Modeling method and device of thermal power generating unit coordination control system, electronic equipment and storage medium

Abstract

The invention discloses a modeling method and device for a thermal power generating unit coordination control system, electronic equipment and a storage medium. The method comprises the following steps: acquiring current time sequence parameters of the thermal power generating unit coordinated control system at each acquisition time point within the current preset time length and current working condition parameters of the thermal power generating unit coordinated control system through a sensor network; inputting the current time sequence parameter and the current working condition parameter into a physical information neural network, and outputting a current modeling parameter of the thermal power generating unit coordination control system through the physical information neural network; and modeling the physical information neural network based on the current modeling parameters of the thermal power generating unit coordinated control system. According to the embodiment of the invention, the contradiction between the complexity of a mechanism model and the generalization of a data-driven model can be broken through, and the prediction precision and stability are remarkably improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the technical field of thermal power generation control, and in particular, to a modeling method, device, electronic device, and storage medium for a coordinated control system of a thermal power unit. Background Technique

[0002] A thermal power unit refers to a combination of a series of devices that use the heat energy generated by burning fossil fuels such as coal, oil, or natural gas. Through a boiler, water is heated into steam, and then the steam is used to drive a steam turbine to rotate, which in turn drives a generator to generate electricity. It is the core part of a traditional thermal power plant and is responsible for converting the chemical energy in the fuel into electrical energy. The coordinated control system of a thermal power unit is one of the very important control systems in modern large-scale thermal power plants. Its main function is to optimize the operating parameters of the boiler and the steam turbine so that the entire unit can efficiently and stably respond to changes in the grid load demand. With the expansion of the grid scale and the increase in the proportion of new energy power generation, thermal power units need to respond more flexibly to the requirements of grid dispatching, which requires the coordinated control system to have higher precision and faster response speed. Therefore, continuously improving and perfecting the coordinated control system is of great significance for improving the overall performance of thermal power units and ensuring the safe and stable operation of the power grid.

[0003] For the modeling method of the coordinated control system of a thermal power unit, currently, the following two methods are mainly adopted: 1) Traditional mechanism model: This method is based on transfer functions or state-space equations. By establishing dynamic equations of subsystems such as boilers and steam turbines, such as combustion lag compensation and pressure set curve design, combined with parameter identification technology, a model is constructed. However, the traditional mechanism model needs to make simplified assumptions about the thermodynamic process and is difficult to cover the non-linear dynamic characteristics under variable operating conditions of the unit, such as coal quality changes and equipment aging, resulting in an increase in prediction errors when the unit is at low load or rapidly changing load, and causing main steam pressure fluctuations. 2) Data-driven model: This method uses historical operation data to train models such as neural networks or support vector machines, such as the dynamic estimation method of coal calorific value. Through the input-output data mapping relationship, prediction is achieved, relying on feature engineering and black-box optimization. The pure data-driven method lacks explicit constraints on the laws of thermodynamics, resulting in the output results possibly violating physical laws and having insufficient reliability. Summary of the Invention

[0004] The present application provides a modeling method, device, electronic device, and storage medium for a coordinated control system of a thermal power unit, which can break through the contradiction between the complexity of the mechanism model and the generalization of the data-driven model, and significantly improve the prediction accuracy and stability.

[0005] In a first aspect, an embodiment of the present application provides a modeling method for a coordinated control system of a thermal power unit, and the method includes:

[0006] Collect the current time series parameters at each acquisition time point within the current predetermined time period of the coordinated control system of the thermal power unit and the current operating condition parameters of the coordinated control system of the thermal power unit through a sensor network;

[0007] Input the current time series parameters and the current operating condition parameters into a physics-informed neural network, and output the current modeling parameters of the coordinated control system of the thermal power unit through the physics-informed neural network;

[0008] Based on the current modeling parameters of the coordinated control system of the thermal power unit, model the coordinated control system of the thermal power unit.

[0009] In a second aspect, an embodiment of the present application further provides a modeling device for a coordinated control system of a thermal power unit. The device includes: a data acquisition module, an input-output module, and a model construction module; wherein,

[0010] The data acquisition module is configured to collect the current time series parameters at each acquisition time point within the current predetermined time period of the coordinated control system of the thermal power unit and the current operating condition parameters of the coordinated control system of the thermal power unit through a sensor network;

[0011] The input-output module is configured to input the current time series parameters and the current operating condition parameters into a physics-informed neural network, and output the current modeling parameters of the coordinated control system of the thermal power unit through the physics-informed neural network;

[0012] The model construction module is configured to model the coordinated control system of the thermal power unit based on the current modeling parameters of the coordinated control system of the thermal power unit.

[0013] In a third aspect, an embodiment of the present application provides an electronic device, including:

[0014] One or more processors;

[0015] A memory for storing one or more programs,

[0016] When the one or more programs are executed by the one or more processors, the one or more processors implement the modeling method for the coordinated control system of the thermal power unit according to any embodiment of the present application.

[0017] In a fourth aspect, an embodiment of the present application provides a storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the modeling method for the coordinated control system of the thermal power unit according to any embodiment of the present application.

[0018] The embodiment of the present application proposes a modeling method, device, electronic device and storage medium for a coordinated control system of a thermal power unit. First, the current time series parameters of the coordinated control system of the thermal power unit at each collection time point within the current predetermined time length and the current operating parameters of the coordinated control system of the thermal power unit are collected through a sensor network; then the current time series parameters and the current operating parameters are input into a physical information neural network, and the current modeling parameters of the coordinated control system of the thermal power unit are output through the physical information neural network; and then the physical information neural network is modeled based on the current modeling parameters of the coordinated control system of the thermal power unit. That is to say, in the technical solution of the present application, modeling the coordinated control system of the thermal power unit through a physical information neural network is equivalent to combining the traditional mechanism model with the data-driven model. The combination of the two can utilize the accuracy and interpretability of the mechanism model, and at the same time, with the help of the data-driven model's ability to handle nonlinear and complex relationships, thereby improving the accuracy and reliability of the prediction results. Therefore, compared with the prior art, the modeling method, device, electronic device and storage medium of the coordinated control system of thermal power units proposed in the embodiments of the present application can overcome the contradiction between the complexity of the mechanism model and the generalization of the data-driven model, and significantly improve the prediction accuracy and stability; moreover, the technical solution of the embodiments of the present application is simple and convenient to implement, easy to popularize, and has a wider range of applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A flow chart of a modeling method for a coordinated control system of a thermal power plant provided in one embodiment of the present application;

[0020] Figure 2 A model architecture diagram of a thermal power unit coordinated control system based on a physical information neural network provided in one embodiment of the present application;

[0021] Figure 3 A schematic diagram of a flow chart of a method for training a physical information neural network provided in one embodiment of the present application;

[0022] Figure 4 A schematic flow chart of a method for training a physical information neural network provided in another embodiment of the present application;

[0023] Figure 5 A schematic diagram of the structure of the deployment hardware of the physical information neural network provided in the embodiment of the present application;

[0024] Figure 6 A schematic diagram of the structure of a modeling device for a coordinated control system of a thermal power plant provided in one embodiment of the present application;

[0025] Figure 7 A schematic diagram of the structure of an electronic device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0026] The present application will be further described in detail below with reference to the accompanying drawings and embodiments. It can be understood that the specific embodiments described herein are only for explaining the present application, rather than limiting the present application. In addition, it should be noted that for the sake of description, only parts related to the present application rather than all structures are shown in the drawings.

[0027] Figure 1 As shown in the flowchart of the modeling method for the coordinated control system of a thermal power unit provided in an embodiment of the present application, this method can be executed by a modeling device or an electronic device of the coordinated control system of the thermal power unit. The device or the electronic device can be implemented in a software and / or hardware manner, and the device or the electronic device can be integrated in any intelligent device with network communication functions. As Figure 1 shown, the modeling method for the coordinated control system of a thermal power unit may include the following steps:

[0028] S101. Collect the current time series parameters at each acquisition time point within the current predetermined duration of the coordinated control system of the thermal power unit and the current operating condition parameters of the coordinated control system of the thermal power unit through a sensor network.

[0029] In one embodiment, the current time series parameters may include at least one of the following: feed water flow rate m w , feed water temperature T at the outlet of the economizer m1 , feed water pressure P at the outlet of the economizer m1 , reheated steam flow rate m r , steam pressure P at the inlet of the low-temperature reheater r1 , steam temperature T at the inlet of the low-temperature reheater r1 , total coal quantity m f , total air volume m a , water quantity of the first and second stage desuperheaters m mc , water quantity of the reheater desuperheater m rc ; the current operating condition parameters include at least one of the following: calorific value C of coal quality, equipment aging coefficient d.

[0030] S102. Input the current time series parameters and the current operating condition parameters into a physics-informed neural network, and output the current modeling parameters of the coordinated control system of the thermal power unit through the physics-informed neural network.

[0031] In one embodiment, the current modeling parameters include at least one of the following: load P, intermediate point temperature T, main steam temperature T m2 , main steam pressure P m2 , reheated steam pressure P r2 , reheated steam temperature T r2 , opening degree V of the steam turbine control valve.

[0032] S103. Model the coordinated control system of the thermal power unit based on the current modeling parameters of the coordinated control system of the thermal power unit.

[0033] In a specific embodiment of the present application, after obtaining the current modeling parameters of the coordinated control system of the thermal power unit, the coordinated control system of the thermal power unit can be modeled according to the following steps. These steps combine the key parameters and dynamic characteristics of the system to ensure that the model can accurately reflect the actual operating conditions. 1) Determine the modeling objective according to the obtained parameters and the operating requirements of the system. For example, achieve rapid response of the load P; optimize the control of the intermediate point temperature T and the main steam temperature T m2 ; improve the stability of the main steam pressure P m2 and the reheat steam pressure P r2 ; coordinate the dynamic matching between the boiler and the steam turbine. 2) The core of the coordinated control system of the thermal power unit is the coupling relationship among the boiler, the steam turbine and the generator. By analyzing the relationships among the parameters, the dynamic framework of the system can be initially constructed. 3) To construct the coordinated control model, the following strategies can be adopted: feedforward control, feedback control, and coupling control. In summary, after obtaining the modeling parameters of the coordinated control system of the thermal power unit, through a series of steps such as parameter classification, system structure analysis, dynamic model construction, and simulation verification, the modeling work of the system can be completed. The final model can not only reflect the dynamic characteristics of the system, but also provide theoretical support for optimizing the control strategy.

[0034] Figure 2 This is the model architecture diagram of the coordinated control system of the thermal power unit based on the physics-informed neural network provided in an embodiment of the present application. As Figure 2 shown, the input data of the physics-informed neural network can include: feedwater flow rate, feedwater temperature at the economizer outlet, feedwater pressure at the economizer outlet, reheat steam flow rate, steam pressure at the inlet of the low-temperature reheater, steam temperature at the inlet of the low-temperature reheater, total coal quantity, total air volume, water quantity for primary and secondary desuperheating, water quantity for reheat desuperheating, calorific value of coal quality, equipment aging coefficient; the output data of the physics-informed neural network can include: load, intermediate point temperature, main steam temperature, main steam pressure, reheat steam pressure, reheat steam temperature, steam turbine control valve opening.

[0035] In one embodiment, the physics-informed neural network can include: an input layer, a hidden layer, a physical constraint layer, and an output layer. 1) Input layer: This is the starting point of the neural network and is responsible for receiving external input data. The input data in this embodiment can include: current time series parameters and current operating condition parameters; among them, the current time series parameters include at least one of the following: feedwater flow rate m w , feedwater temperature T at the economizer outlet m1 , feedwater pressure P at the economizer outlet m1 , reheat steam flow rate m r, the inlet steam pressure P of the low-temperature reheater r1 , the inlet steam temperature T of the low-temperature reheater r1 , the total coal quantity m f , the total air quantity m a , the water quantity of the first and second-stage desuperheating m mc , the water quantity of the reheater desuperheating m rc ; The current operating condition parameters include at least one of the following: calorific value C of coal quality, equipment aging coefficient d. 2) Hidden layer: The hidden layer can contain multiple neurons, which process the input data through an activation function and transfer the processed information to the next layer. The hidden layer in this embodiment can adopt a deep residual network, which can include 6 residual blocks. Each residual block consists of two fully connected layers and contains a skip connection. Each fully connected layer has 256 neurons. The activation function is selected as Swish(β = 1.0) to balance the non-linear expression ability and gradient stability. 3) Physical constraint layer: This layer is the key difference between the physical information neural network and the traditional neural network. The physical constraint layer adds a physical equation (such as a partial differential equation) describing the system behavior as an additional loss term to the training process. This means that in addition to minimizing the prediction error of the observed data, it is also necessary to minimize the degree of violation of the physical laws. Doing so can make the model not only fit the observed data but also maintain physical consistency at the unobserved data points. The partial differential equation in this embodiment is: Among them, Q fuel = m f × C; m r × h(T r1 , P r1 ); Q ws_o = (m w + m mc ) × h(T m2 , P m2 ) + (m r + m rc ) × h(T r2 , P r2 ); Among them, m f is the total coal quantity; C is the calorific value of coal quality; m r is the reheater steam flow; T r1 is the inlet steam temperature of the low-temperature reheater; P r1 is the inlet steam pressure of the low-temperature reheater; m w is the feed water flow; m mc is the water quantity of the first and second-stage desuperheating; T m2 is the main steam temperature; P m2 is the main steam pressure; m rc is the water quantity of the reheater desuperheating; T r2 is the reheater steam temperature; P r2is the reheat steam pressure; h(·) is the water vapor physical property function that solves the enthalpy value according to the temperature and pressure. 4) Output layer: The output layer gives the final prediction results, which may be state variables of the physical system, such as temperature, pressure, etc. The output data in this embodiment may include: load P, intermediate point temperature T, main steam temperature T m2 , Main steam pressure P m2 , reheat steam pressure P r2 , Reheat steam temperature T r2 , steam turbine regulating valve opening V.

[0036] The modeling method of the coordinated control system of a thermal power unit proposed in the embodiment of the present application first collects the current time series parameters and the current operating parameters of the coordinated control system of the thermal power unit at each collection time point within the current predetermined time length through the sensor network; then the current time series parameters and the current operating parameters are input into the physical information neural network, and the current modeling parameters of the coordinated control system of the thermal power unit are output through the physical information neural network; and then the physical information neural network is modeled based on the current modeling parameters of the coordinated control system of the thermal power unit. That is to say, in the technical scheme of the present application, modeling the coordinated control system of the thermal power unit through the physical information neural network is equivalent to combining the traditional mechanism model with the data-driven model. The combination of the two can utilize the accuracy and interpretability of the mechanism model, and at the same time use the ability of the data-driven model to handle nonlinear and complex relationships, thereby improving the accuracy and reliability of the prediction results. Therefore, compared with the prior art, the modeling method of the coordinated control system of a thermal power unit proposed in the embodiment of the present application can break through the contradiction between the complexity of the mechanism model and the generalization of the data-driven model, and significantly improve the prediction accuracy and stability; and the technical scheme of the embodiment of the present application is simple and convenient to implement, easy to popularize, and has a wider range of application.

[0037] Figure 3 A flow chart of a training method for a physical information neural network provided in one embodiment of the present application. The above technical solution is further optimized and expanded, and can be combined with the above optional implementation methods. Figure 3 As shown, the training method of the physical information neural network may include the following steps:

[0038] S301. If the physical information neural network does not meet the preset convergence conditions, the historical time series parameters of the thermal power unit coordinated control system at each acquisition time point within each historical predetermined time period and the historical operating condition parameters of the thermal power unit coordinated control system are extracted from the pre-constructed data training set.

[0039] In one embodiment, the historical time series parameters may include at least one of the following: water flow rate m w , EcoGas outlet water temperature Tm1 、 Feed water pressure P at the gas-saving outlet m1 、 Reheat steam flow rate m r 、 Inlet steam pressure P of the low-temperature reheater r1 、 Inlet steam temperature T of the low-temperature reheater r1 、 Total coal quantity m f 、 Total air volume m a 、 Water quantity of the first and second-stage desuperheaters m mc 、 Water quantity of the reheat desuperheater m rc ; The historical operating condition parameters may include at least one of the following: calorific value C of coal quality, equipment aging coefficient d.

[0040] S302. Input the historical time series parameters at each acquisition time point within each historical predetermined duration of the coordinated control system of the thermal power unit and the historical operating condition parameters of the coordinated control system of the thermal power unit into the physics-informed neural network, and output the prediction modeling parameters of the coordinated control system of the thermal power unit through the physics-informed neural network.

[0041] The prediction modeling parameters in the embodiments of the present application may include at least one of the following: load P, intermediate point temperature T, main steam temperature T m2 、 Main steam pressure P m2 、 Reheat steam pressure P r2 、 Reheat steam temperature T r2 、 Turbine control valve opening V.

[0042] S303. Train the physics-informed neural network based on the prediction modeling parameters of the coordinated control system of the thermal power unit, the true modeling parameters of the coordinated control system of the thermal power unit collected in advance, and the pre-constructed loss function until the physics-informed neural network meets the convergence condition; wherein, the loss function is determined by a data loss term, a physical loss term, a gradient amplitude dynamic adjustment coefficient corresponding to the data loss term, and a gradient amplitude dynamic adjustment coefficient corresponding to the physical loss term.

[0043] The loss function in the embodiments of the present application can be calculated according to the following formula: L = λ(t) × L data + [1 - λ(t)] × L physics ; wherein, L data is the data loss term; L physics is the physical loss term; λ(t) is the gradient amplitude dynamic adjustment coefficient corresponding to the data loss term; 1 - λ(t) is the gradient amplitude dynamic adjustment coefficient corresponding to the physical loss term; wherein, wherein, is the gradient; θ is the weight of the physics-informed neural network.

[0044] By dynamically adjusting the weights of the data loss term and the physical loss term, this application can ensure that the model can not only fit the observed data well (i.e., reduce the data loss), but also follow the known physical laws (i.e., reduce the physical loss). This balance helps to improve the robustness of the model and make it perform more stably on unseen data. In addition, an appropriate weight adjustment strategy can accelerate the convergence speed of the model. If the training in a certain stage focuses too much on one aspect, such as an overly heavy data loss term may lead to overfitting, it may prolong the overall training time. By dynamically adjusting the coefficient, this situation can be effectively avoided, enabling the model to reach the optimal solution faster.

[0045] Figure 4 It is a schematic flowchart of the training method of the physics-informed neural network provided by another embodiment of this application. It is further optimized and extended based on the above technical solutions and can be combined with each of the above optional implementation manners. As Figure 4 shown, the training method of the physics-informed neural network may include the following steps:

[0046] S401. If the physics-informed neural network does not meet the pre-set convergence condition, extract the historical time series parameters at each acquisition time point within each historical preset time length of the coordinated control system of the thermal power unit and the historical operating conditions parameters of the coordinated control system of the thermal power unit from the pre-constructed data training set.

[0047] In one embodiment, the Min-Max method can be used to normalize each parameter. Specifically, it can be calculated according to the following formula: where, x norm is the parameter after normalization; x is the original parameter; x max is the maximum value of the original parameter within its fluctuation range; x min is the minimum value of the original parameter within its fluctuation range. Using the Min-Max method to normalize each parameter is a data preprocessing technique, and its core idea is to linearly transform the data to a specific range (usually [0, 1] or [-1, 1]). This can effectively solve problems such as data scale differences, numerical stability, and gradient propagation, thereby accelerating model convergence, improving generalization ability, and simplifying the optimization process in multi-task learning.

[0048] S402. Input the historical time series parameters at each acquisition time point within each historical preset time length of the coordinated control system of the thermal power unit and the historical operating conditions parameters of the coordinated control system of the thermal power unit into the physics-informed neural network, and output the prediction modeling parameters of the coordinated control system of the thermal power unit through the physics-informed neural network.

[0049] S403. Calculate the original change amount of the prediction modeling parameters of the coordinated control system of the thermal power unit relative to the true modeling parameters of the coordinated control system of the thermal power unit collected in advance.

[0050] In one embodiment, the true modeling parameters of the coordinated control system of the thermal power unit can be subtracted from the prediction modeling parameters of the coordinated control system of the thermal power unit to obtain the original change amount of the prediction modeling parameters of the coordinated control system of the thermal power unit relative to the true modeling parameters of the coordinated control system of the thermal power unit collected in advance.

[0051] S404. Multiply the original change amount by the dynamic learning rate to obtain the true change amount of the prediction modeling parameters of the coordinated control system of the thermal power unit relative to the true modeling parameters of the coordinated control system of the thermal power unit collected in advance.

[0052] The dynamic learning rate in the embodiments of the present application can be calculated according to the following formula: where η t is the dynamic learning rate; η min is the pre-determined minimum learning rate; η max is the pre-determined maximum learning rate; t max is the pre-determined maximum number of iterations.

[0053] The present application adopts the cosine annealing strategy to smoothly adjust the learning rate, which can avoid sudden large changes. This smooth change helps to stabilize the training process and reduce the model performance fluctuations caused by the sudden change of the learning rate. In the initial stage of training, a larger learning rate can help quickly cross the vast area in the parameter space and find the approximate position of a more ideal solution. As the training progresses, the learning rate gradually decreases, enabling the model to more finely adjust the parameters and approach the local optimal solution. Cosine annealing effectively combines the advantages of these two stages through its unique curve shape, thus accelerating the convergence of the entire training process; and by periodically increasing and decreasing the learning rate, cosine annealing can prevent the model from overfitting on the training set because this strategy encourages the model to explore the loss surface at different scales. The result of this is that even when facing unseen data, the model can exhibit good generalization performance.

[0054] S405. Train the physics-informed neural network according to the true change amount and the pre-constructed loss function until the physics-informed neural network meets the convergence condition; wherein, the loss function is determined by the data loss term, the physical loss term, the gradient amplitude dynamic adjustment coefficient corresponding to the data loss term, and the gradient amplitude dynamic adjustment coefficient corresponding to the physical loss term.

[0055] The training method of the physics-informed neural network proposed in the embodiments of this application uses the cosine annealing strategy to adjust the learning rate, which can not only help the model reach the convergence state faster, but also improve the final performance and generalization ability of the model. At the same time, it simplifies the parameter adjustment work in the training process, which is particularly beneficial for tasks that require long-term training or processing of large-scale data sets.

[0056] In one embodiment, the training of the physics-informed neural network can also be automatically triggered by an online incremental update mechanism. For example, when the absolute value of the prediction deviation of key parameters such as unit load, intermediate point temperature, main steam pressure and temperature exceeds the ±5% threshold, or the fluctuation range of fuel calorific value exceeds 10%, the system can automatically trigger the online incremental update mechanism. This mechanism realizes adaptive triggering by dynamically monitoring the model prediction performance index to ensure timely response when the working conditions change significantly. In addition, this embodiment adopts a progressive parameter update scheme, only fine-tuning the weights of the last two layers of the fully connected layer, and keeping the parameters of the underlying feature extraction network frozen to retain the learned feature expression ability. The model update uses the latest data within a 30-minute time window for iterative optimization, combined with an adaptive learning rate adjustment mechanism, which not only ensures the adaptability of the model to the latest working conditions, but also avoids the problem of catastrophic forgetting.

[0057] In one embodiment, the cross-validation method can also be used to verify the prediction results of the physics-informed neural network. Specifically, the 5-fold stratified cross-validation method is adopted, and stratified sampling is carried out according to the unit load intervals (30-50%, 50-70%, 70-100%) to ensure the data distribution consistency of the training set and the validation set under different working conditions. This strategy ensures that each sample only participates in one verification through non-repetitive sampling, effectively reducing the variance bias of model evaluation.

[0058] In one embodiment, the physics-informed neural network can also be evaluated for performance through a performance evaluation system. The core indicators can include: mean absolute error (MAE), root mean square error (RMSE) and coefficient of determination (R 2 ); among them, MAE and RMSE quantify the prediction deviation through the original dimension, and R 2 is used to evaluate the model interpretability. This embodiment can also strictly control the error threshold. For example, the absolute percentage error (APE) of the predicted values of load, intermediate point temperature, main steam pressure and temperature needs to be less than or equal to 1.5%, and the continuous ranked probability score is used to assist in verifying the prediction precision.

[0059] In one example, external data sets can also be tested, mainly including: coal quality mutation scenarios and equipment aging scenarios; 1) Coal quality mutation scenario: Simulate the working condition where the ash content increases from 15% to 20% and the calorific value decreases by 8%. The model prediction error should be kept within 2.1%, which is better than the traditional mechanism model (>5.2%). This test verifies the non-linear mapping ability of the model for industrial analysis indicators of coal quality. 2) Equipment aging scenario: Build an aging model with a 2% decline in boiler efficiency. Through the online incremental learning strategy (only update the parameters of the last two layers of the network), the prediction error is converged to less than 1.0% within 6 hours. This process can effectively test the error accumulation effect caused by equipment performance drift.

[0060] Figure 5 Schematic diagram of the structure of the deployment hardware of the physical information neural network provided by the embodiment of the present application. As Figure 5 shown, the deployment hardware may include: a distributed control system DCS, a switch, a computing cluster, and a workstation; wherein, the distributed control system and the switch transmit data through the OPC UA or MQTT protocol. The computing cluster in the embodiment of the present application may be a Jetson AGX Orin cluster. 1) Edge computing node: Computing power upgrade: Adopt NVIDIA Jetson AGX Orin, support higher concurrent inference tasks, and meet the multi-variable real-time control requirements of thermal power units. Memory expansion: Adopt 64GB LPDDR5 memory to ensure high-frequency data throughput capacity. Storage redundancy: Deploy 2×1TB NVMe SSDs configured in RAID 1 to ensure the high availability of key models and data. Dynamic inference acceleration: Utilize the dynamic input size feature of TensorRT 8.6, adapt to different sensor resolutions, and combine the cross-frame compatibility of the ONNX model to achieve millisecond-level inference latency. Containerized deployment: Package the PyTorch 2.0 model through Docker, and combine Kubernetes to achieve elastic scaling and version rollback of edge nodes. 2) DCS interface module: Dual-protocol redundancy: The main channel adopts OPC UA, and the backup channel deploys MQTT over TLS to ensure that key data is transmitted back through the 4G / 5G slice network in case of network interruption. Data caching strategy: Expand the circular buffer to 2 hours of data, and add an edge cache layer based on the time series database to support breakpoint resumption and historical data traceability. Introduce edge stream processing to filter noise data in real time and extract key features. 3) Security protection mechanism: Transport layer security: Enable TLS1.3 encryption for OPC UA communication, and combine the national secret SM2 / SM4 algorithms to form a hybrid encryption system. Key management: The HSM integrates a TPM 2.0 chip, supports key rotation and hardware-level side-channel attack prevention. Fine-grained RBAC: Design an "auditor" role, only allowing viewing of operation logs; engineer permissions are bound to two-factor authentication.

[0061] In one embodiment, the network design for deploying the hardware may include: 1) Edge layer: Multiple Jetson AGX Orin form a Kubernetes cluster, which is connected to the DCS and the sensor network through a TSN switch (supporting IEEE 802.1AS clock synchronization). A local firewall is deployed to isolate the control network and the data network. 2) Aggregation layer: An industrial-grade core switch realizes VLAN division and transmits OPC UA and model update traffic in segments. An edge computing gateway is deployed to support 5G / fiber dual-mode backhaul. 3) Cloud / central layer: The model training and parameter optimization are centrally managed through a private cloud platform, and A / B testing is used to verify the new model version. An API gateway is deployed in the security isolation area, and only the necessary northbound interfaces are opened.

[0062] Figure 6 FIG. is a schematic structural diagram of a modeling device for a coordinated control system of a thermal power unit provided in an embodiment of the present application. As Figure 6 shown, the modeling device for the coordinated control system of the thermal power unit includes: a data acquisition module 601, an input / output module 602, and a model construction module 603; wherein,

[0063] The data acquisition module 601 is configured to collect the current time series parameters and the current operating condition parameters of the coordinated control system of the thermal power unit at each acquisition time point within a current predetermined time period through a sensor network;

[0064] The input / output module 602 is configured to input the current time series parameters and the current operating condition parameters into a cyber-physical neural network, and output the current modeling parameters of the coordinated control system of the thermal power unit through the cyber-physical neural network;

[0065] The model construction module 603 is configured to model the coordinated control system of the thermal power unit based on the current modeling parameters of the coordinated control system of the thermal power unit.

[0066] The above-mentioned modeling device for the coordinated control system of the thermal power unit can execute the method provided in any embodiment of the present application, and has the corresponding functional modules and beneficial effects for executing the method. For technical details not described in detail in this embodiment, reference can be made to the modeling method for the coordinated control system of the thermal power unit provided in any embodiment of the present application.

[0067] Figure 7 FIG. is a schematic structural diagram of an electronic device provided in an embodiment of the present application. Figure 7 FIG. shows a block diagram of an exemplary electronic device suitable for implementing the embodiments of the present application. Figure 7 The electronic device 12 shown is only an example and should not impose any limitation on the functions and usage scope of the embodiments of the present application.

[0068] AsFigure 7 As shown, the electronic device 12 is presented in the form of a general-purpose computing device. The components of the electronic device 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 that connects different system components (including the system memory 28 and the processing unit 16).

[0069] The bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus structures. By way of example, these architectures include, but are not limited to, Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnect (PCI) bus.

[0070] The electronic device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the electronic device 12, including volatile and non-volatile media, removable and non-removable media.

[0071] The system memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. The electronic device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, a storage system 34 can be used to read and write non-removable, non-volatile magnetic media ( Figure 7 not shown, typically referred to as a "hard disk drive"). Although Figure 7 not shown in the figure, a disk drive for reading and writing removable non-volatile disks (such as "floppy disks") and an optical disk drive for reading and writing removable non-volatile optical disks (such as CD-ROM, DVD-ROM, or other optical media) can be provided. In these cases, each drive can be connected to the bus 18 through one or more data media interfaces. The memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of the embodiments of the present application.

[0072] A program / utility 40 having a set (at least one) of program modules 42 can be stored, for example, in the memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. The program modules 42 generally perform the functions and / or methods described in the embodiments of the present application.

[0073] The electronic device 12 can also communicate with one or more external devices 14 (such as a keyboard, a pointing device, a display 24, etc.), and can also communicate with one or more devices that enable a user to interact with the electronic device 12, and / or communicate with any device that enables the electronic device 12 to communicate with one or more other computing devices (such as a network card, a modem, etc.). Such communication can be carried out through the input / output (I / O) interface 22. Moreover, the electronic device 12 can also communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through the network adapter 20. As shown in the figure, the network adapter 20 communicates with other modules of the electronic device 12 through the bus 18. It should be understood that although Figure 7 not shown in the figure, other hardware and / or software modules can be used in combination with the electronic device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0074] The processing unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, such as implementing the modeling method of the coordinated control system of a thermal power unit provided in the embodiments of the present application.

[0075] The embodiments of the present application also provide a computer storage medium.

[0076] The computer-readable storage medium of the embodiments of the present application can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples (a non-exhaustive list) of the computer-readable storage medium include: an electrical connection having one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, the computer-readable storage medium can be any tangible medium that contains or stores a program, and this program can be used by or in combination with an instruction execution system, apparatus, or device.

[0077] A computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which computer-readable program code is carried. Such propagated data signals may take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device.

[0078] The program code contained on a computer-readable medium may be transmitted using any appropriate medium, including - but not limited to - wireless, wireline, optical fiber cable, RF, and the like, or any suitable combination of the foregoing.

[0079] The computer program code for performing the operations of the present application may be written in one or more programming languages or combinations thereof. The programming languages include object-oriented programming languages such as Java, Smalltalk, C++, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, executed as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0080] The embodiments of the present application also provide a computer program product.

[0081] The various embodiments of the systems and techniques described above in this document may be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGA), application specific integrated circuits (ASIC), application specific standard products (ASSP), system on a chip (SOC), complex programmable logic devices (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include being implemented in one or more computer program products, which may include one or more computer programs. The one or more computer programs may be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0082] Note that the above are only the preferred embodiments of the present application and the technical principles applied. Those skilled in the art will understand that the present application is not limited to the specific embodiments described herein. Various obvious changes, re-adjustments, and substitutions can be made by those skilled in the art without departing from the protection scope of the present application. Therefore, although the present application has been described in more detail through the above embodiments, the present application is not limited to the above embodiments. Without departing from the concept of the present application, more other equivalent embodiments can be included, and the scope of the present application is determined by the scope of the appended claims.

Claims

1. A modeling method for a coordinated control system of a thermal power unit, characterized in that The method includes: Collecting current time series parameters at each acquisition time point within a current predetermined time period of the coordinated control system of a thermal power unit and current operating condition parameters of the coordinated control system of the thermal power unit through a sensor network; Inputting the current time series parameters and the current operating condition parameters into a cyber-physical neural network, and outputting current modeling parameters of the coordinated control system of the thermal power unit through the cyber-physical neural network; Modeling the coordinated control system of the thermal power unit based on the current modeling parameters of the coordinated control system of the thermal power unit.

2. The method according to claim 1, wherein: The current time series parameters include at least one of the following: feed water flow rate m w , feed water temperature T at the economizer outlet m1 , feed water pressure P at the economizer outlet m1 , reheater steam flow rate m r , steam pressure P at the inlet of the low-temperature reheater r1 , steam temperature T at the inlet of the low-temperature reheater r1 , total coal quantity m f , total air volume m a , water flow rate m for the first and second stage desuperheating mc , water flow rate m for reheater desuperheating rc ; The current operating condition parameters include at least one of the following: calorific value C of coal quality, equipment aging coefficient d; The current modeling parameters include at least one of the following: load P, intermediate point temperature T, main steam temperature T m2 , main steam pressure P m2 , reheater steam pressure P r2 , reheater steam temperature T r2 , steam turbine control valve opening V.

3. The method according to claim 1, wherein The method further includes: If the cyber-physical neural network does not meet a pre-set convergence condition, extracting historical time series parameters at each acquisition time point within each historical predetermined time period of the coordinated control system of the thermal power unit and historical operating condition parameters of the coordinated control system of the thermal power unit from a pre-constructed data training set; Inputting the historical time series parameters at each acquisition time point within each historical predetermined time period of the coordinated control system of the thermal power unit and the historical operating condition parameters of the coordinated control system of the thermal power unit into the cyber-physical neural network, and outputting predicted modeling parameters of the coordinated control system of the thermal power unit through the cyber-physical neural network; Training the cyber-physical neural network based on the predicted modeling parameters of the coordinated control system of the thermal power unit, the true modeling parameters of the coordinated control system of the thermal power unit collected in advance, and a pre-constructed loss function until the cyber-physical neural network meets the convergence condition; wherein, the loss function is determined by a data loss term, a physical loss term, a gradient amplitude dynamic adjustment coefficient corresponding to the data loss term, and a gradient amplitude dynamic adjustment coefficient corresponding to the physical loss term.

4. The method according to claim 3, characterized in that, The loss function is calculated according to the following formula: L = λ(t) × L data + [1 - λ(t)] × L physics ; where, L data is the data loss term; L physics is the physical loss term; λ(t) is the dynamic adjustment coefficient of the gradient amplitude corresponding to the data loss term; 1 - λ(t) is the dynamic adjustment coefficient of the gradient amplitude corresponding to the physical loss term; where, where, is the gradient; θ is the weight of the physical information neural network.

5. The method according to claim 4, wherein The physical loss term is calculated according to the following formula: where L physics is the physical loss term; N is the length window of the historical predetermined duration; E is the internal energy of the boiler; and t is each acquisition time point within the historical predetermined duration.

6. The method according to claim 5, wherein Among them, Q fuel = m f × C; m r × h(T r1 , P r1 )); Q ws_o = (m w + m mc ) × h(T m2 , P m2 ) + (m r + m rc ) × h(T r2 , P r2 )); Among them, m f is the total coal quantity; C is the calorific value of coal quality; m r is the reheater steam flow rate; T r1 is the inlet steam temperature of the low-temperature reheater; P r1 is the inlet steam pressure of the low-temperature reheater; m w is the feed water flow rate; m mc is the water quantity of the first and second stage desuperheating; T m2 is the main steam temperature; P m2 is the main steam pressure; m rc is the water quantity of the reheater desuperheating; T r2 is the reheater steam temperature; P r2 is the reheater steam pressure; h(·) is the water vapor property function for solving the enthalpy value according to temperature and pressure.

7. The method according to claim 3, characterized in that Training the cyber-physical neural network based on the predicted modeling parameters of the coordinated control system of the thermal power unit, the true modeling parameters of the coordinated control system of the thermal power unit collected in advance, and a pre-constructed loss function includes: Calculating an original change amount of the predicted modeling parameters of the coordinated control system of the thermal power unit relative to the true modeling parameters of the coordinated control system of the thermal power unit collected in advance; Multiply the original change amount by the dynamic learning rate to obtain the true change amount of the prediction modeling parameter of the coordinated control system of the thermal power unit relative to the true modeling parameter of the coordinated control system of the thermal power unit collected in advance; wherein, the dynamic learning rate is calculated according to the following formula: where η t is the dynamic learning rate; η min is the minimum learning rate determined in advance; η max is the maximum learning rate determined in advance; t max is the maximum number of iterations determined in advance. Training the cyber-physical neural network according to the true change amount and the pre-constructed loss function.

8. A modeling device for a coordinated control system of a thermal power unit, characterized in that, The device includes: a data acquisition module, an input / output module, and a model construction module; wherein, The data acquisition module is configured to collect current time series parameters at each acquisition time point within a current predetermined time period of the coordinated control system of a thermal power unit and current operating condition parameters of the coordinated control system of the thermal power unit through a sensor network; The input / output module is configured to input the current time series parameters and the current operating condition parameters into a cyber-physical neural network, and output current modeling parameters of the coordinated control system of the thermal power unit through the cyber-physical neural network; The model construction module is used to model the coordinated control system of the thermal power unit based on the current modeling parameters of the coordinated control system of the thermal power unit.

9. An electronic device, characterized in that, It includes: One or more processors; A memory for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the modeling method of the coordinated control system of the thermal power unit according to any one of claims 1 to 7.

10. A storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the modeling method of the coordinated control system of the thermal power unit according to any one of claims 1 to 7.

Images