Hearth negative pressure closed-loop control system health degree monitoring method based on machine learning

Abstract

The embodiment of the invention discloses a hearth negative pressure closed-loop control system health degree monitoring method based on machine learning, and relates to the technical field of thermal power plant thermal process automatic control, and the method comprises the steps: obtaining the real-time operation data of a hearth negative pressure closed-loop control system, inputting the real-time operation data into a first neural network model, obtaining a feedforward parameter standard value, and calculating the health degree of the hearth negative pressure closed-loop control system; determining an operation state according to the real-time operation data; calculating to obtain a hearth negative pressure adjustment deviation parameter corresponding to the operation state; calculating to obtain a moving blade deviation parameter based on the real-time operation data; weighting to obtain a fusion parameter; according to the method, the second-level quantitative evaluation of the health degree of the chamber negative pressure closed-loop control system, the hour-level early warning of early faults and the quantitative decision of income are realized, the operation safety and the intelligent level of operation and maintenance of a thermal power generating unit are effectively improved, and a scientific solution is provided for state monitoring and optimal maintenance of the system.

Description

Technical Field

[0001] This application relates to the field of automatic control technology for thermal processes in thermal power plants, and in particular to a method for monitoring the health of a furnace negative pressure closed-loop control system based on machine learning. Background Technology

[0002] To ensure the safe operation of boilers, furnace negative pressure is a key controlled variable, making boiler control performance crucial. Currently, relevant technologies generally rely on a closed-loop control system for furnace negative pressure monitoring within a distributed control system (DCS) using PI (or PID) cascade + feedforward control strategies.

[0003] Specifically, the lack of quantitative evaluation methods for the furnace negative pressure closed-loop control system in related technologies makes it impossible to provide early warnings of performance degradation. Furthermore, the fixed maintenance cycle of current furnace negative pressure closed-loop control systems easily leads to over-dimensionality or untimely maintenance. On the other hand, related technologies do not consider load variations in thermal power units. The furnace negative pressure closed-loop control system does not employ differentiated evaluation benchmarks under different load segments and varying load rates, making it impossible to accurately and adaptively evaluate the performance of the furnace negative pressure closed-loop control system based on operating conditions. In addition, related technologies focus on only a single indicator, failing to simultaneously consider feedforward quality, steady-state / dynamic / actuator multi-dimensional characteristics, or introduce machine learning models to adaptively generate "personalized" standard values, resulting in crude evaluation results and a high false alarm rate.

[0004] Therefore, developing a machine learning-based method for monitoring the health of a furnace negative pressure closed-loop control system is of great significance for improving the operational safety and intelligent operation and maintenance level of thermal power units. Summary of the Invention

[0005] This application provides a machine learning-based method for monitoring the health of a furnace negative pressure closed-loop control system. This method addresses the lack of quantitative assessment methods for the health status of traditional furnace negative pressure control systems, which leads to the inability to provide early warnings of performance degradation. The technical solution is as follows: In a first aspect, embodiments of this application provide a method for monitoring the health of a furnace negative pressure closed-loop control system based on machine learning, including: Acquire real-time operating data of the furnace negative pressure closed-loop control system; The real-time running data is input into the trained first neural network model to obtain the standard value of the feedforward parameter. The feedforward adjustment deviation parameter is calculated based on the actual value of the feedforward parameter in the real-time running data and the standard value of the feedforward parameter. The operating status of the furnace negative pressure closed-loop control system is determined based on the real-time operating data, and the furnace negative pressure adjustment deviation parameter corresponding to the operating status is calculated by combining the real-time operating data. The blade deviation parameter is calculated based on the real-time operating data. The fusion parameter is obtained by weighting the feedforward adjustment deviation parameter, the furnace negative pressure adjustment deviation parameter, and the moving blade deviation parameter; The health monitoring results of the furnace negative pressure closed-loop control system are obtained based on the fusion parameters.

[0006] In one alternative of the first aspect, the real-time operating data includes the collected active power of the thermal power unit, furnace negative pressure, blower blade opening and induced draft blade opening. The step of inputting the real-time running data into the trained first neural network model to obtain the standard values ​​of the feedforward parameters includes: The active power and the blower blade opening are input into the first neural network model so that the first neural network model can calculate the predicted value of the blower blade opening based on the learned mapping relationship between the input and output quantities, and the predicted value of the blower blade opening is used as the standard value of the feedforward parameter. The process of using the induced draft blade opening from the real-time operating data as the actual value of the feedforward parameter to calculate the feedforward adjustment deviation parameter applies the following formula: ; in, It is the feedforward adjustment deviation parameter; It is the actual value of the feedforward parameter; It is the predicted value based on the opening degree of the induced draft blades; It is the proportionality coefficient.

[0007] In one alternative embodiment of the first aspect, the training process of the first neural network model includes: Acquire historical operating data of the furnace negative pressure closed-loop control system, including historical active power, historical forced draft fan blade opening, and historical induced draft fan blade opening; The historical active power and historical supply fan blade opening are used as sample inputs, and the historical induced draft fan blade opening is used as the corresponding sample label to construct a sample set. The sample set is input into the first neural network model, so that the first neural network model outputs the corresponding predicted value based on the sample input; A loss function is constructed based on the difference between the predicted value and the sample label; The convergence of the first neural network model is determined based on the value of the loss function, and the trained first neural network model is obtained based on the model parameters at the time of convergence.

[0008] In one alternative embodiment of the first aspect, determining the operating status of the furnace negative pressure closed-loop control system based on the real-time operating data includes: Active power is collected within a preset time window based on a preset sampling frequency, and the difference between the maximum and minimum values ​​of active power within the preset time window is calculated. If the difference is less than the active power fluctuation threshold, then the furnace negative pressure closed-loop control system is determined to be in a steady state. Otherwise, it is determined that the furnace negative pressure closed-loop control system is in a dynamic state.

[0009] In one alternative embodiment of the first aspect, when the furnace negative pressure closed-loop control system is in a steady state, the calculation of the furnace negative pressure adjustment deviation parameter corresponding to the operating state based on the real-time operating data includes: Obtain the furnace negative pressure at each sampling moment within the preset time window, calculate the furnace negative pressure difference between the furnace negative pressure at each sampling moment and the furnace negative pressure setpoint, and calculate the actual value of steady-state adjustment deviation based on the furnace negative pressure difference, applying the formula: ; The active power within the preset time window is input into the trained second neural network model, so that the second neural network model can calculate the corresponding furnace negative pressure standard deviation based on the learned mapping relationship between the input and output quantities. The steady-state regulation deviation standard value is calculated based on the standard deviation of the furnace negative pressure, using the following formula: ; The furnace negative pressure regulation deviation parameter when the operating state is steady is calculated based on the actual value of the steady-state regulation deviation and the standard value of the steady-state regulation deviation, using the following formula: ; in, This is the actual value of the steady-state adjustment deviation; It is the actual value of the furnace negative pressure at the i-th data acquisition moment; This is the furnace negative pressure setpoint; N is the number of data collection points within the time window. It is the standard value of steady-state adjustment deviation; It is the tolerance coefficient; It is the standard deviation of the furnace negative pressure output by the second neural network model; It is the furnace negative pressure regulation deviation parameter when the operating state is steady; It is the proportionality coefficient.

[0010] In one alternative embodiment of the first aspect, when the furnace negative pressure closed-loop control system is in a dynamic state, the calculation of the furnace negative pressure adjustment deviation parameter corresponding to the operating state based on the real-time operating data includes: The actual value of the dynamic adjustment deviation is calculated based on the difference between the furnace negative pressure and the furnace negative pressure setpoint, using the following formula: ; The load correction coefficient is determined based on the ratio of the current load to the rated load of the furnace negative pressure closed-loop control system, and the preset variable load rate correction coefficient is obtained. The standard value of dynamic adjustment deviation is calculated based on the load correction coefficient, the variable load rate correction coefficient, and the furnace negative pressure setpoint, using the following formula: ; Based on the actual value of the dynamic adjustment deviation and the standard value of the dynamic adjustment deviation, the furnace negative pressure adjustment deviation parameter when the operating state is dynamic is calculated, and the formula is applied: ; in, It is the actual value of the dynamic adjustment deviation; It is the negative pressure in the furnace when the operating state is dynamic; This is the furnace negative pressure setting value; It is the load correction factor; It is the variable load rate correction factor; It is a dynamic adjustment deviation standard value; It is the furnace negative pressure adjustment deviation parameter when the operating state is dynamic; It is the proportionality coefficient.

[0011] In one alternative of the first aspect, the real-time operating data includes the target opening degree of the induced draft blade and the actual opening degree of the induced draft blade from the blade control system. The calculation of the blade deviation parameters based on the real-time operating data includes: The actual deviation of the induced draft blades between the target opening and the actual opening is calculated using the following formula: ; The blade deviation parameter is calculated based on the actual deviation of the induced draft blade and the preset standard deviation of the induced draft blade, using the following formula: ; in, This is the actual deviation of the induced draft fan blades; It is the target opening degree of the induced draft blades; This refers to the actual opening of the induced draft fan blades; It is the blade deviation parameter; It is the proportionality coefficient.

[0012] Secondly, embodiments of this application also provide a machine learning-based furnace negative pressure closed-loop control system health monitoring device, comprising: The data acquisition unit is used to acquire real-time operating data of the furnace negative pressure closed-loop control system; The parameter calculation unit is used to input the real-time running data into the trained first neural network model to obtain the standard value of the feedforward parameter, and to calculate the feedforward adjustment deviation parameter based on the actual value of the feedforward parameter in the real-time running data and the standard value of the feedforward parameter. The parameter calculation unit is also used to determine the operating status of the furnace negative pressure closed-loop control system based on the real-time operating data, and calculate the furnace negative pressure adjustment deviation parameter corresponding to the operating status in combination with the real-time operating data. The parameter calculation unit is also used to calculate the blade deviation parameter based on the real-time operating data; The parameter calculation unit is also used to obtain a fusion parameter based on the weighted sum of the feedforward adjustment deviation parameter, the furnace negative pressure adjustment deviation parameter, and the moving blade deviation parameter; The health monitoring unit is used to obtain the health monitoring results of the furnace negative pressure closed-loop control system based on the fusion parameters.

[0013] Thirdly, embodiments of this application also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method provided by the first aspect or any implementation thereof of the embodiments of this application.

[0014] Fourthly, this application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method provided by the first aspect of the embodiments of this application or any implementation thereof.

[0015] The beneficial effects of the technical solutions provided in some embodiments of this application include at least the following: This application provides a machine learning-based method for monitoring the health of a furnace negative pressure closed-loop control system. By using a neural network model to generate standard values ​​for feedforward parameters, it avoids the shortcomings of related technologies that rely solely on fixed standard values ​​and cannot adaptively monitor health based on real-time operating data. Furthermore, this application considers different operating states and calculates corresponding furnace negative pressure adjustment deviation parameters based on those states, overcoming the shortcomings of related technologies that do not consider steady-state / dynamic differences. This application integrates feedforward adjustment deviation parameters, furnace negative pressure adjustment deviation parameters considering operating states, and blade deviation parameters, enabling multi-dimensional health monitoring. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in this application or related technologies, the drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating a machine learning-based method for monitoring the health of a closed-loop furnace negative pressure control system, as provided in an embodiment of this application. Figure 2 This is a schematic diagram of the structure of a machine learning-based furnace negative pressure closed-loop control system health monitoring device provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] The terms "comprising" and "having," and any variations thereof, in the specification, claims, and accompanying drawings of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or modules is not limited to the steps or modules listed, but may optionally include steps or modules not listed, or may optionally include other steps or modules inherent to such process, method, product, or apparatus.

[0020] It should be noted that the terms "first" and "second" used in this application are merely to distinguish similar objects and do not represent a specific ordering of the objects. It is understood that "first" and "second" can be interchanged in a specific order or sequence where permitted. It should be understood that the objects distinguished by "first" and "second" can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in an order other than those described or illustrated herein.

[0021] The present application will now be described in detail with reference to specific embodiments.

[0022] Next, combine Figure 1 This application introduces a machine learning-based method for monitoring the health of a furnace negative pressure closed-loop control system, as provided in its embodiments. For details, please refer to... Figure 1 , Figure 1 This illustration shows a flowchart of a machine learning-based closed-loop control system health monitoring method for furnace negative pressure provided in an embodiment of this application. Figure 1 As shown, the method includes the following steps: S101, acquire real-time operating data of the furnace negative pressure closed-loop control system; S102, the real-time running data is input into the trained first neural network model to obtain the standard value of the feedforward parameter, and the feedforward adjustment deviation parameter is calculated based on the actual value of the feedforward parameter in the real-time running data and the standard value of the feedforward parameter. S103, determine the operating status of the furnace negative pressure closed-loop control system based on the real-time operating data, and calculate the furnace negative pressure adjustment deviation parameter corresponding to the operating status based on the real-time operating data. S104, calculate the blade deviation parameter based on the real-time operating data; S105, the fusion parameter is obtained by weighting the feedforward adjustment deviation parameter, the furnace negative pressure adjustment deviation parameter, and the moving blade deviation parameter; S106, The health monitoring results of the furnace negative pressure closed-loop control system are obtained based on the fusion parameters.

[0023] This application enables second-level quantitative assessment of the health status of the chamber negative pressure closed-loop control system, hour-level early warning of early faults, and quantitative decision-making on benefits, effectively improving the operational safety and intelligent level of thermal power units and providing a scientific solution for condition monitoring and optimized maintenance of thermal power unit control systems.

[0024] In some embodiments, in S101, real-time operating data can be collected by sensors in the furnace negative pressure closed-loop control system. The real-time operating data includes, but is not limited to, the active power of the thermal power unit, the furnace negative pressure, the opening degree of the forced draft fan blades, and the opening degree of the induced draft fan blades.

[0025] Understandably, the furnace negative pressure closed-loop control system is a PID control system. The PID controller generates control signals for the actuators. Input data includes the set furnace negative pressure setpoint and the feedforward input of the blower motor blade opening. The output value is the actual furnace negative pressure. The controller calculates the control signal based on the deviation between the setpoint and the actual furnace negative pressure according to a proportional (P), integral (I), and derivative (D) control law to adjust the actuator's action. The actuator, including the induced draft fan motor blades, receives the control signal from the controller and changes its own opening to control the furnace negative pressure. In the feedforward loop, the pre-adjustment amount of the induced draft fan motor blades is calculated based on the opening deviation of the blower motor blades and transmitted to the controller to preemptively counteract the interference of blower fan changes on the furnace negative pressure. The furnace negative pressure can be monitored in real time by a negative pressure sensor and fed back to the PID controller to form a feedback loop for closed-loop control.

[0026] Understandably, the forced draft fan is located at the boiler's "air inlet" and is responsible for supplying the air necessary for combustion into the furnace. The induced draft fan is located at the boiler's "flue gas outlet" and is responsible for drawing the high-temperature flue gas generated after combustion out of the furnace, causing it to flow along the flue, and finally being discharged into the atmosphere through the chimney.

[0027] In some embodiments, the training process of the first neural network model in S102 includes: S201, Obtain historical operating data of the furnace negative pressure closed-loop control system, including historical active power, historical blower blade opening and historical induced draft blade opening; S202, Preprocessing the data, including: S2021, Outlier Removal: To suppress the impact of extreme outliers on modeling, the following approach is adopted. The method involves data cleaning of all data in the input model.

[0028] For example, the following uses active power data as an example to illustrate how to clean the power sequence: First, the active power data at corresponding times are arranged according to the wind speed inside the furnace, and the power data is divided into multiple data segments; then, the mean and standard deviation of the power in each segment are calculated: ; in, This represents the mean of each data segment after dividing the data into segments according to wind speed ranges. It represents the standard deviation of the data segments. The larger the standard deviation, the more drastic the fluctuation of the data. Indicates the number of power data points contained within a data segment; This represents the i-th power data within the interval.

[0029] like Then the The corresponding data, such as the opening degree of the moving blades and the opening degree of the induced draft fan, were identified as outliers and removed. Finally, for data gaps appearing in the active power data after outlier removal, linear interpolation was used to fill them in. Linear interpolation constructs a linear function using only two valid data points before and after the gap, enabling rapid output of the filling results. This perfectly matches the real-time requirements of the furnace negative pressure control system for "second-level quantitative evaluation," and the error is more controllable.

[0030] S2022, normalization is performed. The cleaned data is uniformly processed using Min-Max Normalization, mapping the data to the [0, 1] interval so that the model can learn better. The formula is as follows: ; in, It is normalized data; The original value, These are the minimum and maximum values ​​of the variable in the training set.

[0031] S203, using historical active power and historical supply fan blade opening as sample inputs, and using the historical induced draft fan blade opening as the corresponding sample label, a sample set is constructed. The dataset can be evenly divided into training, validation, and test sets in a 3:1:1 ratio to ensure that they are equally distributed and contain different working conditions.

[0032] S204, the sample set is input into the first neural network model so that the first neural network model outputs the corresponding predicted value based on the sample input.

[0033] The first neural network model in this embodiment specifically adopts a Long Short-Term Memory (LSTM) network model, whose structure includes an input layer, an LSTM layer, and an output layer. The input layer receives the active power of the thermal power unit and the turbine blade opening as input features. The LSTM layer is used to capture long-term dependencies in the time series, and the output layer generates the turbine blade opening as the model's output. The LSTM model structure and formula are as follows: 1) Input layer Input feature is power and the opening of the blower blades The input vector at time t is represented as: ; These are the normalized active power and the fan blade opening. It represents the 2-dimensional real space.

[0034] 2) LSTM layer LSTM layers handle time-series dependencies through gating mechanisms, with the core variable being cell state. (Long-term memory) and hidden state (Short-term output) update. Assume the LSTM layer has d hidden units, and the parameter formulas are as follows: Forget Gate: Determines which information to discard from the cell state.

[0035] ; in, It is the forget gate weight matrix. It is the forgetting gate bias. It is the sigmoid activation function (output range 0-1, 0 means complete forgetting, 1 means complete retention). Indicates the hidden state at the previous moment. With current input The concatenated vector.

[0036] Input Gate: Determines the proportion of new information that should be stored.

[0037] ; ; in, It is the input gate related weight matrix. It corresponds to the bias. It is the hyperbolic tangent activation function (output range -1 to 1). This is the candidate cell state.

[0038] Cell state update: integrating historical memory with new information.

[0039] ; in, This indicates element-wise multiplication (multiplying elements one by one).

[0040] Output Gate: Generates the hidden state at the current moment.

[0041] ; ; in, It is the output gate weight matrix. It is the output gate bias vector. It is the output of the LSTM layer, i.e., the hidden state at time t.

[0042] 3) Output layer Map the hidden states of the LSTM layer to predicted values ​​of the induced draft blade opening. : ; in, It is the output layer weight matrix. This is the output layer bias. All the weight matrices mentioned above are optimized through training. and bias vector .

[0043] S205, Construct a loss function based on the difference between the predicted value and the sample label; S206, determine whether the first neural network model has converged based on the value of the loss function, and obtain the trained first neural network model based on the model parameters at the time of convergence.

[0044] For example, with the goal of minimizing the deviation between the predicted and measured values ​​of the induced draft blade opening, the model parameters—including the LSTM layer gating weights (forget gate)—are iteratively adjusted using the mean square error (MSE) loss function. Input gate (etc.) and output layer mapping weights .

[0045] The trained model can accurately predict the induced draft fan blade opening based on active power and supply fan blade opening, providing a basis for subsequent health index calculations.

[0046] In some embodiments, S102 specifically includes: The real-time running data can be input into the trained first neural network model to obtain the standard values ​​of the feedforward parameters, and then S1021 can be executed: S1021, the data cleaning and data normalization operations in S202 can be used to preprocess the input data. Please refer to the descriptions of S2021 and S2022 for details, which will not be repeated here.

[0047] S1022, the active power and the blower blade opening are input into the first neural network model so that the first neural network model calculates the predicted value of the blower blade opening based on the learned mapping relationship between the input and output quantities, and the predicted value of the blower blade opening is used as the standard value of the feedforward parameter. Furthermore, the feedforward adjustment deviation parameter is calculated: S1023, the process of using the induced draft blade opening in the real-time operating data as the actual value of the feedforward parameter to calculate the feedforward adjustment deviation parameter applies the following formula: ; in, It is the feedforward adjustment deviation parameter; It is the actual value of the feedforward parameter; It is the predicted value based on the opening degree of the induced draft blades; It is a proportional coefficient used to adjust the degree of influence of the deviation on the feedforward adjustment deviation parameter, ensuring that the feedforward adjustment deviation parameter can reasonably reflect the health of the system.

[0048] Here, it is assumed that the feedforward adjustment deviation parameter is linearly related to the deviation, that is, when and When the deviation between them increases, the feedforward adjustment deviation parameter decreases, indicating that there may be a deviation or fault in the system; when and When they are perfectly synchronized, the feedforward adjustment deviation parameter is 100, indicating that the system is operating in its optimal state. Generally, the proportional gain... The value satisfies the following conditions: ; in, This is the maximum allowable deviation value of the system. Through appropriate selection... This ensures that the feedforward adjustment deviation parameter is between 0 and 100.

[0049] In some embodiments, S103, determining the operating status of the furnace negative pressure closed-loop control system based on the real-time operating data includes: S1031, Collect active power within a preset time window based on a preset collection frequency, and calculate the difference between the maximum and minimum values ​​of active power within the preset time window.

[0050] Specifically, a 5-minute preset time window can be set, and active power can be collected once per minute within the time window to determine the fluctuation range of active power within the preset time window, that is, to calculate the difference between the maximum and minimum active power values ​​within the preset time window: ; in, This is the setpoint value of the active power of the thermal power unit at the i-th moment within the preset time window. This represents the maximum active power within a preset time window. This is the minimum active power within a preset time window.

[0051] S1032, compare the difference between the maximum and minimum active power values ​​within the preset time window and the active power fluctuation threshold; If the difference is less than the active power fluctuation threshold: ; This confirms that the furnace negative pressure closed-loop control system is in a steady state. Otherwise, it is determined that the furnace negative pressure closed-loop control system is in a dynamic state.

[0052] in, The active power fluctuation threshold can be set according to the unit characteristics. , which represents the maximum allowable range of fluctuations when the system is stable.

[0053] It should be noted that, in addition to the power setting, other key parameters (such as temperature and pressure) can also be considered to determine the operating status of the system and whether the system remains stable within the allowable fluctuation range.

[0054] In some embodiments, when the furnace negative pressure closed-loop control system is in a steady state, in S103, the furnace negative pressure adjustment deviation parameter corresponding to the operating state is calculated by combining the real-time operating data, including: Obtain the furnace negative pressure at each sampling moment within the preset time window, calculate the furnace negative pressure difference between the furnace negative pressure at each sampling moment and the furnace negative pressure setpoint, and calculate the actual value of steady-state adjustment deviation based on the furnace negative pressure difference, applying the formula: ; The active power within the preset time window is input into the trained second neural network model, so that the second neural network model can calculate the corresponding furnace negative pressure standard deviation based on the learned mapping relationship between the input and output quantities. The steady-state regulation deviation standard value is calculated based on the standard deviation of the furnace negative pressure, using the following formula: ; The furnace negative pressure regulation deviation parameter when the operating state is steady is calculated based on the actual value of the steady-state regulation deviation and the standard value of the steady-state regulation deviation, using the following formula: ; in, This is the actual value of the steady-state adjustment deviation; It is the actual value of the furnace negative pressure at the i-th data acquisition moment; This is the furnace negative pressure setpoint; N is the number of data collection points within the time window. It is the standard value of steady-state adjustment deviation; This is the tolerance factor, which can be set to 1.3; It is the standard deviation of the furnace negative pressure output by the second neural network model; It is the furnace negative pressure regulation deviation parameter when the operating state is steady; It is a proportionality coefficient. For high-precision systems, it can be set to 8, which is used to adjust the degree of influence of the furnace negative pressure deviation on the furnace negative pressure regulation deviation parameter.

[0055] It should be noted that the model structure and training process of the second neural network model are the same as those of the first neural network model in S102, except that the input data and output data are different. The second neural network model takes the active power and furnace negative pressure under steady state as input data and the output data is the predicted standard deviation of the furnace negative pressure.

[0056] In some embodiments, when the furnace negative pressure closed-loop control system is in a dynamic state, in S103, the furnace negative pressure adjustment deviation parameter corresponding to the operating state is calculated by combining the real-time operating data, including: The actual value of the dynamic adjustment deviation is calculated based on the difference between the furnace negative pressure and the furnace negative pressure setpoint, using the following formula: ; The load correction coefficient is determined based on the ratio of the current load to the rated load of the furnace negative pressure closed-loop control system, and the preset variable load rate correction coefficient is obtained. The standard value of dynamic adjustment deviation is calculated based on the load correction coefficient, the variable load rate correction coefficient, and the furnace negative pressure setpoint, using the following formula: ; Based on the actual value of the dynamic adjustment deviation and the standard value of the dynamic adjustment deviation, the furnace negative pressure adjustment deviation parameter when the operating state is dynamic is calculated, and the formula is applied: ; in, It is the actual value of the dynamic adjustment deviation; It is the negative pressure in the furnace when the operating state is dynamic; This is the furnace negative pressure setting value, for example, set to 200Pa; It is the load correction factor; It is the variable load rate correction factor; It is a dynamic adjustment deviation standard value; It is the furnace negative pressure adjustment deviation parameter when the operating state is dynamic; It is a proportional coefficient, which is used in conjunction with the typical scenario of dynamic adjustment deviation in thermal power units. Based on the default value of the maximum allowable deviation of the system, it can be set to 20.

[0057] In some embodiments, load correction factor The calculation formula is as follows: ; Where Load is the ratio of the current load to the rated load, specifically the percentage of the current load to the rated load. When the ratio is above 50%, the load correction factor is 1; when the ratio is 20% or below, the load correction factor is 1.5; when the ratio is between 20% and 50%, the load correction factor is calculated using a weighted average.

[0058] When the load rate change is below 1.5%, the load rate change correction factor is 1; for every 0.1% increase in the load rate change, the correction factor increases by 0.1. The calculation formula is as follows: ; Rate is the variable load rate (expressed as a percentage).

[0059] In some embodiments, the real-time operating data includes the target opening degree and actual opening degree of the induced draft blades of the blade control system, and S104 includes: S1041, Calculate the actual deviation of the induced draft blades between the target opening and the actual opening of the induced draft blades, using the formula: ; S1042, the blade deviation parameter is calculated based on the actual deviation of the induced draft blade and the preset standard deviation of the induced draft blade, using the following formula: ; in, This is the actual deviation of the induced draft fan blades; It is the target opening degree of the induced draft blades; This refers to the actual opening of the induced draft fan blades; It is the blade deviation parameter; It is a proportionality coefficient used to adjust the degree of influence of deviation on health; the preset standard deviation of the induced draft fan blades is a fixed reference value used to assess whether the actual deviation is within the allowable range, and the default standard value is 1.

[0060] In some embodiments, the scaling factor The choice needs to be determined based on the actual requirements of the system and the allowable deviation range. Generally speaking, The value should satisfy the following conditions: ; in, This is the maximum permissible blade deviation of the system. Through proper selection... This ensures that the health of the moving blade deviation is between 0 and 100.

[0061] Further, in step S105, a fusion parameter is obtained by weighting the feedforward adjustment deviation parameter, the furnace negative pressure adjustment deviation parameter, and the moving blade deviation parameter. Apply the formula: ; in, It is the weight of the feedforward adjustment deviation parameter; These are the weights of the blade deviation parameters; It is the weight of the furnace negative pressure regulation deviation parameter when the operating state is steady state. It is the weight of the furnace negative pressure adjustment deviation parameter when the operation is dynamic.

[0062] Understandably, when the system is in a steady state, the furnace negative pressure regulation deviation parameter when the system is in a dynamic state is not included in the weighted calculation; when the system is in a dynamic state, the furnace negative pressure regulation deviation parameter when the system is in a steady state is not included in the weighted calculation.

[0063] Optionally, the weight values ​​of the feedforward adjustment deviation parameter, the moving blade deviation parameter, and the furnace negative pressure adjustment deviation parameter can be set based on empirical values.

[0064] Optionally, when determining the weight of each indicator on the system's health, in order to enhance the scientific rigor and accuracy of the overall health assessment, a neural network model can be used to optimize the weight setting, making the calculation of the overall health more reasonable and reliable.

[0065] Furthermore, based on the fusion parameters, the health monitoring results of the furnace negative pressure closed-loop control system are obtained, and S106 includes: The values ​​of the fusion parameters can be statistically analyzed over time, and trend curves of the fusion parameters changing over time can be plotted. The health status changes of the furnace negative pressure closed-loop control system can be observed intuitively through the curves. If the curve remains stable at a high value for a long period (e.g., above 80 points), it indicates that the furnace negative pressure closed-loop control system is operating well and the various subsystems are working together stably. If the curve shows a significant drop (e.g., from 90 points to 60 points in a short period of time), it suggests that there may be an abnormality in the furnace negative pressure closed-loop control system, which needs to be investigated in a timely manner.

[0066] You can also set a warning threshold for the fusion parameters (e.g., 70 points). When the value of the fusion parameter calculated in real time is lower than this threshold, an audible and visual alarm or a system pop-up notification will be triggered to notify the operators that the health status of the furnace negative pressure closed-loop control system has deteriorated and that relevant equipment (such as induced draft fan, forced draft fan, negative pressure control system, etc.) should be checked immediately to prevent accidents (such as positive pressure combustion in the furnace, excessive negative pressure leading to air leakage, etc.).

[0067] The following are apparatus embodiments of this application, which can be used to execute the method embodiments of this application. For details not disclosed in the apparatus embodiments of this application, please refer to the method embodiments of this application.

[0068] Please see below. Figure 2 This is a schematic diagram of a machine learning-based furnace negative pressure closed-loop control system health monitoring device provided in an exemplary embodiment of this application. The device includes: The data acquisition unit is used to acquire real-time operating data of the furnace negative pressure closed-loop control system; The parameter calculation unit is used to input the real-time running data into the trained first neural network model to obtain the standard value of the feedforward parameter, and to calculate the feedforward adjustment deviation parameter based on the actual value of the feedforward parameter in the real-time running data and the standard value of the feedforward parameter. The parameter calculation unit is also used to determine the operating status of the furnace negative pressure closed-loop control system based on the real-time operating data, and calculate the furnace negative pressure adjustment deviation parameter corresponding to the operating status in combination with the real-time operating data. The parameter calculation unit is also used to calculate the blade deviation parameter based on the real-time operating data; The parameter calculation unit is also used to obtain a fusion parameter based on the weighted sum of the feedforward adjustment deviation parameter, the furnace negative pressure adjustment deviation parameter, and the moving blade deviation parameter; The health monitoring unit is used to obtain the health monitoring results of the furnace negative pressure closed-loop control system based on the fusion parameters.

[0069] It should be noted that the device provided in the above embodiments, when executing the machine learning-based furnace negative pressure closed-loop control system health monitoring method, is only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the equipment can be divided into different functional modules to complete all or part of the functions described above. Furthermore, the device provided in the above embodiments and the machine learning-based furnace negative pressure closed-loop control system health monitoring method embodiments belong to the same concept, and their implementation process is detailed in the method embodiments, which will not be repeated here.

[0070] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of any of the methods described above.

[0071] Please see Figure 3 This is a structural block diagram of an electronic device provided in an embodiment of this application.

[0072] like Figure 3 As shown, the electronic device 300 includes a processor 301 and a memory 302.

[0073] In this embodiment, the processor 301 is the control center of the computer system, and can be a processor of a physical machine or a processor of a virtual machine. The processor 301 may include one or more processing cores, such as a 4-core processor or an 8-core processor. The processor 301 can be implemented using at least one hardware form selected from DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array).

[0074] Processor 301 may also include a main processor and a coprocessor. The main processor is a processor used to process data in the wake-up state, also known as a CPU (Central Processing Unit). The coprocessor is a low-power processor used to process data in the standby state.

[0075] Memory 302 may include one or more computer-readable storage media, which may be non-transitory. Memory 302 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In some embodiments of this application, the non-transitory computer-readable storage media in memory 302 is used to store at least one instruction, which is executed by processor 301 to implement the method in the embodiments of this application.

[0076] In some embodiments, the electronic device 300 further includes a peripheral device interface 303 and at least one peripheral device 304. The processor 301, memory 302, and peripheral device interface 303 can be connected via a bus or signal line. Each peripheral device 304 can be connected to the peripheral device interface 303 via a bus, signal line, or circuit board. Specifically, the peripheral device 304 includes: a display screen, a camera, and audio circuitry. The peripheral device interface 303 can be used to connect at least one I / O (Input / Output) related peripheral device to the processor 301 and memory 302.

[0077] In some embodiments of this application, the processor 301, memory 302, and peripheral device interface 303 are integrated on the same chip or circuit board; in other embodiments of this application, any one or two of the processor 301, memory 302, and peripheral device interface 303 can be implemented on separate chips or circuit boards. This application does not specifically limit the implementation in this regard.

[0078] The electronic device structural block diagram shown in the embodiments of this application does not constitute a limitation on the electronic device 300. The electronic device 300 may include more or fewer components than shown, or combine certain components, or use different component arrangements.

[0079] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the methods in any of the foregoing embodiments. The computer-readable storage medium may include, but is not limited to, any type of disk, including floppy disks, optical disks, DVDs, CD-ROMs, microdrives, as well as magneto-optical disks, ROMs, RAMs, EPROMs, EEPROMs, DRAMs, VRAMs, flash memory devices, magnetic cards or optical cards, nanosystems (including molecular memory ICs), or any type of medium or device suitable for storing instructions and / or data.

[0080] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of software products. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for monitoring the health of a furnace negative pressure closed-loop control system based on machine learning, characterized in that, include: Acquire real-time operating data of the furnace negative pressure closed-loop control system; The real-time running data is input into the trained first neural network model to obtain the standard value of the feedforward parameter. The feedforward adjustment deviation parameter is calculated based on the actual value of the feedforward parameter in the real-time running data and the standard value of the feedforward parameter. The operating status of the furnace negative pressure closed-loop control system is determined based on the real-time operating data, and the furnace negative pressure adjustment deviation parameter corresponding to the operating status is calculated by combining the real-time operating data. The blade deviation parameter is calculated based on the real-time operating data. The fusion parameter is obtained by weighting the feedforward adjustment deviation parameter, the furnace negative pressure adjustment deviation parameter, and the moving blade deviation parameter; The health monitoring results of the furnace negative pressure closed-loop control system are obtained based on the fusion parameters.

2. The method according to claim 1, characterized in that, The real-time operating data includes the collected active power of the thermal power unit, furnace negative pressure, blower blade opening, and induced draft blade opening. The step of inputting the real-time running data into the trained first neural network model to obtain the standard values ​​of the feedforward parameters includes: The active power and the blower blade opening are input into the first neural network model so that the first neural network model can calculate the predicted value of the blower blade opening based on the learned mapping relationship between the input and output quantities, and the predicted value of the blower blade opening is used as the standard value of the feedforward parameter. The process of using the induced draft blade opening from the real-time operating data as the actual value of the feedforward parameter to calculate the feedforward adjustment deviation parameter applies the following formula: ； in, It is the feedforward adjustment deviation parameter; It is the actual value of the feedforward parameter; It is the predicted value based on the opening degree of the induced draft blades; It is the proportionality coefficient.

3. The method according to claim 2, characterized in that, The training process of the first neural network model includes: Acquire historical operating data of the furnace negative pressure closed-loop control system, including historical active power, historical forced draft fan blade opening, and historical induced draft fan blade opening; The historical active power and historical supply fan blade opening are used as sample inputs, and the historical induced draft fan blade opening is used as the corresponding sample label to construct a sample set. The sample set is input into the first neural network model, so that the first neural network model outputs the corresponding predicted value based on the sample input; A loss function is constructed based on the difference between the predicted value and the sample label; The convergence of the first neural network model is determined based on the value of the loss function, and the trained first neural network model is obtained based on the model parameters at the time of convergence.

4. The method according to claim 2, characterized in that, Determining the operating status of the furnace negative pressure closed-loop control system based on the real-time operating data includes: Active power is collected within a preset time window based on a preset sampling frequency, and the difference between the maximum and minimum values ​​of active power within the preset time window is calculated. If the difference is less than the active power fluctuation threshold, then the furnace negative pressure closed-loop control system is determined to be in a steady state. Otherwise, it is determined that the furnace negative pressure closed-loop control system is in a dynamic state.

5. The method according to claim 4, characterized in that, When the furnace negative pressure closed-loop control system is in a steady state, the furnace negative pressure adjustment deviation parameter calculated based on the real-time operating data for the corresponding operating state includes: Obtain the furnace negative pressure at each sampling moment within the preset time window, calculate the furnace negative pressure difference between the furnace negative pressure at each sampling moment and the furnace negative pressure setpoint, and calculate the actual value of steady-state adjustment deviation based on the furnace negative pressure difference, applying the formula: ； The active power within the preset time window is input into the trained second neural network model, so that the second neural network model can calculate the corresponding furnace negative pressure standard deviation based on the learned mapping relationship between the input and output quantities. The steady-state regulation deviation standard value is calculated based on the standard deviation of the furnace negative pressure, using the following formula: ； The furnace negative pressure regulation deviation parameter when the operating state is steady is calculated based on the actual value of the steady-state regulation deviation and the standard value of the steady-state regulation deviation, using the following formula: ； in, This is the actual value of the steady-state adjustment deviation; It is the actual value of the furnace negative pressure at the i-th data acquisition moment; This is the furnace negative pressure setpoint; N is the number of data collection points within the time window. It is the standard value of steady-state adjustment deviation; It is the tolerance coefficient; It is the standard deviation of the furnace negative pressure output by the second neural network model; It is the furnace negative pressure regulation deviation parameter when the operating state is steady; It is the proportionality coefficient.

6. The method according to claim 4, characterized in that, When the furnace negative pressure closed-loop control system is in a dynamic state, the furnace negative pressure adjustment deviation parameter calculated based on the real-time operating data for the corresponding operating state includes: The actual value of the dynamic adjustment deviation is calculated based on the difference between the furnace negative pressure and the furnace negative pressure setpoint, using the following formula: ； The load correction coefficient is determined based on the ratio of the current load to the rated load of the furnace negative pressure closed-loop control system, and the preset variable load rate correction coefficient is obtained. The standard value of dynamic adjustment deviation is calculated based on the load correction coefficient, the variable load rate correction coefficient, and the furnace negative pressure setpoint, using the following formula: ； Based on the actual value of the dynamic adjustment deviation and the standard value of the dynamic adjustment deviation, the furnace negative pressure adjustment deviation parameter when the operating state is dynamic is calculated, and the formula is applied: ； in, It is the actual value of the dynamic adjustment deviation; It is the negative pressure in the furnace when the operating state is dynamic; This is the furnace negative pressure setting value; It is the load correction factor; It is the variable load rate correction factor; It is a dynamic adjustment deviation standard value; It is the furnace negative pressure adjustment deviation parameter when the operating state is dynamic; It is the proportionality coefficient.

7. The method according to claim 1, characterized in that, The real-time operating data includes the target opening degree of the induced draft blades and the actual opening degree of the induced draft blades from the blade control system. The calculation of the blade deviation parameters based on the real-time operating data includes: The actual deviation of the induced draft blades between the target opening and the actual opening is calculated using the following formula: ； The blade deviation parameter is calculated based on the actual deviation of the induced draft blade and the preset standard deviation of the induced draft blade, using the following formula: ； in, This is the actual deviation of the induced draft fan blades; It is the target opening degree of the induced draft blades; This refers to the actual opening of the induced draft fan blades; It is the blade deviation parameter; It is the proportionality coefficient.

8. A machine learning-based furnace negative pressure closed-loop control system health monitoring device, characterized in that, include: The data acquisition unit is used to acquire real-time operating data of the furnace negative pressure closed-loop control system; The parameter calculation unit is used to input the real-time running data into the trained first neural network model to obtain the standard value of the feedforward parameter, and to calculate the feedforward adjustment deviation parameter based on the actual value of the feedforward parameter in the real-time running data and the standard value of the feedforward parameter. The parameter calculation unit is also used to determine the operating status of the furnace negative pressure closed-loop control system based on the real-time operating data, and calculate the furnace negative pressure adjustment deviation parameter corresponding to the operating status in combination with the real-time operating data. The parameter calculation unit is also used to calculate the blade deviation parameter based on the real-time operating data; The parameter calculation unit is also used to obtain a fusion parameter based on the weighted sum of the feedforward adjustment deviation parameter, the furnace negative pressure adjustment deviation parameter, and the moving blade deviation parameter; The health monitoring unit is used to obtain the health monitoring results of the furnace negative pressure closed-loop control system based on the fusion parameters.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 7.

Images