimates of all channels, the method further includes:

[0031] Import the time delay matrix into the model predictive controller;

[0032] The model predictive controller performs time axis registration on its internal predictive models according to the time delay matrix, calculates control sequences based on the time axis registered predictive models, and sends them to each actuator.

[0033] In a second aspect, the present invention provides a device for estimating channel delay in a boiler-turbine coordination system for a thermal power unit, comprising:

[0034] The data acquisition module is used to collect time-series samples corresponding to the input and output variables of the boiler-turbine coordination system during the operation of the thermal power unit; the input variables are the commands issued by the actuators, and the output variables are the sensor measurements.

[0035] The preprocessing module is used to perform zero-phase filtering and differential processing on the time-series samples of each variable to obtain stable differential signal samples.

[0036] The resampling module is used to resample the differential signal samples to obtain resampled signal samples;

[0037] The calculation module is used to calculate the conditional mutual information value between each input variable and each output variable under different lag orders based on the resampled signal samples.

[0038] The delay determination module is used to determine the delay estimate of each channel between each input variable and each output variable, starting from zero lag, when the conditional mutual information value first appears significantly higher than zero and is a local maximum. The delay estimate of all channels is then combined to obtain the delay matrix.

[0039] Thirdly, the present invention provides a computer device, the computer device including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the channel delay estimation method for the boiler-turbine coordination system of a thermal power unit as described in any one of the first aspects.

[0040] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0041] This application provides a method for estimating channel delays in the boiler-turbine coordination system of a thermal power unit. This method quantifies the causal contribution of inputs to outputs through conditional mutual information values, effectively eliminating the interference of spurious correlations on the identification results. It avoids the delay distortion problem of traditional data-driven methods under deep peak-shaving conditions in thermal power units, and can still accurately distinguish the true physical delays of each channel under multivariate coupling conditions. Furthermore, this method bypasses the complex and mismatch-prone modeling process, directly utilizing process data to accurately locate delays. Compared to model-driven methods, this method greatly reduces the dependence on prior system knowledge, making the algorithm easier to implement and reducing computational overhead. Compared to conventional data-driven methods, due to the introduction of causal feature constraints, the accuracy and robustness of its estimation results are significantly improved, enabling a more realistic reproduction of the physical lag characteristics of the process. Attached Figure Description

[0042] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings:

[0043] Figure 1 A flowchart illustrating the method for estimating channel delay in the boiler-turbine coordination system of a thermal power unit provided in this application embodiment;

[0044] Figure 2 This is a schematic diagram illustrating the effect of symmetric mean filtering provided in an embodiment of this application;

[0045] Figure 3 This is a schematic diagram showing the comparison of data before and after differential processing provided in an embodiment of this application;

[0046] Figure 4 A schematic diagram of the resampling algorithm provided in the embodiments of this application;

[0047] Figure 5 The time-series sample data corresponding to the input and output variables provided in the embodiments of this application;

[0048] Figure 6 A schematic diagram illustrating the delay estimation results for the nine channels provided in this embodiment of the application;

[0049] Figure 7 This is a schematic diagram showing the comparison of the control effect on the main steam pressure provided in the embodiments of this application;

[0050] Figure 8A schematic diagram showing the comparison of the control effect on the outlet enthalpy of the steam-water separator provided in the embodiments of this application;

[0051] Figure 9 This is a schematic diagram comparing the control effects of unit power provided in the embodiments of this application;

[0052] Figure 10 A schematic diagram of the structure of the channel delay estimation device for the boiler-turbine coordination system of a thermal power unit provided in this application embodiment. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0054] Please refer to Figure 1 This is a flowchart illustrating a method for estimating channel delay in the boiler-turbine coordination system of a thermal power unit according to an embodiment of this application. The following is a detailed explanation of this method. Figure 1 The method for estimating channel delay in the boiler-turbine coordination system of the thermal power unit shown is introduced.

[0055] S1. During the operation of the thermal power unit, collect time-series samples corresponding to the input and output variables of the boiler-turbine coordination system.

[0056] In practical implementation, the boiler-turbine coordination system has the following characteristics: input variables and Each output variable is sampled according to a set sampling period during normal operation of the thermal power unit. Continuously collect time-series samples corresponding to input and output variables to obtain the final dataset. The matrix, where, The sample size is represented by each column, which corresponds to a time series sample of an input or output variable.

[0057] In one possible embodiment, the actuators of the boiler-turbine coordination system include a coal feeder, a feedwater control valve, and a turbine high-pressure control valve. The input variables are the commands issued by the actuators, specifically including coal feed rate commands, feedwater flow rate commands, and turbine control valve opening commands. The sensors of the boiler-turbine coordination system include a main steam pressure sensor, a steam-water separator outlet enthalpy detection unit, and a unit power detection unit. The output variables are the sensor measurements, specifically including main steam pressure, steam-water separator outlet enthalpy, and unit power. These variables collectively constitute the multi-input multi-output signal set of the boiler-turbine coordination system, covering the main energy and mass transfer channels from fuel, feedwater, and steam intake to pressure, temperature, and power.

[0058] S2. Perform zero-phase filtering and differential processing on the time series samples of each variable to obtain stable differential signal samples.

[0059] In one possible embodiment, step S2 includes: performing symmetric mean filtering on the time series samples of each variable to obtain filtered time series samples of each variable; and simultaneously performing differential processing on the filtered time series samples of each variable to obtain stationary differential signal samples.

[0060] In practical implementation, high-frequency interference is introduced by sensor noise and actuator jitter in the industrial field. Therefore, the time series samples of each variable are first filtered to suppress noise interference and improve the signal-to-noise ratio. If the filtering introduces additional phase shift, it will directly cause systematic deviation in the time delay estimation results. Therefore, zero-phase filtering is performed on the time series samples of each variable to obtain the filtered time series samples of each variable.

[0061] Common zero-phase filtering methods include symmetrical mean filtering, forward-backward filtering, frequency-domain zero-phase filtering, finite impulse response (FIR) combined with time-shift filtering, and overlapping segmented filtering. Considering both accuracy and convenience in engineering applications, this embodiment employs symmetrical mean filtering to achieve zero-phase filtering of each input and output signal. Symmetrical mean filtering is a zero-phase filtering method whose core idea is to take the average of the signal values ​​within a preset time window centered on the current sampling point as the filtered output for that point. The window is symmetrical, thus ensuring that the signal phase remains consistent before and after filtering.

[0062] For any variable, the time series sample sequence Let the sample length of the filtering window be... ( (For odd numbers), the mathematical expression for symmetric mean filtering is as follows:

[0063]

[0064] in, For the first The filtered value at each sampling time; For the first The original values ​​at each sampling time; Within the filter window, relative to the center point The offset, with a value range of 1. ; The sample length of the filter window. It is an odd number.

[0065] Please refer to Figure 2This is a schematic diagram illustrating the effect of symmetric mean filtering provided in an embodiment of this application. The blue curve (Original) corresponds to the original data, and the red curve (Filtered) corresponds to the filtered data. From the... and The filtering results show that the symmetrical mean filtering effectively removes high-frequency noise from the original data, making the curve smoother and without introducing additional phase shift.

[0066] Considering the limited sample size of actual processes, the bounded increment of controllers, and the non-stationarity of thermal power units under varying load conditions such as deep peak shaving, which causes the statistical characteristics of some variables to change over time, directly using the raw data for statistical modeling and analysis can easily lead to problems such as biased parameter estimation, spurious correlation, inaccurate hypothesis testing, and extrapolation failure, resulting in inaccurate subsequent time delay estimation. Therefore, the filtered time series samples of each variable are differentially processed to obtain stationary differential signal samples.

[0067] Suppose the multiple-input multiple-output (MIMO) process to be identified has linear time-invariant (LTI) properties, and the discrete-time input-output relationship between the input and output variables is as follows:

[0068]

[0069] in, Input variables of Transformation; For output variables of Transformation; Let this be the transfer function matrix of the process. Differences are applied to both the input and output variables simultaneously.

[0070]

[0071]

[0072] in, Input variables exist The first-order difference value at the sampling time; Input variables exist The first-order difference value at the sampling time.

[0073] but Transformed into:

[0074]

[0075]

[0076] in, Differential signal of Transformation; Differential signal of Transformation; The delay operator is a unit.

[0077] Substitute the original output relationship We can obtain:

[0078]

[0079] The above derivation shows that: after simultaneously differentiating the input and output, the transfer function matrix... The differential processing does not change the system's time delay parameters, ensuring that the time delay estimation results after differential processing can still accurately reflect the signal transmission delay of the original physical channel, thus ensuring the physical consistency between the preprocessing operation and the subsequent time delay identification.

[0080] Please refer to Figure 3 This is a schematic diagram comparing data before and after differential processing according to an embodiment of this application. The blue curve on the left (Original) represents the data before differential processing (i.e., the filtered time series sample), and the red curve on the right (Differenced) represents the data after differential processing (i.e., the differential signal sample). It can be seen that after differential processing, the input variable... and The curve exhibits an approximate pulse shape, while the output variable... and The curve can be regarded as a , The response to the pulse input. Furthermore, compared to before differential processing, the curves of each variable are more stable after differential processing, indicating that differential processing effectively eliminates deterministic non-stationary components such as trend terms, enabling the differential signal samples to meet the stationarity requirements of time-series causality tests. This fundamentally avoids false causal inferences caused by non-stationary data, providing a reliable sample basis for subsequent time delay estimation.

[0081] S3. Resample the differential signal sample to obtain the resampled signal sample.

[0082] In one possible embodiment, step S3 includes: mapping the differential signal samples to Euclidean space and normalizing them; estimating the local probability density of each sample point using the K-nearest neighbor algorithm; calculating the inverse probability weight of each sample point based on the local probability density of each sample point; and performing weighted random sampling without replacement on the differential signal samples according to the distribution of the inverse probability weights to obtain the resampled signal samples.

[0083] In the specific implementation process, the differential signal samples are first mapped to Euclidean space and then normalized to eliminate the influence of the dimensional differences of different variables on the calculation results. The normalization formula is as follows:

[0084]

[0085] in, These are the normalized sample values, with a range of [value range missing]. ; These are the original sample values; It is the minimum value of this variable among all samples; This is the maximum value of the variable across all samples.

[0086] Secondly, utilize The K-Nearest Neighbors (KNN) nonparametric method estimates the local probability density at each sample point. In the high-dimensional space where the sample resides, a given sample point... around The smaller the volume occupied by each neighboring point, the higher the probability density of that point. The corresponding calculation formula is as follows:

[0087]

[0088] in, For sample points The local probability density; ; This represents the total sample size. The specified number of neighboring nodes; For sample points Its first Euclidean distance between nearest neighbors; for The volume of a unit hypersphere in 3D space is calculated using the following formula:

[0089]

[0090] in, The dimension of the feature space is equal to the total number of variables in the difference signal sample; This is a gamma function.

[0091] Then, based on the local probability density of each sample point, the inverse probability weight of each sample point is calculated:

[0092]

[0093] in, For sample points Inverse probability weights; For sample points The local probability density; Sample points The local probability density; This represents the total sample size.

[0094] Finally, the inverse probability weights for each sample point are used. As a probability weight for its selection, the higher the weight of the sample (i.e., the sample in the sparse region of the feature space, corresponding to the key dynamic sample in the dynamic transition process of the unit), the greater the probability of being selected. Weighted random sampling without replacement is performed on the differential signal samples. Each sample is not replaced after being selected, that is, each original sample is selected at most once. The selected samples constitute the resampled signal sample set.

[0095] Considering that the boiler-generator coordination system of thermal power units operates in a steady or quasi-steady state most of the time, only entering a dynamic transition process when load commands change or disturbances occur, this results in a large proportion of stationary samples with values ​​near zero in the differential sample set, while the proportion of effective samples that truly reflect the dynamic causal relationship of the system is extremely low. If all differential samples are used with equal weight for subsequent causal tests, a large number of redundant steady-state samples will dominate the statistical results, making it difficult to detect significant peaks in the conditional mutual information value at the correct lag time, leading to a high false negative rate in time delay estimation.

[0096] This application provides a resampling method based on inverse probability weighting (i.e., KNN-IPWS resampling), the principle of which is as follows: Figure 4 As shown. The original sample point set will be... The set of sample points obtained by KNN-IPWS resampling is denoted as .

[0097] In this embodiment, the local sparsity of the sample in the feature space is used as the sampling probability to reduce the retention probability of redundant samples that occur frequently in the original data (i.e., samples that are in dense regions of the feature space and correspond to the steady-state or quasi-steady-state operation of the unit), while increasing the retention probability of feature samples that are in boundary or sparse regions (i.e., samples that correspond to dynamic transition processes such as unit load changes and combustion disturbances). This significantly increases the proportion of dynamic causal information samples in the resampled sample set, thereby improving the sensitivity of subsequent instantaneous information transmission causal testing and reducing the false negative rate of time delay estimation.

[0098] S4. Based on the resampled signal samples, calculate the conditional mutual information value between each input variable and each output variable under different lag orders.

[0099] The formula for calculating the Conditional Mutual Information (CMI) value is as follows:

[0100]

[0101] in, for Output variables at the sampling time; for Output variables at the sampling time; The time lag is the time delay.

[0102] for Input variables at the sampling time; for Input variables at the sampling time;

[0103] For variables The joint probability density;

[0104] For a given under conditions The conditional probability density;

[0105] For a given under conditions The conditional probability density;

[0106] Indicates that given historical information Under the conditions, and The conditional mutual information value between them. That is, given the input value at the previous time step. and the output value of the previous time step Under the condition of lag Input variables at time The output variable at the current time step How much additional causal information is provided.

[0107] For continuous variables in the boiler-generator coordination system of thermal power units, such as main steam pressure, steam-water separator outlet enthalpy, and unit power, the conditional mutual information value cannot be directly calculated by summation and needs to be estimated using numerical methods. Preferably, kernel density estimation (KDE) or KSG (Kraskov-Stögbauer-Grassberger) estimators can be used to calculate the probability density and conditional mutual information (CMI) values ​​in the formulas.

[0108] S5. For each channel between an input variable and an output variable, determine the lag order corresponding to the maximum conditional mutual information value as the time delay estimate of that channel, and combine the time delay estimates of all channels to obtain the time delay matrix.

[0109] In the specific implementation process, the boiler-turbine coordination system has 3 input variables and 3 output variables, forming a total of 9 input-output channels, namely: the channel between the coal feed command and the main steam pressure, the channel between the coal feed command and the steam-water separator outlet enthalpy, the channel between the coal feed command and the unit power, the channel between the feedwater flow command and the main steam pressure, the channel between the feedwater flow command and the steam-water separator outlet enthalpy, the channel between the feedwater flow command and the unit power, the channel between the turbine control valve opening command and the main steam pressure, the channel between the turbine control valve opening command and the steam-water separator outlet enthalpy, and the channel between the turbine control valve opening command and the unit power.

[0110] From the curves showing the change of conditional mutual information with lag order for each channel, the lag order corresponding to the maximum conditional mutual information value is determined. This lag order is then used as the time delay estimate for that channel. By combining the time delay estimates of all channels, a time delay matrix is ​​obtained. The elements in the time delay matrix... Indicates the first Input variables and the first The estimated time delay between channels of each output variable.

[0111] In one possible embodiment, after obtaining the delay matrix by integrating the delay estimates of all channels, the method further includes: importing the delay matrix into the model predictive controller; the model predictive controller performs time axis registration on its internal predictive models according to the delay matrix, and calculates control sequences based on the time axis registered predictive models and sends them to each actuator.

[0112] In practice, after obtaining the time delay matrix, it is imported as a key parameter into the internal predictive model of the Model Predictive Controller (MPC). The MPC adjusts the alignment of its input-output time series based on this time delay matrix, enabling it to anticipate future control actions when calculating future control sequences. It takes several seconds to produce a response in the actual physical output.

[0113] For example, when the coal feed instruction To the main steam pressure The time delay between responses is At a given second, the model predictive controller incorporates this information into its prediction time domain, providing a 45-second lead time when controlling incremental output. This ensures that the pressure response trough and coal feeding compensation are precisely matched on the time axis, thereby effectively suppressing pressure fluctuations and significantly improving the control quality of thermal power units under deep peak shaving conditions.

[0114] The following is a specific embodiment to illustrate the method for estimating channel delay in the boiler-turbine coordination system of thermal power units provided in this application.

[0115] Taking the simulation of an ultra-supercritical unit as an example, the input variables include the coal feed command. Water supply flow instructions Steam turbine control valve opening command The output variables include the main steam pressure. Enthalpy value at the outlet of the steam-water separator Unit power . Figure 5 These are the time series sample data corresponding to the input and output variables.

[0116] against Figure 5 The time series sample shown is obtained by using the time delay estimation method for the boiler-turbine coordination system of thermal power units provided in this application, resulting in time delay estimation results for a total of 9 channels, as shown below. Figure 6 As shown. The 9 channels include: arrive passage arrive passage arrive passage arrive passage arrive passage arrive passage arrive passage arrive passage arrive The channel. The red solid line (BT) represents the input-output causal strength characterized by conditional mutual information (CMI) at each lag order Lag. The pink area is the 95% confidence interval corresponding to CMI. The black dashed line is the causal invalidation threshold when CMI=0. The horizontal axis Lag represents the lag order.

[0117] from Figure 6 It can be seen that each input-output coupling curve has a significant peak, and the CMI value at the peak position is significantly higher than the zero baseline (black dashed line). The lag order corresponding to this peak is the estimated time delay value of the corresponding channel. At other lag steps deviating from the peak, the CMI value falls back to near the zero line, and the causal effect is not significant. The Lag position and peak height of the peaks in different coupling paths are different, accurately reflecting the coal feed instruction. Water supply flow instructions Steam turbine control valve opening command Three types of operational disturbances are transmitted to the main steam pressure. Enthalpy value at the outlet of the steam-water separator Unit power The transmission speed and coupling strength differ. Meanwhile, the narrow, stable 95% confidence interval demonstrates that the method of this invention reduces the statistical error of mutual information estimation and exhibits outstanding resistance to temporal noise.

[0118] will pass Figure 6 Please refer to the process of embedding the identified delay parameters into the MPC model identification. Figure 7 This is a schematic diagram comparing the control effect on the main steam pressure provided in the embodiments of this application. The horizontal axis represents the simulation time (seconds), and the vertical axis represents the steam pressure (i.e., the main steam pressure). Please refer to... Figure 8 This is a schematic diagram comparing the control effect on the enthalpy value of the steam-water separator outlet provided in the embodiments of this application. The horizontal axis represents the simulation time (seconds), and the vertical axis represents the enthalpy value at the midpoint (i.e., the enthalpy value at the steam-water separator outlet). Please refer to... Figure 9 This is a schematic diagram comparing the control effect of unit power provided in the embodiments of this application. The horizontal axis represents the simulation time (seconds), and the vertical axis represents the power (i.e., unit power).

[0119] Figure 7 , Figure 8 and Figure 9 The black dashed line represents the parameter setting value, the blue solid line (MPC based on KNN-IPWS-MIT delay identification) represents the MPC control effect of introducing the delay of this invention, and the red dashed line (MPC without delay identification) represents the traditional MPC control effect without introducing the delay of this invention.

[0120] like Figure 7 As shown, under the main steam pressure control condition, the two curves have similar initial descent rates, but the red curve without the time delay of this invention has a slower descent rate and a slower convergence rate in the later stage. The blue curve with the time delay of this invention, however, approaches the set value more quickly, eventually smoothly conforming to the set reference, exhibiting smaller steady-state deviation, lower fluctuations during the transition phase, and earlier steady-state convergence of the main steam pressure.

[0121] like Figure 8 As shown, the performance difference between the two curves is significant in the control of the enthalpy value at the outlet of the steam-water separator. The red curve without the time delay of this invention exhibits large-amplitude bidirectional oscillations, first plunging sharply to a trough, followed by a significant overshoot peak, with prolonged oscillation convergence. In contrast, the blue curve with the time delay of this invention significantly reduces the maximum dynamic overshoot and weakens the oscillation amplitude. After the trough falls back, it smoothly approaches the set value, the oscillation period is greatly shortened, and the dynamic process is smoother. This effectively avoids the problem of repeated enthalpy fluctuations caused by inaccurate time delay prediction, and has a significant effect on improving the parameter stability of the intermediate links of the boiler.

[0122] like Figure 9As shown, in the unit power control, the two curves initially show similar downward trends. The red curve, which does not employ the time delay of this invention, lags behind in approaching steady state, and exhibits a slight, persistent offset at its tail. In contrast, the blue curve, which incorporates the time delay of this invention, reaches the target setting faster, has a shorter dynamic transition range, and its steady-state curve closely follows the set dashed line, resulting in better power output stability and improved load adaptation quality on the grid side of the unit.

[0123] In summary, traditional MPC without precise time delay suffers from prediction errors due to deviations in model time delay parameters, leading to larger dynamic overshoot, longer settling times, and subsequent oscillations in the controlled parameters. By using the precise time delay obtained through the method described in this application to optimize MPC modeling, the controller compensates for hysteresis characteristics ahead of time. This achieves triple optimization—dynamic overshoot suppression, reduced settling time, and lower steady-state error—in all three coupled channels: main steam pressure, steam-water separator outlet enthalpy, and unit power. Experimental results demonstrate that the proposed time delay estimation method for the boiler-turbine coordination system of thermal power units significantly improves model prediction accuracy, enhances tracking control performance under conditions of large-scale setpoint step changes, exhibits better control robustness, and is suitable for the complex actual operating conditions of multi-stage heat transfer delays on the thermal power boiler side.

[0124] Based on the same inventive concept, please refer to Figure 10 This application also provides a device for estimating channel delay in the boiler-turbine coordination system of a thermal power unit, the device comprising:

[0125] The data acquisition module is used to collect time-series samples of the input and output variables of the boiler-turbine coordination system during the operation of the thermal power unit; the input variables are the commands issued by the actuators, and the output variables are the sensor measurements.

[0126] The preprocessing module is used to perform zero-phase filtering and differential processing on the time-series samples of each variable to obtain stable differential signal samples.

[0127] The resampling module is used to resample the differential signal samples to obtain resampled signal samples;

[0128] The calculation module is used to calculate the conditional mutual information value between each input variable and each output variable under different lag orders based on the resampled signal samples;

[0129] The delay determination module is used to determine the delay estimate of each channel between each input variable and each output variable, starting from zero lag, when the conditional mutual information value first appears significantly higher than zero and is a local maximum. The delay estimate of all channels is then combined to obtain the delay matrix.

[0130] Optionally, input variables include coal feed rate command, water flow rate command, and turbine control valve opening command; output variables include main steam pressure, steam-water separator outlet enthalpy, and unit power.

[0131] Optionally, the preprocessing module is specifically used for:

[0132] Symmetric mean filtering is applied to the time series samples of each variable to obtain the filtered time series samples of each variable.

[0133] The filtered time series samples of each variable are simultaneously differentially processed to obtain stable differential signal samples.

[0134] Optionally, the resampling module is specifically used for:

[0135] The differential signal samples are mapped to Euclidean space and normalized. The K-nearest neighbor algorithm is used to estimate the local probability density of each sample point.

[0136] Calculate the inverse probability weight of each sample point based on the local probability density of each sample point;

[0137] The differential signal samples are subjected to weighted random sampling without replacement according to the distribution of inverse probability weights to obtain resampled signal samples.

[0138] Optionally, the formula for calculating the local probability density is as follows:

[0139]

[0140] in, For sample points The local probability density; ; This represents the total sample size. The specified number of neighboring nodes; for The volume of a unit hypersphere in 3D space; For sample points Its first The Euclidean distance between the nearest neighbors.

[0141] Optionally, the formula for calculating the inverse probability weight is as follows:

[0142]

[0143] in, For sample points Inverse probability weights; For sample points The local probability density; Sample points The local probability density; This represents the total sample size.

[0144] Optionally, the formula for calculating the conditional mutual information value is as follows:

[0145]

[0146] in, for Output variables at the sampling time; for Output variables at the sampling time; The time lag is the time delay. for Input variables at the sampling time; for Input variables at the sampling time; For variables The joint probability density; For a given under conditions The conditional probability density; For a given under conditions The conditional probability density; Indicates that given historical information Under the conditions, and The conditional mutual information values ​​between them.

[0147] Optionally, the device also includes an import module, which is used for:

[0148] After combining the time delay estimates of all channels to obtain the time delay matrix, the time delay matrix is ​​imported into the model predictive controller.

[0149] The model predictive controller performs time axis registration on its internal predictive models according to the time delay matrix, calculates control sequences based on the time axis registered predictive models, and sends them to each actuator.

[0150] It should be noted that each module in the boiler-turbine coordination system channel delay estimation device of the thermal power unit in this embodiment corresponds one-to-one with each step in the boiler-turbine coordination system channel delay estimation method of the thermal power unit in the aforementioned embodiment. Therefore, the specific implementation of this embodiment can refer to the implementation of the aforementioned boiler-turbine coordination system channel delay estimation method of the thermal power unit, and will not be repeated here.

[0151] Based on the same inventive concept, this application also provides a computer device, which includes a processor, a memory, and a computer program stored in the memory. The computer program is executed by the processor to implement the aforementioned method for estimating the channel delay of the boiler-turbine coordination system of a thermal power unit.

[0152] Based on the same inventive concept, this application also provides a computer storage medium storing a computer program, which is executed by a processor to implement the aforementioned method for estimating channel delay of the boiler-turbine coordination system of a thermal power unit.

[0153] In some embodiments, the computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or it may be a device including one or any combination of the above-mentioned memories. The computer may be a variety of computing devices, including smart terminals and servers.

[0154] In some embodiments, executable instructions may take the form of a program, software, software module, script, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a standalone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0155] As an example, executable instructions may, but do not necessarily, correspond to files in the file system. They may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a Hyper Text Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple collaborative files (e.g., a file that stores one or more modules, subroutines, or code sections).

[0156] As an example, executable instructions can be deployed to execute on a single computing device, or on multiple computing devices located in one location, or on multiple computing devices distributed across multiple locations and interconnected via a communication network.

[0157] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0158] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for estimating channel delay in the boiler-turbine coordination system of a thermal power unit, characterized in that, include: During the operation of thermal power units, time-series samples corresponding to the input and output variables of the boiler-turbine coordination system are collected. The input variable is the instruction issued by the actuator, and the output variable is the sensor measurement value; Zero-phase filtering and differential processing are performed on the time-series samples of each variable to obtain stable differential signal samples; The differential signal samples are resampled to obtain resampled signal samples; Based on the resampled signal samples, calculate the conditional mutual information value between each input variable and each output variable at different lag orders; For each channel between an input variable and an output variable, the lag order corresponding to the maximum conditional mutual information value is determined as the time delay estimate for that channel. The time delay estimates of all channels are then combined to obtain the time delay matrix.

2. The method for estimating channel delay in the boiler-turbine coordination system of a thermal power unit according to claim 1, characterized in that, The input variables include coal feed rate command, water flow rate command, and turbine control valve opening command; the output variables include main steam pressure, steam-water separator outlet enthalpy, and unit power.

3. The method for estimating channel delay in the boiler-turbine coordination system of a thermal power unit according to claim 1, characterized in that, The step of performing zero-phase filtering and differential processing on the time-series samples of each variable to obtain stable differential signal samples includes: Symmetric mean filtering is applied to the time series samples of each variable to obtain the filtered time series samples of each variable. The filtered time series samples of each variable are simultaneously differentially processed to obtain stable differential signal samples.

4. The method for estimating channel delay in the boiler-turbine coordination system of a thermal power unit according to claim 1, characterized in that, The step of resampling the differential signal samples to obtain resampled signal samples includes: The differential signal samples are mapped to Euclidean space and normalized. The K-nearest neighbor algorithm is used to estimate the local probability density of each sample point. Calculate the inverse probability weight of each sample point based on the local probability density of each sample point; According to the distribution of the inverse probability weights, the differential signal samples are subjected to weighted random sampling without replacement to obtain resampled signal samples.

5. The method for estimating channel delay in the boiler-turbine coordination system of a thermal power unit according to claim 4, characterized in that, The formula for calculating the local probability density is as follows: ； in, For sample points The local probability density; ; This represents the total sample size. The specified number of neighboring nodes; for The volume of a unit hypersphere in 3D space; For sample points Its first The Euclidean distance between the nearest neighbors.

6. The method for estimating channel delay in the boiler-turbine coordination system of a thermal power unit according to claim 4, characterized in that, The formula for calculating the inverse probability weight is as follows: ； in, For sample points Inverse probability weights; For sample points The local probability density; Sample points The local probability density; This represents the total sample size.

7. The method for estimating channel delay in the boiler-turbine coordination system of a thermal power unit according to claim 1, characterized in that, The formula for calculating the conditional mutual information value is as follows: ； in, for Output variables at the sampling time; for Output variables at the sampling time; The time lag is the time delay. for Input variables at the sampling time; for Input variables at the sampling time; For variables The joint probability density; For a given under conditions The conditional probability density; For a given under conditions The conditional probability density; Indicates that given historical information Under the conditions, and The conditional mutual information values ​​between them.

8. The method for estimating channel delay in the boiler-turbine coordination system of a thermal power unit according to claim 1, characterized in that, After synthesizing the delay estimates from all channels to obtain the delay matrix, the method further includes: Import the time delay matrix into the model predictive controller; The model predictive controller performs time axis registration on its internal predictive models according to the time delay matrix, calculates control sequences based on the time axis registered predictive models, and sends them to each actuator.

9. A device for estimating channel delay in the boiler-turbine coordination system of a thermal power unit, characterized in that, include: The data acquisition module is used to collect time-series samples of the input and output variables of the boiler-turbine coordination system during the operation of the thermal power unit. The input variable is the instruction issued by the actuator, and the output variable is the sensor measurement value; The preprocessing module is used to perform zero-phase filtering and differential processing on the time-series samples of each variable to obtain stable differential signal samples. The resampling module is used to resample the differential signal samples to obtain resampled signal samples; The calculation module is used to calculate the conditional mutual information value between each input variable and each output variable under different lag orders based on the resampled signal samples; The delay determination module is used to determine the delay estimate of each channel between each input variable and each output variable, starting from zero lag, when the conditional mutual information value first appears significantly higher than zero and is a local maximum. The delay estimate of all channels is then combined to obtain the delay matrix.

10. A computer device, characterized in that, The computer device includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method for estimating channel delay of the boiler-turbine coordination system of a thermal power unit as described in any one of claims 1-8.

Images