Weighted fuzzy predictive control-based dry and wet state conversion control method and system for thermal power generating unit

Abstract

The invention discloses a thermal power generating unit dry and wet state conversion control method and system based on weighted fuzzy predictive control, and the method comprises the steps: dynamically adjusting operation data according to the difference value between control data and a dry state control data set value, and taking the operation data as a dry state operation data deviation value; dynamically adjusting the operation data according to a difference value between the control data and a wet state control data set value, and taking the operation data as a wet state operation data deviation value; predicting dry state control data by using the dry state operation data deviation value, and predicting wet state control data by using the wet state operation data deviation value; performing fuzzy control on the predicted dry state control data to obtain a dry state judgment coefficient; performing fuzzy control on the predicted wet state control data to obtain a wet state judgment coefficient; and weighting the dry state operation data and the wet state operation data by using the dry state judgment coefficient and the wet state judgment coefficient, updating the operation data, and performing dry and wet state conversion control on the thermal power generating unit by using the updated operation data.

Description

Technical Field

[0001] The present invention belongs to the technical field of thermal automation control, and in particular relates to a dry-wet state conversion control method and system for a thermal power unit based on weighted fuzzy predictive control. Background Art

[0002] As the installed capacity of renewable energy generators like wind, photovoltaic, and hydropower gradually increases, the proportion of thermal power generation in the power system continues to decline. However, due to the high volatility and low stability of renewable energy generation, thermal power generation remains a pillar of my country's energy sector. For example, deep peak regulation of ultra-supercritical (supercritical) thermal power units helps maintain stable and flexible power generation. To improve power generation stability and flexibility, it is imperative to incorporate renewable energy and implement deep peak regulation of ultra-supercritical (supercritical) thermal power units.

[0003] As installed capacity of renewable energy continues to grow, the grid's demand for flexibility from ultra-supercritical (ULC) thermal power plants is also increasing. However, current peak-shaving capabilities and regulation rates still cannot fully meet the grid's regulation needs. This problem manifests itself in two key areas. First, the peak-shaving depth of thermal power plants remains insufficient in the long term, making their regulation capabilities unable to support the fluctuations of large-scale renewable energy generation. Second, the response speed of thermal power plants participating in deep peak-shaving is also significantly insufficient, primarily manifested in prolonged load ramping and reduction times.

[0004] Due to these issues, when the operating conditions of ultra-supercritical (ULC) thermal power units change, their input parameters need to be adjusted accordingly. This is especially true during deep peak load regulation, often facing numerous challenges such as large fluctuations in feedwater flow, limited adjustment ranges for auxiliary actuators, and the need for manual feedwater adjustment under special operating conditions. Therefore, designing automatic control strategies and optimizing operating modes are crucial for addressing these issues. Furthermore, during deep peak load regulation, the dry and wet operating conditions of thermal power units differ, necessitating significant differences in control strategies. Smoothly switching between dry and wet states through automatic control is currently a key challenge in optimizing dry and wet operation of ultra-supercritical (ULC) thermal power units. Summary of the Invention

[0005] In response to the problems existing in the prior art, the present invention provides a dry-wet state conversion control method and system for thermal power units based on weighted fuzzy predictive control, which can improve the reliability and flexibility of dry-wet state conversion of thermal power units.

[0006] To achieve the above technical objectives, the present invention adopts the following technical solution: a dry-wet state conversion control method for a thermal power unit based on weighted fuzzy predictive control, characterized in that it includes the following steps: Step 1: setting the dry state control data setting value of the thermal power unit in the dry state operation mode and the wet state control data setting value of the thermal power unit in the wet state operation mode respectively; Step 2: Acquire status data of the thermal power unit during actual operation, wherein the status data includes operation data and control data; Step 3: Dynamically adjust the operating data according to the difference between the control data and the set value of the dry-state control data, as the dry-state operating data deviation value; dynamically adjust the operating data according to the difference between the control data and the set value of the wet-state control data, as the wet-state operating data deviation value; Step 4: Input the dry state operation data deviation value into the first LSTM model to predict the dry state control data; input the wet state operation data deviation value into the second LSTM model to predict the wet state control data; Step 5: Perform fuzzy control on the predicted dry state control data to obtain the dry state judgment coefficient; perform fuzzy control on the predicted wet state control data to obtain the wet state judgment coefficient; Step 6: Use the dry state judgment coefficient and the wet state judgment coefficient to weight the dry state operation data and the wet state operation data, update the operation data, and use the updated operation data to perform dry-wet state conversion control of the thermal power unit.

[0007] Furthermore, the operating data include: turbine main steam valve opening, coal consumption, feed water flow and recirculation water flow; the control data include: main steam pressure, output power, midpoint temperature, main steam temperature and water tank water level.

[0008] Furthermore, when the thermal power unit is in a dry operation mode, the operating data include: the opening of the turbine main steam valve, the amount of coal burned and the feed water flow, and the control data include: the main steam pressure, the output power and the main steam temperature; when the thermal power unit is in a wet operation mode, the operating data include: the opening of the turbine main steam valve, the amount of coal burned, the feed water flow and the recirculating water flow; the control data include: the main steam pressure, the output power mid-point temperature and the water level of the water tank.

[0009] Furthermore, the determination process of the dry state judgment coefficient in step 5 is: The basic domain and fuzzy domain of dry state control data are set, the quantization factor is determined, and the fuzzy subset corresponding to the fuzzy domain is set. The membership function of the fuzzy subset of the dry state control data is described by the triangular membership function. The basic domain, fuzzy domain and fuzzy subset corresponding to the dry state judgment coefficient are set, and the membership function of the fuzzy subset of the dry state judgment coefficient is described by the triangular membership function; According to the influence of dry state control data on dry state judgment coefficient, fuzzy control rules of fuzzy subsets of dry state control data and fuzzy subsets of dry state judgment coefficient are established; The predicted dry state control data is fuzzified by quantization factors, the membership function of the corresponding fuzzy subset of the dry state control data is found, and the corresponding fuzzy control rules are found. The fuzzy reasoning is performed using the Mamdani algorithm to obtain the fuzzy set of dry state judgment coefficients. The center of the fuzzy set of the dry state judgment coefficient is calculated using the center of gravity method as the dry state judgment coefficient.

[0010] Furthermore, the determination process of the wet state judgment coefficient in step 5 is as follows: The basic domain and fuzzy domain of wet state control data are set, the quantization factor is determined, and the fuzzy subset corresponding to the fuzzy domain is set. The membership function of the fuzzy subset of the wet state control data is described by the triangular membership function. The basic domain, fuzzy domain and fuzzy subset corresponding to the wet state judgment coefficient are set, and the membership function of the fuzzy subset of the wet state judgment coefficient is described by the triangular membership function; According to the influence of wet state control data on wet state judgment coefficient, the fuzzy control rules of the fuzzy subset of wet state control data and the fuzzy subset of wet state judgment coefficient are established; The predicted wet state control data is fuzzified by quantization factors, the membership function of the corresponding fuzzy subset of the wet state control data is found, and the corresponding fuzzy control rules are found. The fuzzy reasoning is performed using the Mamdani algorithm to obtain the fuzzy set of wet state judgment coefficients. The center of the fuzzy set of the wet state judgment coefficient is calculated using the center of gravity method as the wet state judgment coefficient.

[0011] Furthermore, the updating process of the operation data is as follows:

[0012] in, Indicates the first i The updated value of the data, Indicates the first i The dry running data corresponding to the data is represents the dry state judgment coefficient, Indicates the first i The wet operation data corresponding to the data, Indicates the wet state judgment coefficient.

[0013] Furthermore, the main steam pressure and output power in the predicted dry state control data are fuzzy controlled to obtain the dry state judgment coefficient, and the difference between 1 and the dry state judgment coefficient is used as the wet state judgment coefficient.

[0014] Furthermore, the process of obtaining the dry state judgment coefficient is as follows: Set the basic domain of main steam pressure to [p T,w , p T,d ], the fuzzy domain is set to {1, 2, 3}, and the quantization factor K pT Set to 3 / (p T,d -p T,w ), the corresponding fuzzy subset in the fuzzy domain of the main steam pressure is {S pT , M pT , B pT}, and describe the membership function of the fuzzy subset of the main steam pressure through the triangle membership function, where S pT 、M pT 、B pT Respectively indicate the main steam pressure is small, medium and large; The basic domain of output power is set to [N E,w , N E,d ], the fuzzy domain is set to {1, 2, 3}, and the quantization factor K pT Set to 3 / (N E,d -N E,w ), the corresponding fuzzy subset in the fuzzy domain of the output power is {S NE , M NE , B NE}, and describe the membership function of the fuzzy subset of output power through the triangular membership function, where S NE 、M NE 、B NE Respectively indicate the situations where the output power is small, medium, and large; The basic domain of the dry state judgment coefficient is set to [0, 1], the fuzzy domain is set to {1, 2, 3, 4, 5}, and the quantization factor K is set to [0, 1]. pT Set to 5, the corresponding fuzzy subset in the fuzzy domain of the dry state judgment coefficient is {VS j , S j , M j , B j ,VB j}, and describe the membership function of the fuzzy subset of the dry state judgment coefficient through the triangle membership function, where VS j 、S j 、M j 、B j ,VB j They represent the high and low dry state judgment coefficients respectively; According to the influence of main steam pressure and output power on dry state judgment coefficient, fuzzy control rules are formulated; After the predicted main steam pressure and output power are fuzzified by quantitative factors, the corresponding membership functions are searched and the corresponding fuzzy control rules are found in the fuzzy control rules. The fuzzy reasoning is performed using the Mamdani algorithm to obtain the fuzzy set of dry state judgment coefficients. The center of the fuzzy set is calculated using the center of gravity method as the dry state judgment coefficient.

[0015] Furthermore, the fuzzy control rules are specifically as follows: When the fuzzy value of the main steam pressure is in the fuzzy subset S pT And the fuzzy value of the output power is located in the fuzzy subset S NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset VS j ; When the fuzzy value of the main steam pressure is in the fuzzy subset S pT And the fuzzy value of the output power is located in the fuzzy subset M NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset S j ; When the fuzzy value of the main steam pressure is in the fuzzy subset S pT And the fuzzy value of the output power is in the fuzzy subset B NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset M j ; When the fuzzy value of the main steam pressure is in the fuzzy subset M pT And the fuzzy value of the output power is located in the fuzzy subset S NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset S j ; When the fuzzy value of the main steam pressure is in the fuzzy subset M pT And the fuzzy value of the output power is located in the fuzzy subset M NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset M j ; When the fuzzy value of the main steam pressure is in the fuzzy subset M pT And the fuzzy value of the output power is in the fuzzy subset B NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset B j ; When the fuzzy value of the main steam pressure is in the fuzzy subset B pT And the fuzzy value of the output power is located in the fuzzy subset S NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset M j ; When the fuzzy value of the main steam pressure is in the fuzzy subset B pT And the fuzzy value of the output power is located in the fuzzy subset M NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset Bj ; When the fuzzy value of the main steam pressure is in the fuzzy subset B pT And the fuzzy value of the output power is in the fuzzy subset B NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset VB j .

[0016] Furthermore, the present invention also provides a dry-wet state conversion control system for a thermal power unit based on weighted fuzzy predictive control, comprising: a dry-state model predictive controller, a wet-state model predictive controller, a first LSTM model, a second LSTM model, a dry-state fuzzy identifier, and a wet-state fuzzy identifier; The dry state model prediction controller dynamically adjusts the operation data according to the difference between the control data in the actual operation process of the thermal power unit and the dry state control data set value to obtain the dry state operation data deviation value; The wet state model prediction controller dynamically adjusts the operation data according to the difference between the control data in the actual operation process of the thermal power unit and the wet state control data set value to obtain the wet state operation data deviation value; The first LSTM model predicts dry-state control data according to the dry-state operation data deviation value, and the second LSTM model predicts wet-state control data according to the wet-state operation data deviation value; The dry state fuzzy identifier performs fuzzy control according to the predicted dry state control data to obtain a dry state judgment coefficient; The wet state fuzzy identifier performs fuzzy control according to the predicted wet state control data to obtain a wet state judgment coefficient; The dry state judgment coefficient and the wet state judgment coefficient are used to weight the dry state operation data and the wet state operation data, the operation data is updated, and the updated operation data is used to perform dry-wet state conversion control of the thermal power unit.

[0017] Compared with the prior art, the present invention has the following beneficial effects: the dry-wet state conversion control method and system of a thermal power unit based on weighted fuzzy predictive control of the present invention respectively use a first LSTM model and a second LSTM model to realize the prediction of dry-state control data and wet-state control data, and utilize the synergistic effect of the forget gate, input gate, and output gate of LSTM to capture time series characteristics, dynamically process information in the time series, make reasonable decisions at each time step, and realize accurate prediction of control data; at the same time, the present invention performs fuzzy control on the predicted dry-state control data and wet-state control data respectively, and uses the judgment coefficient of fuzzy control to adjust the operating data. This method can better handle the nonlinear system of dry-wet state conversion, has strong flexibility and interpretability, and is suitable for the coordinated control of dry-wet state conversion multivariable systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1This is a schematic diagram of an ultra-supercritical thermal power unit; Figure 2 Schematic diagram of the dry-wet state conversion control method of a thermal power unit based on weighted fuzzy predictive control according to the present invention; Figure 3 Schematic diagram of the membership function of the control data predicted in the present invention, wherein the left figure represents the fuzzy membership function of the main steam pressure, and the right figure represents the fuzzy membership function of the output power; Figure 4 Schematic diagram of the membership function of the dry state judgment coefficient in the present invention. DETAILED DESCRIPTION

[0019] The technical solution of the present invention will be further explained below with reference to the accompanying drawings.

[0020] like Figure 1 This is a schematic diagram of a supercritical (ultracritical) thermal power unit according to the present invention. This supercritical (ultracritical) thermal power unit comprises a boiler system, a steam turbine, a condenser, a low-pressure feedwater heater, a deaerator, a high-pressure feedwater heater, and a generator. The boiler system includes a burner, an economizer, a water-cooled wall, a steam-water separator, a water storage tank, and a superheater; the steam turbine includes a high-pressure cylinder, an intermediate-pressure cylinder, and a low-pressure cylinder. In this supercritical (ultracritical) thermal power unit, pulverized coal is burned in the burner at a fuel quantity of uB. The resulting flue gas then transfers heat to the economizer, water-cooled wall, superheater, and reheater. The feedwater flow rate is uW. After being heated in the economizer, the feedwater enters the water-cooled wall. At high loads, all the feedwater is heated to superheated steam and enters the superheater, with the intermediate temperature at Tmp. At low loads, the feedwater is heated to a gas-liquid mixture. After passing through the steam-water separator, saturated steam enters the superheater for further heating. The saturated water is stored in a storage tank with a water level of Hst. The feedwater in the storage tank is recirculated and re-enters the economizer at a recirculating water flow rate of uR. The main steam from the superheater outlet enters the high-pressure cylinder of the steam turbine to perform work. The turbine main steam valve opening is uT, the main steam pressure is pT, and the main steam temperature is Tms. The steam from the high-pressure cylinder outlet enters the primary reheater. The reheated steam at the outlet passes through the intermediate-pressure and low-pressure cylinders of the steam turbine in sequence to perform work. The output power of the supercritical (ultra-supercritical) thermal power unit is NE. Steam is extracted from each stage of the steam turbine and heated by feedwater heaters. Low-quality steam at the turbine outlet enters the condenser, condensing it into feedwater. This feedwater is then reheated by the economizer through the feedwater heater tubes, completing the steam-water circulation system of the boiler-steam-water reheat system. Therefore, operating data for supercritical (ultracritical) thermal power units includes turbine main steam valve opening, coal consumption, feedwater flow, and recirculation water flow. Control data includes main steam pressure, output power, intermediate point temperature, main steam temperature, and water storage tank level.

[0021] Thermal power units are classified into dry and wet modes based on the operating conditions of the boiler separator. During dry mode, the separator outlet is steam, while during wet mode, the separator outlet is saturated steam and saturated water. When the unit is in dry mode, operating data includes: turbine main steam valve opening, coal consumption, and feedwater flow; control data includes: main steam pressure, output power, and main steam temperature. When the unit is in wet mode, operating data includes: turbine main steam valve opening, coal consumption, feedwater flow, and recirculation water flow; control data includes: main steam pressure, output power midpoint temperature, and water tank water level.

[0022] like Figure 2 Schematic diagram of a dry-wet state conversion control method for a thermal power unit based on weighted fuzzy predictive control according to the present invention. The dry-wet state conversion control method for a thermal power unit comprises the following steps: Step 1: setting the dry state control data setting value of the thermal power unit in the dry state operation mode and the wet state control data setting value of the thermal power unit in the wet state operation mode respectively; Step 2: Obtain status data during the actual operation of the thermal power unit, which includes operation data and control data; Step 3: Dynamically adjust the operating data according to the difference between the control data and the set value of the dry-state control data, as the dry-state operating data deviation value; dynamically adjust the operating data according to the difference between the control data and the set value of the wet-state control data, as the wet-state operating data deviation value; adjusting the operating data according to the set value can improve the accuracy and stability of the thermal power unit in the dry or wet state operation mode, so that the output of the thermal power unit is close to the set value, and improve the avoidance of excessive fluctuation of the thermal power unit; Step 4: Input the dry-state operation data deviation value into the first LSTM model to predict the dry-state control data; input the wet-state operation data deviation value into the second LSTM model to predict the wet-state control data. As a prediction model, LSTM can process long sequence information, capture complex nonlinear relationships, support multi-input and multi-output prediction, and has a certain degree of interpretability. By leveraging the synergistic effect of the forget gate, input gate, and output gate of LSTM, it can capture time series characteristics, dynamically process information in the time series, make reasonable decisions at each time step, and achieve accurate prediction of control data. Step 5: Fuzzy control is performed on the predicted dry-state control data to obtain the dry-state judgment coefficient; fuzzy control is performed on the predicted wet-state control data to obtain the wet-state judgment coefficient. Fuzzy control uses fuzzy rules to represent the relationship between the operating data and control data of the thermal power unit. It can naturally adapt to nonlinear systems. The rules are intuitive and easy to adjust. It has strong robustness and is suitable for dry-wet state conversion control of complex thermal power units. Step 6: Use the dry-state judgment coefficient and the wet-state judgment coefficient to weight the dry-state and wet-state operating data, update the operating data, and use the updated operating data to control the dry-wet state transition of the thermal power unit. This method can reduce parameter fluctuations and prevent the unit from switching repeatedly between dry and wet states. This improves the stability of the thermal power unit, enables the unit to respond quickly to load changes, reduces the unit's fault tolerance, and ensures equipment safety.

[0023] The updating process of the operating data in the present invention is:

[0024] in, Indicates the first i The updated value of the data, Indicates the first i The dry running data corresponding to the data is represents the dry state judgment coefficient, Indicates the first i The wet operation data corresponding to the data, Indicates the wet state judgment coefficient.

[0025] In one technical solution of the present invention, objective functions that simultaneously consider the optimization of operating data and control data are constructed in the first LSTM model and the second LSTM model, with the goal of minimizing the objective function. Taking into account the high requirements for control stability during the dry-wet state conversion process of ultra-supercritical thermal power units, augmented dynamic constraints on state data are introduced, thereby improving the stability and robustness of ultra-supercritical thermal power units.

[0026] The objective function in the present invention The construction process is:

[0027] in, N represents the time increment, i express N The index of Indicates the current moment, represents the predicted control data, Indicates the control data setting value, represents the two-norm, Indicates the increment of running data.

[0028] The augmented dynamic constraints of state data in the present invention include:

[0029] in, Indicates running data, Indicates the minimum value of the running data, Indicates the maximum value of the running data, Indicates the minimum value of the running data increment, Indicates the maximum value of the running data increment. represents the minimum value of the predicted control data, The maximum value of the wet control data.

[0030] The determination process of the dry state judgment coefficient in the present invention is: The basic domain and fuzzy domain of dry state control data are set, the quantization factor is determined, and the fuzzy subset corresponding to the fuzzy domain is set. The membership function of the fuzzy subset of the dry state control data is described by the triangular membership function. The basic domain, fuzzy domain and fuzzy subset corresponding to the dry state judgment coefficient are set, and the membership function of the fuzzy subset of the dry state judgment coefficient is described by the triangular membership function; According to the influence of dry state control data on dry state judgment coefficient, fuzzy control rules of fuzzy subsets of dry state control data and fuzzy subsets of dry state judgment coefficient are established; The predicted dry state control data is fuzzified by quantization factors, the membership function of the corresponding fuzzy subset of the dry state control data is found, and the corresponding fuzzy control rules are found. The fuzzy reasoning is performed using the Mamdani algorithm to obtain the fuzzy set of dry state judgment coefficients. The center of the fuzzy set of the dry state judgment coefficient is calculated using the center of gravity method as the dry state judgment coefficient.

[0031] The determination process of the wet state judgment coefficient in the present invention is: The basic domain and fuzzy domain of wet state control data are set, the quantization factor is determined, and the fuzzy subset corresponding to the fuzzy domain is set. The membership function of the fuzzy subset of the wet state control data is described by the triangular membership function. The basic domain, fuzzy domain and fuzzy subset corresponding to the wet state judgment coefficient are set, and the membership function of the fuzzy subset of the wet state judgment coefficient is described by the triangular membership function; According to the influence of wet state control data on wet state judgment coefficient, the fuzzy control rules of the fuzzy subset of wet state control data and the fuzzy subset of wet state judgment coefficient are established; The predicted wet state control data is fuzzified by quantization factors, the membership function of the corresponding fuzzy subset of the wet state control data is found, and the corresponding fuzzy control rules are found. The fuzzy reasoning is performed using the Mamdani algorithm to obtain the fuzzy set of wet state judgment coefficients. The center of the fuzzy set of the wet state judgment coefficient is calculated using the center of gravity method as the wet state judgment coefficient.

[0032] In one technical solution of the present invention, since power generation and main steam pressure are the primary factors characterizing unit load and dry / wet states during the operation of an ultra-supercritical thermal power unit, fuzzy control is performed on the main steam pressure and output power in the predicted dry-state control data to obtain a dry-state determination coefficient. The difference between 1 and the dry-state determination coefficient is used as the wet-state determination coefficient. Therefore, the process for obtaining the dry-state determination coefficient is as follows: Set the basic domain of main steam pressure to [p T,w , p T,d ], the fuzzy domain is set to {1, 2, 3}, and the quantization factor K pT Set to 3 / (p T,d -p T,w ), the corresponding fuzzy subset in the fuzzy domain of the main steam pressure is {S pT , M pT , B pT}, and describe the membership function of the fuzzy subset of the main steam pressure through the triangular membership function, such as Figure 3 As shown in the left figure, where S pT 、M pT 、B pT Respectively indicate the main steam pressure is small, medium and large; The basic domain of output power is set to [N E,w , N E,d ], the fuzzy domain is set to {1, 2, 3}, and the quantization factor K pT Set to 3 / (N E,d -N E,w ), the corresponding fuzzy subset in the fuzzy domain of the output power is {S NE , M NE , B NE}, and describe the membership function of the fuzzy subset of output power through the triangular membership function, such as Figure 3 As shown in the right figure, S NE 、M NE 、B NE Respectively indicate the situations where the output power is small, medium, and large; The basic domain of the dry state judgment coefficient is set to [0, 1], the fuzzy domain is set to {1, 2, 3, 4, 5}, and the quantization factor K is set to [0, 1]. pT Set to 5, the corresponding fuzzy subset in the fuzzy domain of the dry state judgment coefficient is {VS j , S j , M j , B j , VB j}, and describe the membership function of the fuzzy subset of the dry state judgment coefficient through the triangle membership function, such as Figure 4As shown, where VS j 、S j 、M j 、B j ,VB j They represent the high and low dry state judgment coefficients respectively; According to the influence of main steam pressure and output power on dry state judgment coefficient, fuzzy control rules are formulated, as shown in Table 1: When the fuzzy value of main steam pressure is within the fuzzy subset S pT And the fuzzy value of the output power is located in the fuzzy subset S NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset VS j ; When the fuzzy value of the main steam pressure is in the fuzzy subset S pT And the fuzzy value of the output power is located in the fuzzy subset M NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset S j ; When the fuzzy value of the main steam pressure is in the fuzzy subset S pT And the fuzzy value of the output power is in the fuzzy subset B NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset M j ; When the fuzzy value of the main steam pressure is in the fuzzy subset M pT And the fuzzy value of the output power is located in the fuzzy subset S NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset S j ; When the fuzzy value of the main steam pressure is in the fuzzy subset M pT And the fuzzy value of the output power is located in the fuzzy subset M NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset M j ; When the fuzzy value of the main steam pressure is in the fuzzy subset M pT And the fuzzy value of the output power is in the fuzzy subset B NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset B j ; When the fuzzy value of the main steam pressure is in the fuzzy subset B pT And the fuzzy value of the output power is located in the fuzzy subset S NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset M j ; When the fuzzy value of the main steam pressure is in the fuzzy subset B pT And the fuzzy value of the output power is located in the fuzzy subset M NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset B j ; When the fuzzy value of the main steam pressure is in the fuzzy subset B pT And the fuzzy value of the output power is in the fuzzy subset B NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset VB j .

[0033] Table 1 Fuzzy control rules

[0034] After the predicted main steam pressure and output power are fuzzified by quantitative factors, the corresponding membership functions are searched and the corresponding fuzzy control rules are found in the fuzzy control rules. The fuzzy reasoning is performed using the Mamdani algorithm to obtain the fuzzy set of dry state judgment coefficients. The center of the fuzzy set is calculated using the center of gravity method as the dry state judgment coefficient.

[0035] when j When 1=1, the supercritical thermal power unit implements dry state control, the main steam valve opening uT of the turbine controls the main steam pressure pT, the fuel quantity uB controls the output power NE, and the feed water flow uW controls the main steam temperature Tms; when j When 1=0, the supercritical thermal power unit implements wet control, the main steam valve opening uT of the steam turbine controls the main steam pressure pT, the fuel quantity uB controls the output power NE, the feed water flow uW controls the intermediate point temperature Tmp, and the recirculating water flow uR controls the water level Hst of the water storage tank; when j When 1∈(0,1), the supercritical thermal power unit is between dry and wet states, the turbine main steam valve opening uT controls the main steam pressure pT, the fuel quantity uB controls the output power NE, the feed water flow uW controls the main steam temperature Tms and the intermediate point temperature Tmp, and the recirculating water flow uR controls the water level Hst in the water storage tank.

[0036] In one technical solution of the present invention, a dry-wet state conversion control system for a thermal power unit based on weighted fuzzy predictive control is also provided, comprising: a dry-state model predictive controller, a wet-state model predictive controller, a first LSTM model, a second LSTM model, a dry-state fuzzy recognizer, and a wet-state fuzzy recognizer; The dry state model prediction controller dynamically adjusts the operating data according to the difference between the control data in the actual operation process of the thermal power unit and the set value of the dry state control data to obtain the dry state operating data deviation value; The wet state model predictive controller dynamically adjusts the operating data according to the difference between the control data in the actual operation process of the thermal power unit and the set value of the wet state control data, and obtains the wet state operating data deviation value; The first LSTM model predicts dry-state control data based on the deviation value of the dry-state operation data, and the second LSTM model predicts wet-state control data based on the deviation value of the wet-state operation data; The dry state fuzzy identifier performs fuzzy control according to the predicted dry state control data to obtain the dry state judgment coefficient; The wet state fuzzy identifier performs fuzzy control according to the predicted wet state control data to obtain the wet state judgment coefficient; The dry state judgment coefficient and the wet state judgment coefficient are used to weight the dry state operation data and the wet state operation data, the operation data is updated, and the updated operation data is used to perform dry-wet state conversion control of the thermal power unit.

[0037] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions based on the principles of the present invention are within the scope of protection of the present invention. It should be noted that for those skilled in the art, various improvements and modifications that do not depart from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A dry-wet state conversion control method for a thermal power unit based on weighted fuzzy predictive control, characterized in that: The steps include: Step 1: setting the dry state control data setting value of the thermal power unit in the dry state operation mode and the wet state control data setting value of the thermal power unit in the wet state operation mode respectively; Step 2: Acquire status data of the thermal power unit during actual operation, wherein the status data includes operation data and control data; Step 3: Dynamically adjust the operating data according to the difference between the control data and the set value of the dry-state control data, as the dry-state operating data deviation value; dynamically adjust the operating data according to the difference between the control data and the set value of the wet-state control data, as the wet-state operating data deviation value; Step 4: Input the dry state operation data deviation value into the first LSTM model to predict the dry state control data; input the wet state operation data deviation value into the second LSTM model to predict the wet state control data; Step 5: Perform fuzzy control on the predicted dry state control data to obtain the dry state judgment coefficient; perform fuzzy control on the predicted wet state control data to obtain the wet state judgment coefficient; Step 6: Use the dry state judgment coefficient and the wet state judgment coefficient to weight the dry state operation data and the wet state operation data, update the operation data, and use the updated operation data to perform dry-wet state conversion control of the thermal power unit.

2. A dry-wet state conversion control method for a thermal power unit based on weighted fuzzy predictive control according to claim 1, characterized in that: The operating data include: turbine main steam valve opening, coal consumption, feed water flow and recirculation water flow; the control data include: main steam pressure, output power, intermediate point temperature, main steam temperature and water tank water level.

3. The dry-wet state conversion control method of a thermal power unit based on weighted fuzzy predictive control according to claim 2 is characterized in that: When the thermal power unit is in dry operation mode, the operating data include: turbine main steam valve opening, coal consumption and feed water flow, and the control data include: main steam pressure, output power and main steam temperature; when the thermal power unit is in wet operation mode, the operating data include: turbine main steam valve opening, coal consumption, feed water flow and recirculating water flow; the control data include: main steam pressure, output power mid-point temperature and water level in the water storage tank.

4. The dry-wet state conversion control method for a thermal power unit based on weighted fuzzy predictive control according to claim 1, characterized in that: The process of determining the dry state judgment coefficient in step 5 is as follows: The basic domain and fuzzy domain of dry state control data are set, the quantization factor is determined, and the fuzzy subset corresponding to the fuzzy domain is set. The membership function of the fuzzy subset of the dry state control data is described by the triangular membership function. The basic domain, fuzzy domain and fuzzy subset corresponding to the dry state judgment coefficient are set, and the membership function of the fuzzy subset of the dry state judgment coefficient is described by the triangular membership function; According to the influence of dry state control data on dry state judgment coefficient, fuzzy control rules of fuzzy subsets of dry state control data and fuzzy subsets of dry state judgment coefficient are established; The predicted dry state control data is fuzzified by quantization factors, the membership function of the corresponding fuzzy subset of the dry state control data is found, and the corresponding fuzzy control rules are found. The fuzzy reasoning is performed using the Mamdani algorithm to obtain the fuzzy set of dry state judgment coefficients. The center of the fuzzy set of the dry state judgment coefficient is calculated using the center of gravity method as the dry state judgment coefficient.

5. The dry-wet state conversion control method of a thermal power unit based on weighted fuzzy predictive control according to claim 1, characterized in that: The determination process of the wet state judgment coefficient in step 5 is: The basic domain and fuzzy domain of wet state control data are set, the quantization factor is determined, and the fuzzy subset corresponding to the fuzzy domain is set. The membership function of the fuzzy subset of the wet state control data is described by the triangular membership function. The basic domain, fuzzy domain and fuzzy subset corresponding to the wet state judgment coefficient are set, and the membership function of the fuzzy subset of the wet state judgment coefficient is described by the triangular membership function; According to the influence of wet state control data on wet state judgment coefficient, the fuzzy control rules of the fuzzy subset of wet state control data and the fuzzy subset of wet state judgment coefficient are established; The predicted wet state control data is fuzzified by quantization factors, the membership function of the corresponding fuzzy subset of the wet state control data is found, and the corresponding fuzzy control rules are found. The fuzzy reasoning is performed using the Mamdani algorithm to obtain the fuzzy set of wet state judgment coefficients. The center of the fuzzy set of the wet state judgment coefficient is calculated using the center of gravity method as the wet state judgment coefficient.

6. The dry-wet state conversion control method for a thermal power unit based on weighted fuzzy predictive control according to claim 1, characterized in that: The updating process of the operating data is as follows: in, Indicates the first i The updated value of the data, Indicates the first i The dry running data corresponding to the data is represents the dry state judgment coefficient, Indicates the first i The wet operation data corresponding to the data, Indicates the wet state judgment coefficient.

7. The dry-wet state conversion control method for a thermal power unit based on weighted fuzzy predictive control according to claim 3 is characterized in that: The main steam pressure and output power in the predicted dry state control data are subjected to fuzzy control to obtain the dry state judgment coefficient, and the difference between 1 and the dry state judgment coefficient is used as the wet state judgment coefficient.

8. The dry-wet state conversion control method for a thermal power unit based on weighted fuzzy predictive control according to claim 7, characterized in that: The process of obtaining the dry state judgment coefficient is as follows: Set the basic domain of main steam pressure to [p T,w , p T,d ], the fuzzy domain is set to {1, 2, 3}, and the quantization factor K pT Set to 3 / (p T,d -p T,w ), the corresponding fuzzy subset in the fuzzy domain of the main steam pressure is {S pT , M pT , B pT }, and describe the membership function of the fuzzy subset of the main steam pressure through the triangle membership function, where S pT 、M pT 、B pT Respectively indicate the main steam pressure is small, medium and large; The basic domain of output power is set to [N E,w , N E,d ], the fuzzy domain is set to {1, 2, 3}, and the quantization factor K pT Set to 3 / (N E,d -N E,w ), the corresponding fuzzy subset in the fuzzy domain of the output power is {S NE , M NE , B NE }, and describe the membership function of the fuzzy subset of output power through the triangular membership function, where S NE 、M NE 、B NE Respectively indicate the situations where the output power is small, medium, and large; The basic domain of the dry state judgment coefficient is set to [0, 1], the fuzzy domain is set to {1, 2, 3, 4, 5}, and the quantization factor K is set to [0, 1]. pT Set to 5, the corresponding fuzzy subset in the fuzzy domain of the dry state judgment coefficient is {VS j , S j , M j , B j , VB j }, and describe the membership function of the fuzzy subset of the dry state judgment coefficient through the triangle membership function, where VS j 、S j 、M j 、B j ,VB j They represent the high and low dry state judgment coefficients respectively; According to the influence of main steam pressure and output power on dry state judgment coefficient, fuzzy control rules are formulated; After the predicted main steam pressure and output power are fuzzified by quantitative factors, the corresponding membership functions are searched and the corresponding fuzzy control rules are found in the fuzzy control rules. The fuzzy reasoning is performed using the Mamdani algorithm to obtain the fuzzy set of dry state judgment coefficients. The center of the fuzzy set is calculated using the center of gravity method as the dry state judgment coefficient.

9. A dry-wet state conversion control method for a thermal power unit based on weighted fuzzy predictive control according to claim 8, characterized in that: The fuzzy control rules are specifically: When the fuzzy value of the main steam pressure is in the fuzzy subset S pT And the fuzzy value of the output power is located in the fuzzy subset S NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset VS j ; When the fuzzy value of the main steam pressure is in the fuzzy subset S pT And the fuzzy value of the output power is located in the fuzzy subset M NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset S j ; When the fuzzy value of the main steam pressure is in the fuzzy subset S pT And the fuzzy value of the output power is in the fuzzy subset B NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset M j ; When the fuzzy value of the main steam pressure is in the fuzzy subset M pT And the fuzzy value of the output power is located in the fuzzy subset S NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset S j ; When the fuzzy value of the main steam pressure is in the fuzzy subset M pT And the fuzzy value of the output power is located in the fuzzy subset M NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset M j ; When the fuzzy value of the main steam pressure is in the fuzzy subset M pT And the fuzzy value of the output power is in the fuzzy subset B NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset B j ; When the fuzzy value of the main steam pressure is in the fuzzy subset B pT And the fuzzy value of the output power is located in the fuzzy subset S NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset M j ; When the fuzzy value of the main steam pressure is in the fuzzy subset B pT And the fuzzy value of the output power is located in the fuzzy subset M NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset B j ; When the fuzzy value of the main steam pressure is in the fuzzy subset B pT And the fuzzy value of the output power is in the fuzzy subset B NE When the fuzzy value of the dry state judgment coefficient is located in the fuzzy subset VB j .

10. A dry-wet state conversion control system for thermal power units based on weighted fuzzy predictive control, characterized in that: include: A dry state model predictive controller, a wet state model predictive controller, a first LSTM model, a second LSTM model, a dry state fuzzy recognizer, and a wet state fuzzy recognizer; The dry state model prediction controller dynamically adjusts the operation data according to the difference between the control data in the actual operation process of the thermal power unit and the dry state control data set value to obtain the dry state operation data deviation value; The wet state model prediction controller dynamically adjusts the operation data according to the difference between the control data in the actual operation process of the thermal power unit and the wet state control data set value to obtain the wet state operation data deviation value; The first LSTM model predicts dry-state control data according to the dry-state operation data deviation value, and the second LSTM model predicts wet-state control data according to the wet-state operation data deviation value; The dry state fuzzy identifier performs fuzzy control according to the predicted dry state control data to obtain a dry state judgment coefficient; The wet state fuzzy identifier performs fuzzy control according to the predicted wet state control data to obtain a wet state judgment coefficient; The dry state judgment coefficient and the wet state judgment coefficient are used to weight the dry state operation data and the wet state operation data, the operation data is updated, and the updated operation data is used to perform dry-wet state conversion control of the thermal power unit.

Images