Intelligent coordinated control system and method based on double-reheat unit

Abstract

The application discloses an intelligent coordination control system and method based on a double-reheat unit, and belongs to the technical field of energy, which comprises the following steps: collecting and preprocessing the operation data of a boiler and a steam turbine through a sensor, combining a physical mechanism of a thermal system with an ST-Attention-LSTM model, constructing a dynamic response prediction model of the thermal system driven by a mechanism and data, realizing multivariate prediction, and setting an initial value of a control parameter; using a K-means clustering classification system to disturb and calculate a weighted compensation amount, and obtaining a corrected control parameter through feedback correction; using a rolling optimization algorithm to dynamically adjust the control parameter, combining a working condition-rule mapping update mechanism with a water quality prediction model, and setting a water flow setting value through adaptive fuzzy control; through unattended and real-time coal yard management, an automatic repair, start-stop and alarm mechanism is set, carbon emission is monitored, and an optimization strategy is formulated; and the application can improve the operation efficiency of the unit, reduce energy consumption, and enhance the stability and response speed of the system.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of energy, in particular to an intelligent coordination control system and method based on a double-reheat unit. BACKGROUND

[0002] Under the background of global energy transformation, coal-fired power generation is gradually changing from the main body of power supply to supporting and regulating power, and is more involved in the operation safety and peak regulation tasks of new power systems. As an advanced technology of coal-fired power generation, the double-reheat technology has complex heating surfaces and thermal inertia, making it difficult to accurately control the steam temperature, which not only affects the operation efficiency and safety of the unit, but also restricts the further development of coal-fired power generation technology. With the increasing demand for grid peak regulation, the optimization of coal-fired power generation thermal control technology has become an urgent need, and realizing rapid and accurate regulation and control of steam temperature is of great significance to ensure the stable operation of the power grid.

[0003] A double-reheat once-through boiler-turbine DBC coordination control system is disclosed in a patent with the authorized announcement number CN102566433B, which includes a two-way direct energy and mass balance coordination control system composed of an energy demand signal of the double-reheat steam turbine to the boiler energy and heat signals and fuel-water ratio signals representing the energy and mass balance of the double-reheat once-through boiler, realizing dynamic decoupling control of the unit. At the same time, considering the complex characteristics of the double-reheat unit, dynamic weighting compensation of the boiler characteristic point heat, steam turbine pressure correction, and asymmetric constraint adaptive correction compensation links are designed, effectively improving the regulation quality of the unit of the thermal power plant with double-reheat.

[0004] The above existing technology has the following problems: 1) The DBC coordination control system mainly relies on the pre-designed two-way direct energy and mass balance coordination control system, limiting its ability to adapt to complex and variable working conditions; 2) Lack of real-time learning and self-optimization ability; 3) The DBC system involves multiple complex compensation links and control logic, which increases the overall complexity of the system, thereby increasing the difficulty of maintenance, debugging, and troubleshooting. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides an intelligent coordinated control system and method based on a double-reheat unit, which collects and pre-processes the operation data of a boiler and a steam turbine through a sensor, combines a physical mechanism of a thermal system with an ST-Attention-LSTM model to construct a mechanism-data dual-driven dynamic response prediction model of the thermal system, realizes multivariate prediction and sets an initial value of a control parameter, classifies system disturbances using a K-means clustering method, calculates a weighted compensation amount, and obtains a corrected control parameter through feedback correction, dynamically adjusts the control parameter using a rolling optimization algorithm, combines a working condition-rule mapping update mechanism with a water quality prediction model to adaptively set a water flow value, manages a coal yard through unattended operation and real-time coal inventory, sets an automatic repair, start-stop and alarm mechanism, monitors carbon emissions and formulates an optimization strategy, and can improve the operation efficiency of the unit, reduce energy consumption, and enhance the stability and response speed of the system.

[0006] To achieve the above object, the present application provides the following technical scheme:

[0007] An intelligent coordinated control system based on a double-reheat unit, comprising an intelligent control module, a coordinated control module and an auxiliary control module.

[0008] The intelligent control module comprises a prediction control unit and an intelligent feedforward unit; the prediction control unit is configured with a multivariate prediction control strategy, which is used to predict the real-time operation state of the boiler and the steam turbine according to historical data, real-time operation data and the physical mechanism of the thermal system, using a pre-constructed mechanism-data dual-driven dynamic response prediction model of the thermal system; the mechanism-data dual-driven dynamic response prediction model of the thermal system is constructed based on an ST-Attention-LSTM model; the intelligent feedforward unit is configured with an intelligent feedforward control strategy; the intelligent feedforward control strategy is used to correct the prediction results through K-means clustering.

[0009] The coordinated control module comprises a boiler main control unit and a steam turbine main control unit; the boiler main control unit dynamically adjusts the valve opening degree and the boiler combustion rate according to the received load instruction, optimizes the combustion process of the boiler, and matches the energy of the boiler with the output energy of the steam turbine; the steam turbine main control unit is used to receive the load instruction and adjust the steam admission amount, speed and power of the steam turbine according to the received load instruction.

[0010] The auxiliary control module comprises a feedwater control unit; the feedwater control unit is configured with an adaptive fuzzy control strategy; the adaptive fuzzy control strategy is used to automatically call or correct the fuzzy control rule library according to a pre-set working condition-rule mapping update mechanism, combines a pre-constructed water quality prediction model to obtain the water quality trend in advance, converts the fuzzy set into a control amount through the gravity center method, and automatically adjusts the set value of the feedwater flow; the water quality prediction model is configured based on a GRU network.

[0011] Specifically, the prediction control unit adopts a multivariable prediction control strategy, and the specific steps include:

[0012] A1: Collecting real-time operation data of the boiler and the steam turbine through sensors , wherein the collected operation data is preprocessed, and the preprocessing includes: representing the first k operation data, k representing the number of collected operation data; the operation data includes secondary reheat steam temperature, pressure, fuel flow, feed water flow, steam turbine admission, and rotating speed;

[0013] A2: Determining input features of the ST-Attention-LSTM model in combination with a physical mechanism of the thermal system, initializing weights and thresholds of the ST-Attention-LSTM model, training the ST-Attention-LSTM model using the preprocessed operation data, and obtaining a mechanism-data dual-driven dynamic response prediction model of the thermal system;

[0014] A3: Setting a sampling time T, based on the mechanism-data dual-driven dynamic response prediction model of the thermal system, calculating, according to the operation data at the current time, the prediction result of the operation state of the boiler and the steam turbine in the next sampling time T through the hidden layer of the mechanism-data dual-driven dynamic response prediction model of the thermal system , wherein the prediction result of the operation state of the boiler and the steam turbine in the next sampling time T includes: representing the first k predicted operation data;

[0015] A4: Setting initial values of control parameters of the boiler and the steam turbine, setting a target function, operation constraints, and initial parameters of a rolling optimization algorithm, calculating a target function value according to the prediction result, checking whether the target function value meets the operation constraints, if not, adjusting the control parameters of the boiler and the steam turbine using the rolling optimization algorithm, and rolling the time window forward by one sampling time, repeating the step A4 to obtain new control parameters .

[0016] Specifically, the training step of the ST-Attention-LSTM model includes:

[0017] A2.1: Dividing the preprocessed operation data into a training set, a validation set, and a test set according to a ratio of 7:2:1, wherein the operation data dimension includes a time series dimension and a feature dimension;

[0018] A2.2: Initializing weights and thresholds of an attention layer of the model, LSTM layer, and thresholds, wherein the input of the attention layer is a hidden state of the LSTM layer;

[0019] A2.3: Using the mean square error of the model prediction value and the actual operation data as the loss function, the ST-Attention-LSTM model is trained using the Adam optimizer;

[0020] A2.4: The generalization ability of the ST-Attention-LSTM model is evaluated using the test set, and if the mean square error exceeds the set threshold, the network structure of the ST-Attention-LSTM model is adjusted again until the accuracy requirement is met.

[0021] Specifically, the logic of the objective function in A4 is: the objective function value comprehensively reflects the optimization objectives of three dimensions of boiler fuel cost, matching degree of boiler output power and demand power, and system operation efficiency, and the importance of the three dimensions is adjusted by weight factors, wherein the boiler fuel cost is determined by the fuel flow, the matching degree of the boiler output power and the demand power is determined by the difference between the two, and the system operation efficiency is calculated from the real-time operation parameters of the unit.

[0022] Specifically, the intelligent feedforward unit adopts an intelligent feedforward control strategy, and the specific steps include:

[0023] B1: Real-time monitoring of disturbance information of the system through sensors ; The disturbance information includes fuel heat value fluctuation, power grid load impact, water hardness change, and equipment fault early warning signal, wherein, l represents the number of disturbance information, represents the l th disturbance information;

[0024] B2: Real-time classification of the disturbance information using K-means clustering, and according to the classified disturbance information and the prediction result of the mechanism-data dual-driven thermal system dynamic response prediction model, the prediction error of the original control parameter is obtained, and the compensation amount of the prediction error is calculated using the intelligent feedforward control algorithm, and the formula is: , wherein, represents the compensation amount of the prediction error , represents the feedforward control function;

[0025] B3: Set the error threshold to H, and add the obtained compensation amount to the original control parameter to obtain the control parameter error after adding the compensation amount ;

[0026] If , the output variable is adjusted according to the prediction result;

[0027] If ​then the control parameters are adjusted using a rolling optimization algorithm, and based on real-time operating state data, the predicted results are corrected using a feedback correction control algorithm to obtain corrected predicted values wherein, denotes t the corrected predicted value of the i-th output variable at time t; i

[0028] B4: based on the corrected predicted results, the control parameters are re-adjusted;

[0029] B5: repeat B1 to B4 until the system reaches a steady state, and after feedback correction, the corrected control parameters are obtained and used to drive the operation of the boiler and the steam turbine.

[0030] Specifically, the logic of the feedback correction control algorithm in B3 is as follows: t the corrected predicted value of the i-th output variable at time t is equal to the original predicted value of the output variable plus the sum of the product of the feedback correction coefficient of the output variable and the corresponding prediction error, plus an additional correction term determined according to the real-time operating conditions of the system, wherein the feedback correction coefficient is obtained by training historical prediction error and actual deviation data, and the additional correction term is used to compensate for the deviation not covered by the model under extreme operating conditions.

[0031] Specifically, the feedwater control unit adopts an adaptive fuzzy control strategy, and the specific steps include:

[0032] C1: set the feedwater flow domain, divide the low, moderate, and high fuzzy sets within the feedwater flow domain, and configure a triangular membership function for each fuzzy set, wherein the vertex position of the membership function is calibrated according to historical optimal feedwater flow data, and the feedwater flow domain is determined according to the rated evaporation capacity of the boiler;

[0033] C2: map the corrected predicted value to the configured triangular membership function, and calculate the membership values of the corrected predicted value belonging to each fuzzy set;

[0034] C3: according to system requirements, develop a fuzzy control rule library, combine the membership values in C2, and according to the fuzzy control rules in the fuzzy control rule library, use the maximum and minimum reasoning method to calculate the output fuzzy set, i.e., determine the fuzzy interval to which the feedwater flow should belong; the system requirements include water quality standards, steam temperature stability range, and unit load response speed.

[0035] Specifically, the feedwater control unit adopts an adaptive fuzzy control strategy, and the specific steps further include:

[0036] ​C4: Introduce the working condition-rule mapping update mechanism, identify the current working condition of the unit in real time, automatically call the corresponding fuzzy control rule from the fuzzy control rule library in C3 according to the working condition type, if there is no matching fuzzy control rule for the current working condition, then correct the existing fuzzy control rule parameters based on the control effect of the historical similar working condition;

[0037] C5: Call the pre-constructed water quality prediction model to obtain the water quality change trend in a preset time period;

[0038] C6: Convert the output fuzzy set into a control quantity by using the barycentric method, the conversion logic is: the control quantity is equal to the sum of the product of each predicted value in the fuzzy set and its membership value, divided by the sum of all membership values, and then multiplied by a weight factor; the control quantity is used to adjust the set value of the feed water flow; the weight factor is dynamically adjusted according to the water quality change trend output by the water quality prediction model;

[0039] C7: Apply the obtained control quantity to the feed water system, monitor the response of the system to the control quantity, collect the actual feed water flow data in real time, compare the actual feed water flow data with the corrected predicted value, calculate the error and error change rate, and use the error and error change rate to automatically adjust the position of the vertex of the membership function, the fuzzy control rule library and the weight factor.

[0040] The application also provides an intelligent coordinated control method based on a double-reheat unit, comprising:

[0041] Step S1: Collect the operation data of the boiler and the steam turbine in real time through a sensor, pretreat the collected operation data, combine the physical mechanism of the thermal system and the ST-Attention-LSTM model, construct a mechanism-data double-driven dynamic response prediction model of the thermal system, and use the mechanism-data double-driven dynamic response prediction model of the thermal system to perform multivariate prediction on the operation state of the boiler and the steam turbine, set the initial value of the control parameter of the boiler and the steam turbine according to the multivariate prediction result;

[0042] Step S2: Set the operation constraint condition of the boiler and the steam turbine, monitor and classify the system disturbance information in real time through K-means clustering, and calculate the weighted compensation amount of the prediction error based on the dynamic response prediction model of the thermal system and the disturbance influence factor, add the compensation amount to the control parameter, correct the prediction error through a feedback correction mechanism, and obtain the corrected control parameter;

[0043] Step S3: Use a rolling optimization algorithm to perform real-time dynamic adjustment on the corrected control parameter of the boiler and the steam turbine, combine the working condition-rule mapping update mechanism and the water quality prediction model, use an adaptive fuzzy control strategy to automatically adjust the set value of the feed water flow, and obtain the best efficiency of the stable operation of the boiler;

[0044] Step S4: Update the coal yard inventory by unattended and real-time coal inventory methods and visual monitoring, set up automatic repair mechanism, automatic start-stop mechanism and alarm mechanism, if abnormality or failure is found, trigger the alarm mechanism immediately, and automatically adjust the start-stop state of the boiler and steam turbine for repair;

[0045] Step S5: Real-time monitoring of carbon emissions of the boiler and steam turbine, and formulating an optimized control strategy according to the carbon emissions, periodically evaluating the energy use of the boiler and steam turbine, identifying energy-saving potential according to the evaluation results, and formulating energy-saving measures.

[0046] The application also provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the intelligent coordinated control method based on a double-reheat unit when executing the computer program.

[0047] Compared with the prior art, the application has the following beneficial effects:

[0048] 1. The application proposes an intelligent coordinated control system based on a double-reheat unit, and optimizes and improves the architecture, operation steps and flow, and the system has the advantages of simple flow, low investment and operation cost, and low production cost.

[0049] 2. The application proposes an intelligent coordinated control method based on a double-reheat unit, which can ensure stable operation of the system, reduce failure and downtime, automatically adjust the set value of the feed water flow through the adaptive fuzzy control strategy, ensure stable operation of the boiler and optimize energy utilization efficiency, reduce energy consumption, monitor carbon emissions in real time and formulate an optimized control strategy to reduce environmental pollution, and through unattended, real-time coal inventory and visual monitoring means, the coal yard inventory can be mastered in real time, and the management efficiency is improved; the application of the rolling optimization algorithm and the adaptive control strategy makes the system adapt to different working conditions and changes, improves the adaptability and flexibility of the system. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 It is an architecture diagram of the intelligent coordinated control system based on a double-reheat unit of the application;

[0051] Figure 2 It is a working flow chart of the prediction control unit of the intelligent coordinated control system based on a double-reheat unit of the application;

[0052] Figure 3 It is a working flow chart of the intelligent feedforward unit of the intelligent coordinated control system based on a double-reheat unit of the application;

[0053] Figure 4The working flow chart of the feedwater control unit of the intelligent coordinated control system based on a double-reheat unit of the application;

[0054] Figure 5 The flow chart of the intelligent coordinated control method based on a double-reheat unit of the application. DETAILED DESCRIPTION

[0055] Embodiment 1

[0056] Please refer to Figure 1 An embodiment provided by the application is an intelligent coordinated control system based on a double-reheat unit, which comprises an intelligent control module, a coordinated control module, and an auxiliary control module.

[0057] The intelligent control module is used for real-time analysis of the operation data of the boiler and the steam turbine and real-time optimization control of the operation process of the boiler and the steam turbine. The intelligent control module comprises a predictive control unit and an intelligent feedforward unit. The predictive control unit is configured with a multivariable predictive control strategy. The multivariable predictive control strategy is used for predicting the real-time operation state of the boiler and the steam turbine by using a pre-constructed mechanism-data double-driven thermal system dynamic response prediction model according to historical data, real-time operation data, and the physical mechanism of the thermal system, and automatically adjusting the control parameters by using a rolling optimization algorithm based on the prediction result. The mechanism-data double-driven thermal system dynamic response prediction model is constructed based on an ST-Attention-LSTM model. The intelligent feedforward unit is configured with an intelligent feedforward control strategy. The intelligent feedforward control strategy is used for correcting the prediction result by K-means clustering.

[0058] The operation data of the boiler and the steam turbine comprise temperature and pressure data, flow data, combustion data, emission data, rotation speed and position data, electrical data, vibration and noise data, and thermal efficiency and efficiency data.

[0059] The control parameters comprise fuel flow, valve opening, temperature set point, and pressure set point. Adjusting the control parameters can affect the output of the system, such as energy consumption, output power, and system efficiency.

[0060] The disturbance information comprises fuel flow, steam pressure, temperature variation, equipment failure, and parameter variation.

[0061] The coordinated control module is used for coordinated control of the boiler and the steam turbine by the multivariable predictive control strategy and the intelligent feedforward control strategy of the intelligent control module. The coordinated control module comprises a boiler main control unit and a steam turbine main control unit. The boiler main control unit dynamically adjusts the valve opening and the boiler combustion rate by receiving the load instruction, optimizes the combustion process of the boiler, and matches the energy of the boiler with the output energy of the steam turbine. The steam turbine main control unit is used for receiving the load instruction and adjusting the steam admission amount, rotation speed, and power of the steam turbine according to the received load instruction.

[0062] An auxiliary control module is configured to manage a coal yard and control water supply; the auxiliary control module comprises a water supply control unit; an adaptive fuzzy control strategy is configured in the water supply control unit; the adaptive fuzzy control strategy is configured to automatically call or modify a fuzzy control rule library according to a preset working condition-rule mapping update mechanism, obtain a water quality change trend in advance in combination with a pre-constructed water quality prediction model, convert a fuzzy set into a control amount through a barycenter method, and automatically adjust a set value of water supply flow; the water quality prediction model is configured based on a GRU network.

[0063] The intelligent control module further comprises a neural network unit; the neural network unit models and controls a complex nonlinear system by simulating the working principle of a human brain neural network.

[0064] The coordinated control module further comprises a digital electro-hydraulic regulation unit, a feedback correction unit, a disturbance monitoring and compensation unit, and an energy balance unit; the digital electro-hydraulic regulation unit is configured to control turbine regulating valves, realize rapid response of load and stability of steam pressure; the feedback correction unit is configured to compensate for prediction errors; the disturbance monitoring and compensation unit is configured to monitor system disturbances in real time and calculate compensation amounts through an intelligent feedforward control algorithm; and the energy balance unit is configured to monitor energy flow in the unit in real time, ensure energy balance between the boiler and the turbine, and avoid energy waste.

[0065] The auxiliary control module further comprises a smart coal yard unit and a fault diagnosis and repair unit; the smart coal yard unit realizes real-time updating and visual monitoring of coal yard inventory through unattended and real-time coal inventorying; the automatic fault diagnosis and repair unit is configured to diagnose potential faults of the unit and repair them through real-time monitoring and data analysis, thereby improving the reliability of the unit; and the APS automatic start-stop unit is configured to automatically control the start and stop processes of the unit through preset programs and logic.

[0066] The intelligent coordinated control system based on a double-reheat unit further comprises a data processing module and an energy management module.

[0067] The data processing module utilizes data processing methods to process, analyze, and monitor unit operating data in real time. The data processing module includes: a data acquisition unit, a data analysis unit, an automatic repair unit, an automated start-up and shutdown unit, and a monitoring and alarm unit. The data acquisition unit uses sensors to collect real-time operating data from the boiler and turbine. The data analysis unit preprocesses and analyzes the collected data and optimizes control strategies based on the analysis results. The automatic repair unit sets up an automatic repair mechanism to automatically diagnose and repair faults or anomalies. The automated start-up and shutdown unit automatically adjusts the unit's start-up and shutdown status according to unit operating needs and energy market demands. The monitoring and alarm unit monitors the system's operating status in real time, promptly alarms when anomalies are detected, and takes appropriate measures.

[0068] The energy management module is used to manage energy and optimize energy conservation during unit operation. The energy management module includes: an energy efficiency assessment unit, a carbon emission monitoring unit, and an energy conservation measure formulation unit. The energy efficiency assessment unit is used to assess and analyze the unit's energy usage. The carbon emission monitoring unit is used to monitor the unit's carbon emissions in real time and reduce carbon emissions through optimized control strategies. The energy conservation measure formulation unit is used to formulate energy conservation measures based on the assessment results, such as optimizing combustion methods and improving thermal efficiency.

[0069] Example 2

[0070] Please see Figures 2-4 In this embodiment, a multivariate predictive control strategy is adopted in the predictive control unit. The specific steps include:

[0071] A1: Real-time acquisition of boiler and turbine operating data via sensors. The collected operational data is preprocessed, including... Indicates the first k One set of running data, k This indicates the amount of operational data collected; the operational data includes secondary reheat steam temperature, pressure, fuel flow rate, feedwater flow rate, turbine steam intake, and rotational speed.

[0072] A2: Combine the physical mechanism of the thermal system to determine the input features of the ST-Attention-LSTM model, initialize the weights and thresholds of the ST-Attention-LSTM model, and then use the preprocessed running data to train the ST-Attention-LSTM model to obtain a mechanism-data dual-driven dynamic response prediction model for the thermal system.

[0073] Further, the prediction control unit configures a mechanism-data double-driven multivariable prediction control strategy, the core of which is to construct an ST-Attention-LSTM, i.e., a space-time attention-long short-term memory network thermal system dynamic response prediction model, and the specific implementation logic is as follows:

[0074] (1) In combination with the physical mechanism of the double-reheat unit, such as the boiler heat balance equation, and historical operation data, the model input features and output targets are determined, wherein the input features include the secondary reheat steam temperature, the turbine inlet pressure, the fuel flow, and the feedwater flow;

[0075] (2) The ST-Attention-LSTM model includes an input layer, a space-time attention layer, an LSTM layer, and an output layer: the input layer receives the preprocessed operation data after the abnormal value elimination, normalization, and time sequence alignment processing; the space-time attention layer strengthens the time sequence correlation capture of key parameters by calculating the attention weights of each input feature; the LSTM layer memorizes long-term time sequence features through a gating unit to avoid the gradient disappearance problem of the traditional RNN; and the output layer outputs the prediction results of the unit operation state to obtain the constructed ST-Attention-LSTM model, wherein the gating unit includes an input gate, a forgetting gate, and an output gate.

[0076] (3) The constructed ST-Attention-LSTM model is trained.

[0077] A3: Set a sampling time T, based on the mechanism-data double-driven thermal system dynamic response prediction model, according to the current operation data, through the implicit layer calculation of the mechanism-data double-driven thermal system dynamic response prediction model, the prediction results of the operation state of the boiler and the turbine in the next sampling time T of the output layer are obtained , wherein, represents the k th predicted operation data;

[0078] A4: Set the initial values of the control parameters of the boiler and the turbine, set the target function, the operation constraint condition, and the initial parameters of the rolling optimization algorithm, calculate the target function value according to the prediction results, and check whether the target function value meets the operation constraint condition, if not, adjust the control parameters of the boiler and the turbine using the rolling optimization algorithm, and roll the time window forward by one sampling time, repeat step A4 to obtain new control parameters .

[0079] The training steps of the ST-Attention-LSTM model in A2 include:

[0080] A2.1: The pre-processed operation data is divided into training set, validation set and test set in the ratio of 7:2:1, and the operation data dimension includes time series dimension and feature dimension;

[0081] A2.2: The attention layer weight, LSTM layer weight and threshold value of the model are initialized, and the input of the attention layer is the hidden state of the LSTM layer;

[0082] A2.3: The mean square error of the model prediction value and the actual operation data is used as the loss function, and the ST-Attention-LSTM model is trained using the Adam optimizer;

[0083] A2.4: The generalization ability of the ST-Attention-LSTM model is evaluated using the test set, and if the mean square error exceeds the set threshold, the network structure of the ST-Attention-LSTM model is adjusted until the accuracy requirement is met.

[0084] The logic of the objective function in A4 is: the objective function value comprehensively reflects the optimization goals of three dimensions of boiler fuel cost, matching degree of boiler output power and demand power, and system operation efficiency, and the importance of the three dimensions is adjusted by weight factors, wherein the boiler fuel cost is determined by the fuel flow, the matching degree of the boiler output power and the demand power is determined by the difference between the two, and the system operation efficiency is calculated by the real-time operation parameters of the unit. Specifically, the objective function formula is:

[0085] ;

[0086] Operating constraints include:

[0087] ;

[0088] wherein, J represents the objective function value, represents the fuel cost used by the boiler, represents the fuel flow function, represents the boiler output power function, represents the demand power when the boiler is working, represents the system efficiency function, t represents the current time, , , represents the weight factor, represents the control parameters of the boiler and the steam turbine, and represent the minimum and maximum values of the control parameters, and represent the minimum and maximum values of the temperature, represents the based on temperature function, and Indicates the minimum and maximum pressure values. Indicates based on The pressure function, and This represents the minimum and maximum values ​​of the flow rate. Indicates based on The flow function.

[0089] The intelligent feedforward unit adopts an intelligent feedforward control strategy, the specific steps of which include:

[0090] B1: Real-time monitoring of system disturbance information via sensors The disturbance information includes fluctuations in fuel calorific value, grid load impacts, changes in water hardness, and equipment fault warning signals, among which, l Indicates the amount of perturbation information. Indicates the first l One disturbance message;

[0091] B2: K-means clustering is used to classify disturbance information in real time. Based on the classified disturbance information and the prediction results of the mechanism-data dual-driven dynamic response prediction model of the thermodynamic system, the prediction error of the original control parameters is obtained. The compensation amount of the prediction error is calculated using an intelligent feedforward control algorithm. The formula is as follows: ,in, Indicates prediction error The amount of compensation, The feedforward control function is represented here. K-means clustering is a prior art in this field and is not an inventive solution of this application, so it will not be described in detail here.

[0092] B3: Set the error threshold to H, and obtain the compensation amount. Add it to the original control parameters to obtain the control parameter error after adding the compensation. ;

[0093] like Then adjust the output variables according to the prediction results;

[0094] like Then, the rolling optimization algorithm is used to adjust the control parameters, and based on real-time operating status data, the feedback correction control algorithm is used to correct the prediction results to obtain the corrected prediction value. ,in, express t Time of the first i Corrected predicted values ​​of each output variable;

[0095] B4: Based on the corrected prediction results, readjust the control parameters;

[0096] B5: Repeat B1 to B4 until the system reaches a steady state, and after feedback correction, obtain the corrected control parameters and use them to drive the operation of the boiler and the steam turbine.

[0097] The optimization problem solved by the rolling optimization algorithm in B3 includes:

[0098]

[0099] wherein, represents the operating state of the system at time t represents the operating state of the system at time represents a dynamic response prediction model of the thermal system, represents the constraint condition value of the system at time represents a minimum value function, represents a time step. The logic of the feedback correction control algorithm in B3 is as follows: n The corrected predicted value of the i-th output variable at time t is equal to the original predicted value of the output variable plus the sum of the product of the feedback correction coefficient of the output variable for each prediction error and the corresponding prediction error, plus an additional correction term determined according to the real-time operating conditions of the system, wherein the feedback correction coefficient is obtained by training historical prediction error and actual deviation data, and the additional correction term is used to compensate for the deviation not covered by the model under extreme operating conditions. Specifically, the formula of the feedback correction control algorithm is as follows:

[0100] t

[0101]

[0102] wherein, represents t the corrected predicted value of the i-th output variable at time t, i represents the original predicted value of the i-th output variable at time t, t represents i the feedback correction coefficient of the i-th output variable for the j-th prediction error at time t, represents t the prediction error of the i-th output variable at time t, i represents j the additional correction term of the i-th output variable at time t. t i t i

[0103] ​​​​​​​​​​​The water supply control unit adopts an adaptive fuzzy control strategy, and the specific steps include:

[0104] C1: setting a water supply flow domain, dividing the low, moderate, and high three fuzzy sets within the water supply flow domain, and configuring a triangular membership function for each fuzzy set, wherein the vertex position of the membership function is calibrated according to historical optimal water supply flow data, and the water supply flow domain is determined according to the rated evaporation capacity of the boiler;

[0105] Wherein, the boundaries of the fuzzy set are usually determined based on experience, system requirements or expert knowledge, the domain of the water supply flow is set to [0, 100], and the water supply flow range [0, 30) is defined as low, [30, 70] is defined as moderate, and (70, 100] is defined as high.

[0106] In the low flow fuzzy set, the triangular membership function is defined as the function value being 1 when the water flow is 0, i.e. completely belonging to the low set, the function value being 0 when the flow is 30, i.e. not belonging to the low set, and passing through an intermediate point in the process of linearly decreasing;

[0107] In the moderate flow fuzzy set, the triangular membership function takes boundary values 0 and 1 at flow rates of 30 and 70 respectively, and passes through an intermediate point in the process of linearly changing;

[0108] In the high flow fuzzy set, the triangular membership function takes boundary value 1 at flow rate 100, i.e. completely belongs to the high set, the function value is 0 at flow rate 70, i.e. does not belong to the high set, and passes through an intermediate point in the process of linearly changing.

[0109] C2: mapping the corrected predicted value to the configured triangular membership function, and calculating the membership value of the corrected predicted value belonging to each fuzzy set;

[0110] Further, the specific steps of C2 include:

[0111] (1) Set the three vertices of the triangular membership function as b , s , q , wherein, s is the center point;

[0112] (2) According to the definition of the fuzzy control strategy, the boundaries of the low, moderate, and high three fuzzy sets are determined to be [0, 30), [30, 70], and (70, 100) respectively;

[0113] (3) Receive the corrected predicted value, and judge whether the corrected predicted value falls within the water flow domain;

[0114] If not in the water flow domain, the triangular membership value of the prediction value is 0;

[0115] If in the water flow domain, determine the range of the fuzzy set to which the prediction value belongs;

[0116] If not in the range of the fuzzy set, the triangular membership value of the prediction value is 0;

[0117] If in the range of the fuzzy set, according to the range of the fuzzy set to which it belongs, use the corresponding triangular membership function to calculate its membership, wherein the formula for calculating the triangular membership function is a prior art content in the field and is not the inventive scheme of the present application, and will not be described here.

[0118] C3: According to system requirements, formulate a fuzzy control rule base, combine the membership values in C2, and according to the fuzzy control rules in the fuzzy control rule base, use the maximum and minimum reasoning method to calculate the output fuzzy set, i.e., determine the fuzzy interval to which the feedwater flow should belong; the system requirements include water quality standard, steam temperature stability range, and unit load response speed, wherein the maximum and minimum reasoning method is a prior art content in the field and is not the inventive scheme of the present application, and will not be described here;

[0119] Further, the process of formulating the fuzzy control rule base mainly uses if-else logic to traverse all fuzzy set combinations of input variables, combines system requirement constraints to formulate the output rules corresponding to each combination, and ensures that there is no omission or conflict in the rules, wherein the if-else logic is a prior art content in the field and is not the inventive scheme of the present application, and will not be described here.

[0120] C4: Introduce a working condition-rule mapping update mechanism to identify the current working condition of the unit in real time, automatically call the corresponding fuzzy control rules from the fuzzy control rule base in C3 according to the working condition type, and if there is no matching fuzzy control rule for the current working condition, correct the existing fuzzy control rule parameters based on the control effect of the similar historical working conditions;

[0121] Further, the working condition-rule mapping update mechanism refers to a closed-loop optimization mechanism that establishes a dynamic association between the working condition type and the fuzzy control rule base by identifying the current running working condition of the unit in real time, and automatically corrects the rule parameters based on the similar historical working conditions when there is no matching rule, and the core purpose is to solve the problem that the traditional fuzzy control relies on fixed rules and is difficult to adapt to the complex and variable working conditions of the secondary reheat unit, such as load fluctuation, fuel switching, and water quality change, to ensure that the control rules are always synchronized with the real-time running state of the unit and improve the control accuracy of the feedwater flow and other key parameters.

[0122] C5: Call the pre-constructed water quality prediction model to obtain the water quality change trend in a preset time period;

[0123] Furthermore, the process of constructing a water quality prediction model includes:

[0124] (1) Determine the prediction target and accuracy requirements. The core objective of the model is to predict the future trend of key water quality parameters of the water supply system, so as to provide advance notice for water supply flow adjustment.

[0125] (2) Define the dimensions of input and output variables. The input variables include historical water quality parameters, operating parameters, raw water characteristic parameters, and environmental disturbance parameters. The output variables are the predicted values ​​of water hardness, silicon content, and conductivity every 10 minutes in the next hour.

[0126] (3) Deploy online water quality analyzers at the feedwater pump inlet and boiler inlet, read operating parameters from the unit DCS system, obtain raw water data from the water plant interface, obtain environmental data from the plant meteorological station, and perform preprocessing and dataset division.

[0127] (4) Load the GRU network architecture and train the GRU network architecture using the partitioned training set to obtain a trained water quality prediction model. The GRU network architecture is existing technology in this field and is not an inventive solution of this application. It will not be described in detail here.

[0128] C6: The centroid method is used to convert the output fuzzy set into a control quantity. The conversion logic is as follows: the control quantity is equal to the sum of the products of each predicted value and its membership value in the fuzzy set, divided by the sum of all membership values, and then multiplied by a weighting factor. The control quantity is used to adjust the setpoint of the water supply flow rate. The weighting factor is dynamically adjusted according to the water quality change trend output by the water quality prediction model. Specifically, the conversion formula is:

[0129] ;

[0130] in, This represents the converted control quantity. Represents the first in the fuzzy set i One predicted value, This represents the membership degree in the fuzzy set A. Represents the membership function. Indicates the weighting factor;

[0131] C7: Apply the obtained control quantity to the water supply system, monitor the system's response to the control quantity, collect actual water supply flow data in real time, compare the actual water supply flow data with the corrected predicted value, calculate the error and error change rate between the two, and use the error and error change rate to automatically determine the vertex position of the membership function, the fuzzy control rule base, and the weight factor.

[0132] Example 3

[0133] Please refer to Figure 5 In another embodiment, the application provides an intelligent coordination control method based on a double-reheat unit, comprising:

[0134] Step S1: Collect the operation data of the boiler and the steam turbine in real time through a sensor, pre-process the collected operation data, construct a mechanism-data dual-driven dynamic response prediction model of the thermal system by combining the physical mechanism of the thermal system and the ST-Attention-LSTM model, and use the mechanism-data dual-driven dynamic response prediction model of the thermal system to perform multivariate prediction on the operation state of the boiler and the steam turbine, set the initial value of the control parameter of the boiler and the steam turbine according to the multivariate prediction result, and set the initial value of the control parameter of the boiler and the steam turbine.

[0135] Step S2: Set the operation constraint condition of the boiler and the steam turbine, monitor and classify the system disturbance information in real time through K-means clustering, calculate the weighted compensation amount of the prediction error based on the dynamic response prediction model of the thermal system and the disturbance influence factor, add the compensation amount to the control parameter, correct the prediction error through a feedback correction mechanism, and obtain the corrected control parameter.

[0136] Step S3: Use a rolling optimization algorithm to dynamically adjust the control parameter of the boiler and the steam turbine in real time, combine a working condition-rule mapping update mechanism and a water quality prediction model self-adaptive fuzzy control strategy to automatically adjust the set value of the feedwater flow, and obtain the best efficiency of the stable operation of the boiler.

[0137] Step S4: Update the coal yard inventory through unattended and real-time coal inventory updating and visual monitoring, set an automatic repair mechanism, an automatic start-stop mechanism and an alarm mechanism, trigger the alarm mechanism immediately if an abnormality or a fault is found, and automatically adjust the start-stop state of the boiler and the steam turbine for repair.

[0138] Step S5: Monitor the carbon emission of the boiler and the steam turbine in real time, develop an optimization control strategy according to the carbon emission, evaluate the energy use of the boiler and the steam turbine at regular intervals, identify the energy-saving potential according to the evaluation result, and develop energy-saving measures.

[0139] Embodiment 4

[0140] An electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the intelligent coordination control method based on a double-reheat unit when executing the computer program.

[0141] The embodiments of the present application are described above with reference to the accompanying drawings, but the present application is not limited to the above-described specific embodiments, and the above-described specific embodiments are merely illustrative, but not restrictive, and a person of ordinary skill in the art can make changes, modifications, replacements and variations to the above-described embodiments without departing from the purpose of the present application and the scope protected by the claims, and these are all within the protection of the present application.

Claims

1. An intelligent coordinated control system based on a double reheat unit, characterized in that, include: Intelligent control module, coordinated control module, auxiliary control module; The intelligent control module includes a predictive control unit and an intelligent feedforward unit. The predictive control unit is configured with a multivariate predictive control strategy, which predicts the real-time operating status of the boiler and turbine using a pre-built mechanism-data dual-driven dynamic response prediction model based on historical data, real-time operating data, and the physical mechanism of the thermal system. The mechanism-data dual-driven dynamic response prediction model is built based on an ST-Attention-LSTM model. The intelligent feedforward unit is configured with an intelligent feedforward control strategy. The intelligent feedforward control strategy is used to correct the prediction results through K-means clustering; The coordination control module includes a boiler main control unit and a turbine main control unit. The boiler main control unit dynamically adjusts the valve opening and boiler combustion rate based on the received load commands to match the boiler's energy with the turbine's output energy. The turbine main control unit receives load commands and adjusts the turbine's steam intake, speed, and power according to the received load commands. The auxiliary control module includes a water supply control unit; the water supply control unit is equipped with an adaptive fuzzy control strategy; the adaptive fuzzy control strategy is used to automatically call or correct the fuzzy control rule library according to the preset working condition-rule mapping update mechanism, combine with the pre-built water quality prediction model to obtain the water quality change trend in advance, convert the fuzzy set into the control quantity through the centroid method, and automatically adjust the set value of the water supply flow rate; the water quality prediction model is based on the GRU network configuration; The predictive control unit employs a multivariate predictive control strategy, and the specific steps include: A1: Real-time acquisition of boiler and turbine operating data via sensors. The collected operational data is preprocessed, including... Indicates the first k One set of running data, k This indicates the amount of operational data collected; the operational data includes secondary reheat steam temperature, pressure, fuel flow rate, feedwater flow rate, turbine steam intake, and rotational speed. A2: Combine the physical mechanism of the thermal system to determine the input features of the ST-Attention-LSTM model, initialize the weights and thresholds of the ST-Attention-LSTM model, and then use the preprocessed running data to train the ST-Attention-LSTM model to obtain a mechanism-data dual-driven dynamic response prediction model for the thermal system. A3: Set the sampling time T. Based on the mechanism-data dual-driven dynamic response prediction model of the thermodynamic system, and according to the current operating data, the hidden layer calculation of the mechanism-data dual-driven dynamic response prediction model of the thermodynamic system yields the prediction results of the boiler and turbine operating status within the next sampling time T in the output layer. ,in, Indicates the first k The predicted running data; A4: Set the initial values ​​of the control parameters for the boiler and turbine, define the objective function, operating constraints, and initial parameters for the rolling optimization algorithm. Calculate the objective function value based on the prediction results and check if the objective function value meets the operating constraints. If not, use the rolling optimization algorithm to adjust the control parameters of the boiler and turbine, and roll the time window forward by one sampling time. Repeat step A4 to obtain the new control parameters. ; The training steps of the ST-Attention-LSTM model include: A2.1: Divide the preprocessed running data into training set, validation set and test set in a ratio of 7:2:

1. The dimensions of the running data include time series dimension and feature dimension. A2.2: Initialize the attention layer weights, LSTM layer weights, and thresholds of the model. The input to the attention layer is the hidden state of the LSTM layer. A2.3: Using the mean squared error between the model's predicted values ​​and the actual running data as the loss function, train the ST-Attention-LSTM model using the Adam optimizer; A2.4: Use the test set to evaluate the generalization ability of the ST-Attention-LSTM model. If the mean squared error exceeds the set threshold, readjust the network structure of the ST-Attention-LSTM model until the accuracy requirements are met.

2. The intelligent coordinated control system based on a double reheat unit as described in claim 1, characterized in that, The logic of the objective function in A4 is as follows: the objective function value comprehensively reflects the optimization objectives of three dimensions: boiler fuel cost, matching degree between boiler output power and demand power, and system operating efficiency. The importance of the three dimensions is adjusted by weighting factors. Among them, boiler fuel cost is determined by fuel flow rate, matching degree between boiler output power and demand power is determined by the difference between the two, and system operating efficiency is calculated from the real-time operating parameters of the unit.

3. The intelligent coordinated control system based on a double reheat unit as described in claim 2, characterized in that, The intelligent feedforward unit adopts an intelligent feedforward control strategy, and the specific steps include: B1: Real-time monitoring of system disturbance information via sensors The disturbance information includes fluctuations in fuel calorific value, grid load impacts, changes in water hardness, and equipment fault warning signals, among which, l Indicates the amount of perturbation information. Indicates the first l One disturbance message; B2: K-means clustering is used to classify disturbance information in real time. Based on the classified disturbance information and the prediction results of the mechanism-data dual-driven dynamic response prediction model of the thermodynamic system, the prediction error of the original control parameters is obtained. The compensation amount of the prediction error is calculated using an intelligent feedforward control algorithm. The formula is as follows: ,in, Indicates prediction error The amount of compensation, Represents the feedforward control function; B3: Set the error threshold to H, and obtain the compensation amount. Add it to the original control parameters to obtain the control parameter error after adding the compensation. ; like Then adjust the output variables according to the prediction results; like Then, the rolling optimization algorithm is used to adjust the control parameters, and based on real-time operating status data, the feedback correction control algorithm is used to correct the prediction results to obtain the corrected prediction value. ,in, express t Time of the first i Corrected predicted values ​​of each output variable; B4: Based on the corrected prediction results, readjust the control parameters; B5: Repeat B1 to B4 until the system reaches a steady state. After feedback correction, the corrected control parameters are obtained and used to drive the operation of the boiler and turbine.

4. The intelligent coordinated control system based on a double reheat unit as described in claim 3, characterized in that, The logic of the feedback correction control algorithm in B3 is as follows: t The corrected predicted value of the i-th output variable at time i is equal to the original predicted value of the output variable plus the sum of the products of the feedback correction coefficient of the output variable for each prediction error and the corresponding prediction error, plus an additional correction term determined according to the real-time operating conditions of the system. The feedback correction coefficient is obtained by training with historical prediction error and actual deviation data, and the additional correction term is used to compensate for deviations not covered by the model under extreme operating conditions.

5. The intelligent coordinated control system based on a double reheat unit as described in claim 4, characterized in that, The water supply control unit adopts an adaptive fuzzy control strategy, and the specific steps include: C1: Set the feedwater flow domain, divide the feedwater flow domain into three fuzzy sets: low, moderate, and high, and configure a triangular membership function for each fuzzy set. The vertex position of the membership function is calibrated based on the historical best feedwater flow data, and the feedwater flow domain is determined based on the boiler's rated evaporation capacity. C2: Corrected predicted values Mapping onto the configured triangular membership function, calculate the membership value of the corrected predicted value belonging to each fuzzy set; C3: Based on the system requirements, formulate a fuzzy control rule base. Combined with the membership values ​​in C2, and based on the fuzzy control rules in the fuzzy control rule base, use the max-min inference method to calculate the output fuzzy set, that is, determine the fuzzy interval to which the water supply flow should belong. The system requirements include water quality compliance standards, steam temperature stability range, and unit load response speed.

6. The intelligent coordinated control system based on a double reheat unit as described in claim 5, characterized in that, The water supply control unit adopts an adaptive fuzzy control strategy, and the specific steps also include: C4: Introduces a condition-rule mapping update mechanism to identify the current operating condition of the unit in real time. Based on the operating condition type, it automatically calls the corresponding fuzzy control rule from the fuzzy control rule library in C3. If there is no matching fuzzy control rule for the current operating condition, it corrects the existing fuzzy control rule parameters based on the control effect of similar historical operating conditions. C5: Call the pre-built water quality prediction model to obtain the water quality change trend within a preset time period; C6: The centroid method is used to convert the output fuzzy set into a control quantity. The conversion logic is as follows: the control quantity is equal to the sum of the products of each predicted value and its membership value in the fuzzy set, divided by the sum of all membership values, and then multiplied by the weighting factor. The control quantity is used to adjust the set value of the water supply flow rate. The weighting factor is dynamically adjusted according to the water quality change trend output by the water quality prediction model. C7: Apply the obtained control quantity to the water supply system, monitor the system's response to the control quantity, collect actual water supply flow data in real time, compare the actual water supply flow data with the corrected predicted value, calculate the error and error change rate between the two, and use the error and error change rate to automatically determine the vertex position of the membership function, the fuzzy control rule base, and the weight factor.

7. A method for intelligent coordinated control of a double reheat unit, implemented based on the intelligent coordinated control system of a double reheat unit as described in any one of claims 1-6, characterized in that, include: Step S1: Real-time acquisition of boiler and turbine operating data through sensors, preprocessing of the acquired operating data, construction of a mechanism-data dual-driven dynamic response prediction model of the thermal system by combining the physical mechanism of the thermal system and the ST-Attention-LSTM model, and multivariate prediction of the operating status of the boiler and turbine by using the mechanism-data dual-driven dynamic response prediction model of the thermal system. Based on the multivariate prediction results, initial values ​​of control parameters of the boiler and turbine are set. Step S2: Set the operating constraints of the boiler and turbine, monitor and classify system disturbance information in real time through K-means clustering, and calculate the weighted compensation amount of the prediction error based on the dynamic response prediction model of the thermal system and the disturbance influence factor. Add the compensation amount to the control parameters, correct the prediction error through the feedback correction mechanism, and obtain the corrected control parameters. Step S3: Use the rolling optimization algorithm to dynamically adjust the control parameters of the corrected boiler and turbine in real time. Combine the operating condition-rule mapping update mechanism and water quality prediction model, and adopt an adaptive fuzzy control strategy to automatically adjust the set value of the feedwater flow rate to obtain the best efficiency for stable boiler operation. Step S4: Update the coal yard inventory through unattended operation and real-time coal inventory and visualize the monitoring. Set up an automatic repair mechanism, an automatic start-up and shutdown mechanism and an alarm mechanism. If an abnormality or fault is detected, the alarm mechanism will be triggered immediately and the start-up and shutdown status of the boiler and steam turbine will be automatically adjusted for repair. Step S5: Monitor the carbon emissions of the boiler and steam turbine in real time, formulate optimized control strategies based on the carbon emissions, regularly assess the energy use of the boiler and steam turbine, identify energy-saving potential based on the assessment results, and formulate energy-saving measures.

8. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the intelligent coordinated control method based on a double reheat unit as described in claim 7.

Images