Remote operation and maintenance control method for waste heat power generation in variable working condition mode

Abstract

The invention discloses a remote operation and maintenance control method for a waste heat power generation variable working condition mode, and the method is based on a flue gas and air system, a flue gas and air turbine coordination model and a turbine system, and the flue gas and air turbine coordination model converts steam parameters outputted by the flue gas and air system into steam parameters inputted by the turbine system; the flue gas system comprises a first evaporation model, a second evaporation model and a steam pocket unit; the method comprises the steps of collecting historical data of a waste heat power station and constructing a flue gas and air system according to the following formula; the flue gas and wind turbine coordination model converts steam parameters output by a flue gas and wind system according to the following formula to obtain turbine steam parameters; constructing a steam turbine system according to the steam turbine steam parameters; extracting flue gas and air system data through a convolutional neural network to construct a flue gas change feature vector and a steam-water change feature vector; establishing a flue gas and wind turbine coordination prediction model according to the flue gas and wind change feature vector and the steam and water change feature vector; performing optimization training on the flue gas turbine coordination prediction model through a long short-term memory neural network; stable power generation of the waste heat furnace is guaranteed.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence waste heat power generation control, and in particular relates to a remote operation and maintenance control method for waste heat power generation under a variable operating mode. Background Art

[0002] In the existing waste heat power plants of cement lines, most of them rely on the operating personnel's own work experience to adjust the control parameters of the flue gas system and the steam-water system to ensure the operation of the cement waste heat power plant. Since the operating personnel of the power plant are relatively fixed, the historical operating data that can be learned and used is relatively limited, and it is difficult to ensure the stable operation of the cement waste heat power plant. The waste heat power plant operates under three different operating conditions (stable operating condition operation, variable operating condition operation, and abnormal operation). The most difficult to control and adjust is the adjustment and setting of system parameters under variable operating conditions. How to efficiently and accurately ensure the stable operation and maintenance of the waste heat furnace under variable operating conditions is a technical problem that technical personnel in this field need to solve. Summary of the invention

[0003] In response to the existing technical difficulties, the present invention provides a remote operation and maintenance control method for waste heat power generation under variable operating mode. The method solves the technical problem that power station boiler operators rely on experience to adjust boiler operating parameters and subjective judgment of flue gas parameter fluctuations to ensure stable power generation of the waste heat boiler.

[0004] In order to more effectively adjust the boiler and steam turbine parameters and improve the power generation efficiency of the power station, the present invention proposes to use convolutional neural network CNN and long short-term memory neural network LSTM model to read the characteristic parameters of the on-site boiler parameters and steam turbine parameter time series, and predict the turbine system parameters and generator parameters N minutes after the flue gas parameters change, so as to provide a reference for effectively adjusting the smoke and air system and the steam turbine system.

[0005] The present invention is implemented by adopting the following technical scheme:

[0006] A remote operation and maintenance control method for waste heat power generation under variable operating mode, the method is based on a smoke and air system, a smoke and air turbine coordination model and a turbine system, the smoke and air turbine coordination model converts the steam parameters output by the smoke and air system into steam parameters input by the turbine system, so the smoke system includes a first evaporation model, a second evaporation model and a steam drum unit; wherein: the method includes the following steps:

[0007] Collect historical data of waste heat power station and build smoke and wind system according to the following formula;

[0008]

[0009] Where: C g Output matrix; D g is the direct transfer matrix; x g1 Flue gas temperature; xg2 Flue gas flow rate; x g3 Flue gas pressure; g1 Boiler outlet temperature T vgout； y g2 Steam outlet flow rate Q vgout ;y g3 is the boiler outlet steam pressure p fo ;u g1 is the boiler inlet temperature T vgin ;u g2 is the flue gas inlet flow rate Q vgin ;u g3 is the boiler inlet air pressure p fi ;

[0010] The smoke-air-steam-turbine coordination model converts the steam parameters output by the smoke-air system into steam parameters of the steam turbine according to the following formula;

[0011]

[0012]

[0013] Where: h ps is the specific enthalpy of superheated steam;

[0014]

[0015] According to the steam parameters of the steam turbine, the steam turbine system is constructed according to the following formula:

[0016]

[0017] Among them, C s Output matrix of steam turbine system; D s Direct transfer matrix; x s1 Water flow rate Q w1 ;x s2 Feed water temperature T w1 ;x s3 Water supply pressure p w ;u s1 Steam flow Q b1 ;u s2 Steam temperature T b1 ;u s3 Steam pressure p b ;y s1 Actual output power of steam turbine; s2 Turbine speed;

[0018] The smoke and wind system data is extracted through convolutional neural network to construct the smoke and wind change feature vector;

[0019]

[0020] Where: Q vgin is the flue gas inlet flow rate, Q vgout is the steam outlet flow rate; p fi is the boiler inlet air pressure, p fo is the boiler outlet steam pressure, T vgin is the boiler inlet temperature, T vgout The boiler outlet temperature is extracted through the convolutional neural network to construct the steam-water change feature vector, which is expressed as:

[0021]

[0022] Where: p b Steam pressure, p w Feed water pressure, T b1 Steam temperature, T w1 Feed water temperature, Q b1 Steam flow, Q w1 Water flow rate;

[0023] The smoke, wind and steam turbine coordination prediction model is established based on the following formula based on the smoke and wind change characteristic vector and the steam and water change characteristic vector:

[0024]

[0025] Among them, A sg , A gs represents the coupling matrix between the smoke and air system and the steam turbine system; B gs , B sg Input coupling matrix of smoke and air system and steam turbine system; x gs1 Indicates the enthalpy of change of feed water flow rate h pe ;x gs2 It represents the specific enthalpy of superheated steam h ps ;x gs3 Indicates valve opening u; u gs1 Indicates the steam temperature at the drum outlet c v ;u gs2 Indicates the steam mass flow rate D at the outlet of the flue gas system drum do ;u gs3 Indicates the steam pressure p at the outlet of the flue gas system drum dr ;

[0026] The flue gas turbine coordination prediction model is optimized and trained through long short-term memory neural network;

[0027] The parameter changes of the flue gas coordinated control prediction model are optimized through the following formula, thereby achieving optimal adjustment and control of the power plant operation parameters:

[0028]

[0029] Among them, C g,C gs ,C sg ,C s Coefficient matrix of smoke, wind and steam turbine system; D g , D gs ,D sg ,D s Smoke, air and steam turbine system direct transmission matrix; x gs1 Indicates the enthalpy of change of feed water flow rate h pe ;x gs2 It represents the specific enthalpy of superheated steam h ps ;x gs3 Indicates valve opening u; u gs1 Indicates the steam temperature at the drum outlet c v ;u gs2 Indicates the steam mass flow rate D at the outlet of the flue gas system drum do ;u gs3 Indicates the steam pressure p at the outlet of the flue gas system drum dr ;y s1 Steam temperature T b1 ;y s2 Steam flow Q b1 ;y s3 Steam pressure p b .

[0030] Furthermore, the flue gas turbine coordination prediction model is optimized and trained through a long short-term memory neural network; including:

[0031] The smoke, air and steam turbine coordination prediction model is optimized by the following formula:

[0032]

[0033] Where: y i for y s1 Steam temperature T b1 ,y s2 Steam flow Q b1 ,y s3 Steam pressure p b The actual value of the sample space; for y s1 Steam temperature T b1 ,y s2 Steam flow Q b1 ,y s3 Steam pressure p b Prediction estimate; n is the dimensional sample space; is the actual measurement space vector; is the prediction space vector;

[0034] After training, the optimized smoke, wind and steam turbine coordination prediction model is obtained as follows:

[0035]

[0036] Where: d is the prediction step length; y 1 Steam temperature T b1 ,y 2 Steam flow Q b1 ,y 3 Steam pressure p b Output vector for the prediction model; x 1 Feedwater flow rate change enthalpy h pe 、x 2 Specific enthalpy of superheated steam h ps 、x 3 The valve opening u is the input vector of the prediction model; f d Build an LSTM model to predict the current d moments.

[0037] Furthermore, the process of constructing a smoke and air system through the first evaporation model, the second evaporation model and the drum unit includes:

[0038] The first evaporation model is constructed according to the following formula based on the heat source heat transfer absorption energy and the inlet energy;

[0039]

[0040] Where: Q represents the amount of heat absorbed; D represents the mass flow rate; D represents the volume; M represents the mass; C pe represents the specific heat of metal at constant pressure in the evaporation system; T represents temperature; h represents specific enthalpy; ρ represents density; superscript: ' represents saturated water; * represents saturated steam; subscript: d represents steam drum; f represents boiler feed water; e represents evaporation system; m represents metal; s represents saturated state;

[0041] The second evaporation model is constructed according to the following formula based on the energy absorbed by the heat source, the energy brought into the inlet, and the energy output by the first evaporation model;

[0042]

[0043] Constructing a drum unit according to the parameters output by the second evaporation model according to the following formula;

[0044]

[0045] Where: u represents the valve opening, c v is the steam temperature at the drum outlet; subscript: v represents the turbine control valve, out represents the outlet;

[0046] When the steam flow rate at the drum outlet, the drum pressure, the control valve opening and the pressure behind the valve are known, the drum outlet flow rate is:

[0047]

[0048] Where: c v D is the steam temperature at the drum outlet; do u is the steam mass flow rate at the outlet of the flue gas system drum; v Steam turbine valve opening; p dr Steam pressure at the outlet of flue gas system drum; p v,out Steam pressure at turbine inlet;

[0049] The flue gas system is obtained by calculating the first evaporation model, the second evaporation model and the drum unit according to the following formula;

[0050]

[0051] According to the flue gas system, the enthalpy of change of water flow rate is obtained according to the following formula: pe ;

[0052]

[0053] Beneficial Effects

[0054] Compared with the traditional technical solution, the beneficial effects brought by the present invention are:

[0055] The present invention adopts convolutional neural network CNN and long short-term memory neural network LSTM models, combined with the characteristic parameters of reading the time series of on-site boiler parameters and steam turbine parameters, to predict the turbine system parameters and generator parameters N minutes after the flue gas parameters change, so as to provide a reference basis for effectively adjusting the smoke and air system and the steam turbine system, more effectively adjust the boiler and steam turbine parameters, improve the power generation efficiency of the power station, and promote production. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 This is a flow chart of a remote operation and maintenance control method for waste heat power generation under variable operating mode of the present invention DETAILED DESCRIPTION

[0057] The following is combined with Figure 1 The present invention is described as follows:

[0058] The present invention provides a remote operation and maintenance control method for waste heat power generation in a variable operating mode. The method collects the flue gas system 101, the smoke-air-steam turbine coordinated control model 202 and the steam turbine system 201. The smoke-air-steam turbine coordinated control model 202 utilizes the advantages of CNN in feature extraction and the advantages of LSTM network in time series data processing, and proposes a collaborative control prediction model of the boiler system and the steam turbine system based on CNN+LSTM. Among them: the convolutional neural network model extracts the characteristics of the boiler system, the smoke-air system, and the steam-water system from the changes in the heat source parameters (smoke-air temperature, flow rate, wind speed), combined with the physical model of the boiler. The LSTM method is used to train the parameters of the collaborative control prediction model of the smoke-air system and the steam turbine system using the historical data of the operation and operation of the smoke-air system and the steam turbine system, optimize the delay characteristics of the system extracted from the prediction model, optimize the parameters of the prediction models of the two systems, and realize the coordinated control of the parameters of the smoke-air system and the steam-water system to ensure that the entire system operates stably following the fluctuations of the heat source. Specifically including; such as Figure 1 As shown,

[0059] The method is based on the smoke and air system 101, the smoke and air turbine coordinated control model 202 and the turbine system 201. The smoke and air turbine coordinated control model 202 converts the steam parameters output by the smoke and air system 101 into the steam parameters input by the turbine system 201. Therefore, the smoke and air system 101 also includes a first evaporation model 103, a second evaporation model 104 and a steam drum unit.

[0060] Collect historical data of waste heat power station and build smoke and wind system according to the following formula;

[0061]

[0062] Among them, C g Output matrix; D g is the direct transfer matrix; y g1 Flue gas outlet temperature; y g2 Total flue gas flow; wherein: the flue gas system includes a first evaporation model, a second evaporation model and a steam drum unit;

[0063] The first evaporation model is constructed according to the following formula based on the heat source heat transfer absorption energy and the inlet energy;

[0064]

[0065] Where Q represents the amount of heat absorbed; D represents the mass flow rate; D represents the volume; M represents the mass; C pe represents the specific heat of metal at constant pressure in the evaporation system; T represents temperature; h represents specific enthalpy; ρ represents density; superscript: ' represents saturated water; * represents saturated steam; subscript: d represents steam drum; f represents boiler feed water; e represents evaporation system; m represents metal; s represents saturated state;

[0066] The second evaporation model is constructed according to the following formula based on the heat source heat transfer absorption energy, the inlet energy and the first evaporation model;

[0067]

[0068] Constructing a drum unit according to the parameters output by the second evaporation model according to the following formula;

[0069]

[0070] Where: u represents the valve opening, c v is the steam temperature at the drum outlet; subscript: v represents the turbine control valve, out represents the outlet;

[0071] When the steam flow rate at the drum outlet, the drum pressure, the control valve opening and the pressure behind the valve are known, the drum outlet flow rate is:

[0072]

[0073] Where: c v D is the steam temperature at the drum outlet; do u is the steam mass flow rate at the outlet of the flue gas system drum; v Steam turbine valve opening; p dr Steam pressure at the outlet of flue gas system drum; p v,out Steam pressure at turbine inlet;

[0074] The flue gas system is obtained by calculating the first evaporation model, the second evaporation model and the drum unit according to the following formula;

[0075]

[0076] According to the flue gas system, the enthalpy of change of water flow rate is obtained according to the following formula: pe ;

[0077]

[0078] The smoke-air-steam-turbine coordination model converts the steam parameters output by the smoke-air system into steam parameters of the steam turbine according to the following formula;

[0079]

[0080] Where: h ps is the specific enthalpy of superheated steam;

[0081]

[0082] According to the steam parameters of the steam turbine, the steam turbine system is constructed according to the following formula:

[0083]

[0084] Among them, C s Output matrix of steam turbine system; D s Direct transfer matrix; y s1 Actual output power of steam turbine; s2 Turbine speed.

[0085] The smoke and wind system data is extracted through convolutional neural network to construct the smoke change feature vector;

[0086] (1) Data preparation

[0087] The Modbus communication protocol is used to collect and read the operating data of the waste heat power plant, including flue gas inlet flow, flue gas outlet flow, boiler inlet air pressure, boiler outlet air pressure, boiler inlet temperature, boiler outlet temperature, steam pressure, feed water pressure, steam temperature, feed water temperature, steam flow, feed water flow, steam drum liquid level and other parameters, which constitute the basic data set for subsequent feature extraction.

[0088] Perform data cleaning on the basic data set to remove outliers, duplicate data, and erroneous or missing data.

[0089] The cleaned data is normalized so that it falls within the interval [0,1] to facilitate model training.

[0090] (2) Building the AlexNet model architecture

[0091] Input layer: The preprocessed waste heat power plant parameter data is passed into the model as input. The input data is set to [batch_size,1,1,num_features], where batch_size represents the number of data samples and num_features represents the number of waste heat power plant parameters.

[0092] 1. First layer model

[0093] Convolutional layer 1: Contains several convolution kernels to extract local features of the input data. It uses an 8x8 convolution kernel, a step size of 1, and a padding of 2 to perform convolution operations on the input data. After the first layer of convolution, the shape of the output feature map becomes [batch_size, 64, 1, num_features].

[0094] Activation function 1: Apply the activation function ReLU function to the output of convolution layer 1 to introduce nonlinear factors and enhance the expressiveness of the model. The shape of the feature map after activation is [batch_size, 64, 1, num_features]. The expression of the ReLU function is f(x) = max(0, x).

[0095] Pooling layer 1: A maximum pooling operation of size 2x2 is used with a stride of 2 to downsample the activated feature map. After pooling, the shape of the output feature map becomes [batch_size, 64, 1, num_features / 2].

[0096] 2. Layer 2 Model

[0097] Convolutional layer 2: Similar to convolutional layer 1, it uses 128 convolution kernels of size 3x3, a stride of 1, and a padding of 1 to perform convolution operations on the output feature map of pooling layer 1 to further extract features. The shape of the output feature map becomes [batch_size, 128, 1, num_features / 2].

[0098] Activation function 2: ReLU activation function is applied again to enhance the nonlinear expression, and the shape of the activated feature map is [batch_size, 128, 1, num_features / 2].

[0099] Pooling layer 2: Use a maximum pooling operation of size 2x2 with a step size of 2 for downsampling, and the shape of the output feature map becomes [batch_size, 128, 1, num_features / 4].

[0100] 3. Layer 3 Model

[0101] Convolutional layer 3: Use 256 convolution kernels of size 3x3, stride 1, padding 1, and convolve the output feature map of pooling layer 2. The shape of the output feature map is [batch_size, 256, 1, num_features / 4].

[0102] Activation function 3: Apply the ReLU activation function, and the shape of the activated feature map is [batch_size, 256, 1, num_features / 4].

[0103] 4. Layer 4 Model

[0104] Convolutional layer 4: Continue to use 256 convolution kernels of size 3x3, with a stride of 1 and a padding of 1 to perform convolution operations on the activated feature map to further extract features. The shape of the output feature map becomes [batch_size, 256, 1, num_features / 4].

[0105] Activation function 4: Add the ReLU activation function, and the shape of the feature map after activation is [batch_size, 256, 1, num_features / 4].

[0106] Pooling layer 3: Through the maximum pooling operation of size 2x2 and the stride of 2, the activated feature map is downsampled and the shape of the output feature map becomes [batch_size, 256, 1, num_features / 8].

[0107] Fully connected layer 1: Flatten the output feature map of pooling layer 3 into a one-dimensional vector and then connect it to the fully connected layer. For example, using a fully connected layer with 4096 neurons, the flattened feature vector is linearly transformed, and the shape of the output vector becomes [batch_size, 4096].

[0108] Activation function 5: Add the ReLU activation function after the fully connected layer 1, and the shape of the output vector after activation is [batch_size, 4096].

[0109] Dropout layer: To prevent overfitting, the dropout rate of the Dropout layer is set to 0.5, that is, half of the neuron outputs are randomly set to 0 during each training, so as to reduce the complexity of the model and improve the generalization ability of the model.

[0110] Fully connected layer 2: Add another fully connected layer with 4096 neurons and perform a linear transformation on the output of the Dropout layer. The shape of the output vector becomes [batch_size, 4096].

[0111] Activation function 6: Apply the ReLU activation function, and the shape of the output vector after activation is [batch_size, 4096].

[0112] Fully connected layer 3 (output layer): The last fully connected layer is used as the output layer. The number of neurons is adjusted as needed to obtain a feature vector of appropriate length. A fully connected layer with 128 neurons is used to perform a linear transformation on the activated output vector, and the shape of the output feature vector becomes [batch_size, 128].

[0113] The smoke and wind system data is extracted through convolutional neural network to construct the smoke change feature vector;

[0114] The characteristic vector of the smoke and wind system is expressed as:

[0115]

[0116] Where: Q vgin is the flue gas inlet flow rate, Q vgout is the flue gas outlet flow rate; p fi is the boiler inlet air pressure, p fo is the boiler outlet air pressure, T vgin is the boiler inlet temperature, T vgout Boiler outlet temperature

[0117] The steam-water change feature vector is constructed by extracting steam turbine system data through convolutional neural network and expressed as:

[0118] The soda change feature vector is expressed as:

[0119]

[0120] Where: p b Steam pressure, p w Feed water pressure, T b1 Steam temperature, T w1 Feed water temperature, Q b1 Steam flow, Q w1 Feedwater flow, H qb Drum liquid level;

[0121] The smoke, wind and turbine coordination prediction model is established based on the following formula based on the smoke change characteristic vector and the steam-water change characteristic vector:

[0122] (1) Smoke and wind system prediction model:

[0123] The state equation of the smoke and wind system is:

[0124]

[0125] Among them, A g Smoke and wind system corresponds to the system matrix; B g Smoke and wind system input matrix; u g Smoke and wind system input vector; Flue gas temperature change rate, T g Flue gas temperature, Q g Flue gas flow, P g Smoke and air duct pressure, u g1 Smoke and wind system input vector; Flue gas flow rate change rate; Smoke and air duct pressure change rate;

[0126] Smoke and wind system output equation:

[0127]

[0128] Among them, C g Output matrix; D g is the direct transfer matrix; y g1 Flue gas outlet temperature; y g2 Total flue gas flow; (2) Turbine system prediction model:

[0129] Turbine system state equation:

[0130]

[0131] Among them A s The steam turbine system corresponds to the system matrix; B s Turbine system input matrix; u s Turbine system input vector; Steam flow rate change rate; Q s Steam flow rate; n turbine speed; P s Steam turbine output power; u s1 Elements in the steam turbine system input vector;

[0132] Steam turbine system output equation:

[0133]

[0134] Among them, C s Output matrix of steam turbine system; D s Direct transfer matrix. s1 Actual output power of steam turbine; s2 Turbine speed.

[0135] (3) State space of smoke, air and steam turbine coordination prediction model:

[0136]

[0137] Among them, A sg , A gs represents the coupling matrix between the smoke and air system and the steam turbine system; B gs , B sg Input coupling matrix of smoke and air system and steam turbine system; x gs1 Indicates the enthalpy of change of feed water flow rate h pe ;x gs2 It represents the specific enthalpy of superheated steam h ps ;x gs3 Indicates valve opening u; u gs1 Indicates the steam temperature at the drum outlet c v ;u gs2 Indicates the steam mass flow rate D at the outlet of the flue gas system drum do ;u gs3 Indicates the steam pressure p at the outlet of the flue gas system drum dr ;

[0138] Optimize and train the flue gas turbine coordination prediction model through long short-term memory neural network; including:

[0139] The smoke, air and steam turbine coordination prediction model is optimized through the following parameters

[0140]

[0141] Where: n is the dimensional sample space; is the actual measurement space vector; is the prediction space vector;

[0142] After training, the optimized smoke, wind and steam turbine coordination prediction model is obtained as follows:

[0143]

[0144] Where: d is the prediction step length; y is the prediction model output parameter vector; x is the prediction model input parameter vector; f d Build an LSTM model to predict the current d moments;

[0145] The above formula is used to optimize the parameter change of the flue gas coordinated control prediction model, thereby achieving optimal adjustment and control of the power plant operation parameters.

[0146] Flue gas coordinated control prediction model output equation:

[0147]

[0148] Among them, C g ,C gs ,C sg ,C s Coefficient matrix of smoke, wind and steam turbine system; D g , D gs ,D sg ,D s Smoke, air and steam turbine system direct transmission matrix; x gs1 Indicates the enthalpy of change of feed water flow rate h pe ;x gs2 It represents the specific enthalpy of superheated steam h ps ;x gs3 Indicates valve opening u; u gs1 Indicates the steam temperature at the drum outlet c v ;u gs2 Indicates the steam mass flow rate D at the outlet of the flue gas system drum do ;u gs3 Indicates the steam pressure p at the outlet of the flue gas system drum dr ;y s1 Steam temperature T b1 ;y s2 Steam flow Q b1 ;y s3 Steam pressure p b .

[0149] Although the present invention has been described above, the present invention is not limited to the above-mentioned specific embodiments. The above-mentioned specific embodiments are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the purpose of the present invention, which are all within the protection of the present invention.

Claims

1. A remote operation and maintenance control method for waste heat power generation under variable operating mode, characterized in that: The method is based on a smoke and air system, a smoke and air turbine coordination model and a turbine system. The smoke and air turbine coordination model converts steam parameters output by the smoke and air system into steam parameters input by the turbine system. Therefore, the smoke and air system includes a first evaporation model, a second evaporation model and a steam drum unit. The method includes the following steps: Collect historical data of waste heat power station and build smoke and wind system according to the following formula; Where: C g Output matrix; D g is the direct transfer matrix; x g1 Flue gas temperature; x g2 Flue gas flow rate; x g3 Flue gas pressure; g1 Boiler outlet temperature T vgout ;y g2 Steam outlet flow rate Q vgout ;y g3 is the boiler outlet steam pressure p fo ;u g1 is the boiler inlet temperature L vgin ;u g2 is the flue gas inlet flow rate Q vgin ;u g3 is the boiler inlet air pressure p fi ; The smoke-air-steam-turbine coordination model converts the steam parameters output by the smoke-air system into steam parameters of the steam turbine according to the following formula; Where: h ps is the specific enthalpy of superheated steam; According to the steam parameters of the steam turbine, the steam turbine system is constructed according to the following formula: Among them, C s Output matrix of steam turbine system; D s Direct transfer matrix; x s1 Water flow rate Q w1 ;x s2 Feed water temperature T w1 ;x s3 Water supply pressure p w ;u s1 Steam flow Q b1 ;u s2 Steam temperature T b1 ;u s3 Steam pressure p b ;y s1 Actual output power of steam turbine; s2 Turbine speed; The smoke and wind system data is extracted through convolutional neural network to construct the smoke and wind change feature vector; Where: Q vgin is the flue gas inlet flow rate, Q vgout is the steam outlet flow rate; p fi is the boiler inlet air pressure, p fo is the boiler outlet steam pressure, T vgin is the boiler inlet temperature, T vgout The boiler outlet temperature is extracted through the convolutional neural network to construct the steam-water change feature vector, which is expressed as: Where: p b Steam pressure, p w Feed water pressure, T b1 Steam temperature, T w1 Feed water temperature, Q b1 Steam flow, Q w1 Water flow rate; The smoke, wind and steam turbine coordination prediction model is established based on the following formula based on the smoke and wind change characteristic vector and the steam and water change characteristic vector: Among them, A sg , A gs represents the coupling matrix between the smoke and air system and the steam turbine system; B gs , B sg Input coupling matrix of smoke and air system and steam turbine system; x gs1 Indicates the enthalpy of change of feed water flow rate h pe ;x gs2 It represents the specific enthalpy of superheated steam h ps ;x gs3 Indicates valve opening u; u gs1 Indicates the steam temperature at the drum outlet c v ;u gs2 Indicates the steam mass flow rate D at the outlet of the flue gas system drum do ;u gs3 Indicates the steam pressure p at the outlet of the flue gas system drum dr ; The flue gas turbine coordination prediction model is optimized and trained through long short-term memory neural network; The parameter changes of the flue gas coordinated control prediction model are optimized through the following formula, thereby achieving optimal adjustment and control of the power plant operation parameters: Among them, C g , C gs , C sg , C s Coefficient matrix of smoke, wind and steam turbine system; D g , D gs , D sg , D s Smoke, air and steam turbine system direct transmission matrix; x gs1 Indicates the enthalpy of change of feed water flow rate h pe ;x gs2 It represents the specific enthalpy of superheated steam h ps ;x gs3 Indicates valve opening u; u gs1 Indicates the steam temperature at the drum outlet c v ;u gs2 Indicates the steam mass flow rate D at the outlet of the flue gas system drum do ;u gs3 Indicates the steam pressure p at the outlet of the flue gas system drum dr ;y s1 Steam temperature T b1 ;y s2 Steam flow Q b1 ;y s3 Steam pressure p b .

2. According to claim 1, a remote operation and maintenance control method for waste heat power generation under variable operating mode is characterized in that: The optimization training process of the flue gas turbine coordination prediction model is carried out through the long short-term memory neural network; including: The smoke, air and steam turbine coordination prediction model is optimized by the following formula: Where: y i for y s1 Steam temperature T b1 ,y s2 Steam flow Q b1 ,y s3 Steam pressure p b The actual value of the sample space; for y s1 Steam temperature T b1 ,y s2 Steam flow Q b1 ,y s3 Steam pressure p b Prediction estimate; n is the dimensional sample space; is the actual measurement space vector; is the prediction space vector; After training, the optimized smoke, wind and steam turbine coordination prediction model is obtained as follows: Where: d is the prediction step length; y1 is the steam temperature T b1 , y2 steam flow Q b1 , y3 steam pressure p b Output vector of the prediction model; x1 enthalpy of change of feed water flow h pe , x2 specific enthalpy of superheated steam h ps , x3 valve opening u is the input vector of the prediction model; f d Build an LSTM model to predict the current d moments.

3. The remote operation and maintenance control method for waste heat power generation under variable operating mode according to claim 1 is characterized in that: The process of constructing a smoke and air system through the first evaporation model, the second evaporation model and the drum unit includes: The first evaporation model is constructed according to the following formula based on the heat source heat transfer absorption energy and the inlet energy; Where: Q represents the amount of heat absorbed; D represents the mass flow rate; D represents the volume; M represents the mass; C pe represents the specific heat of metal at constant pressure in the evaporation system; T represents temperature; h represents specific enthalpy; ρ represents density; superscript: ' represents saturated water; * represents saturated steam; subscript: d represents steam drum; f represents boiler feed water; e represents evaporation system; m represents metal; s represents saturated state; The second evaporation model is constructed according to the following formula based on the energy absorbed by the heat source, the energy brought into the inlet, and the energy output by the first evaporation model; Constructing a drum unit according to the parameters output by the second evaporation model according to the following formula; Where: u represents the valve opening, c v is the steam temperature at the drum outlet; subscript: v represents the turbine control valve, out represents the outlet; When the steam flow rate at the drum outlet, the drum pressure, the control valve opening and the pressure behind the valve are known, the drum outlet flow rate is: Where: c v D is the steam temperature at the drum outlet; do u is the steam mass flow rate at the outlet of the flue gas system drum; v Steam turbine valve opening; p dr Steam pressure at the outlet of flue gas system drum; p v，out Steam pressure at turbine inlet; The flue gas system is obtained by calculating the first evaporation model, the second evaporation model and the drum unit according to the following formula; According to the flue gas system, the enthalpy of change of water flow rate is obtained according to the following formula: pe ;

Images