Thermal power plant water vapor quality monitoring method and system based on multi-mode situation awareness technology

Abstract

The invention discloses a thermal power plant water vapor quality monitoring method and system based on a multi-mode situation awareness technology, and relates to the technical field of thermal power plant water vapor quality monitoring, and the method comprises the following steps: deploying sensors at raw water, circulating water, demineralized water, condensed water, boiler water and steam of a thermal power plant based on a thermal power plant thermodynamic system design drawing; performing data acquisition on water vapor, packaging acquired data points into data frames, adding source addresses and timestamp information, and establishing a water vapor monitoring database for data recording; for equipment related to different water vapor collection positions in the current thermal power plant, equipment parameter adjustment values of historical equipment under the water vapor abnormal type event are collected, and a standard parameter adjustment value interval is calculated based on standard equipment to determine standard equipment adjustment parameters. According to the invention, by integrating historical data and real-time monitoring data, an equipment personalized compensation mechanism and an intelligent decision-making model are established, and accurate monitoring and efficient adjustment of water vapor abnormity are realized.

Description

Technical Field

[0001] This invention relates to the field of water and steam quality monitoring technology in thermal power plants, and more specifically, to a method and system for monitoring water and steam quality in thermal power plants based on multimodal situational awareness technology. Background Technology

[0002] The water and steam system of a thermal power plant is a key guarantee for the safe and stable operation of core equipment such as boilers and steam turbines. The quality of water and steam is directly related to the corrosion, scaling and salt accumulation of thermal equipment, which in turn affects the unit's operating efficiency, service life and safe production. If water and steam abnormalities occur and are not handled in a timely and accurate manner, they may lead to equipment failure and shutdown, or even cause major safety accidents. Traditional water vapor quality monitoring methods often employ standardized adjustment strategies, neglecting the individual differences among different operating equipment within thermal power plants. Due to factors such as equipment age, production batches, and operating environment, even when facing the same type of water vapor anomaly, the response characteristics and adjustment requirements of different equipment vary. Existing monitoring systems do not scientifically classify equipment states and lack mechanisms for establishing compensation coefficients for individual equipment. The "one-size-fits-all" adjustment approach is prone to over- or under-adjustment. Furthermore, traditional methods often rely on manual experience or fixed thresholds for judgment, lacking in-depth analysis of historical adjustment patterns. This makes it impossible to accurately grasp the evolution trend and optimal handling strategies for different water vapor anomalies, resulting in insufficient scientific decision-making and delayed response when facing complex and ever-changing water vapor anomaly situations. Summary of the Invention

[0003] To address the problems in related technologies, this invention proposes a method and system for monitoring the water and steam quality of thermal power plants based on multimodal situational awareness technology, in order to overcome the aforementioned technical problems existing in the existing related technologies.

[0004] Therefore, the specific technical solution adopted by the present invention is as follows: A method and system for monitoring steam and water quality in thermal power plants based on multimodal situational awareness technology, the method comprising the following steps: S1. Based on the thermal power plant's thermal system design drawing, sensors are deployed at the raw water, circulating water, demineralized water, condensate, boiler water, and steam in the thermal power plant to collect water and steam data. The collected data points are packaged into data frames and the source address and timestamp information are added to establish a water and steam monitoring database for data recording. S2. For the equipment involved in different water and steam collection locations in the current thermal power plant, collect the historical equipment parameter adjustment values ​​under water and steam anomaly events, calculate the standard parameter adjustment value range based on the standard equipment to determine the standard equipment adjustment parameters, group the equipment status based on the historical equipment parameter adjustment values ​​and the standard equipment adjustment parameters, and calculate the parameter adjustment compensation coefficient for different groups of equipment. S3. Collect the dosage of medicines and the adjustment parameters of standard equipment under different types of water vapor anomalies in history. Construct a standard multimodal model based on reinforcement learning. Based on the water vapor data collected in the water vapor monitoring database, select the optimal action according to the current type of water vapor anomaly event and adjust the compensation coefficients in combination with the parameters of different grouped equipment to execute the output. At the same time, collect data again from the sensors after the action is executed to obtain a new state to update the standard multimodal model.

[0005] In a preferred embodiment, S1 includes the following steps: S11. Based on the current thermal power plant thermal system design drawings, deploy sensors at the raw water, circulating water, demineralized water, condensate, boiler water, and steam of the thermal power plant. The sensors include residual chlorine sensors, turbidity sensors, temperature sensors, pH sensors, acid-base meters, conductivity sensors, pressure sensors, dissolved oxygen sensors, silicate monitors, and phosphate analyzers. According to different water and steam monitoring requirements, water and steam data are collected through the sensors based on the set data acquisition frequency. S12. Pack the data collected by the sensor into data frames, define the source address of the data collection according to the sensor number, record the data collection time as a timestamp, sort and combine the source address and timestamp in sequence to form a data header, establish a water vapor monitoring database through MySQL, create a file for each water vapor, and record the data collected by the corresponding sensor.

[0006] In a preferred embodiment, S2 includes the following steps: S21. Statistically analyze the equipment involved in different water and steam collection locations in the current thermal power plant, record the equipment number, identify abnormal events of water and steam quality parameters, extract equipment adjustment records, and calculate the adjustment effect index of different equipment under different abnormal events. S22. Establish the standard parameter adjustment value range through the abnormal adjustment parameters of standard equipment. Based on the parameter adjustment values ​​of historical equipment and the standard parameter adjustment value range, group the equipment status of the current thermal power plant operation equipment and calculate the parameter adjustment compensation coefficient of different groups of equipment.

[0007] In a preferred embodiment, S21 includes the following steps: S211. Collect historical water and steam data from thermal power plants, filter out all periods when water and steam parameters exceed the standard from the historical data, and record the abnormal event type and start time. End time ; S212. Extract the start time for each type of abnormal event. Until the end time The adjustment values ​​of the relevant equipment parameters are used to calculate the adjustment efficiency value. ,in These represent the area of ​​the abnormal parameter curve deviating from the standard value before the adjustment action and the area of ​​the abnormal parameter curve deviating from the standard value after the adjustment action until the parameter returns to normal, respectively. This refers to the absolute value of the equipment parameter adjustment.

[0008] In a preferred embodiment, S22 includes the following steps: S221. For each type of abnormal event j, calculate the adjustment amount of the standard equipment group in historical abnormal events. For the adjustment amount, the lower quartile, upper quartile, and interquartile range of the standard group adjustment amount are used to define the standard parameter adjustment value range. ,in These represent the lower quartile, upper quartile, and interquartile range of the standard group's regulation values, respectively. The average regulation efficiency value is calculated based on the regulation efficiency values ​​within the standard parameter regulation value range. and average adjustment amount , as standard equipment adjustment parameters; S222. For equipment i in the current thermal power plant, calculate the average value of the adjustment amount under abnormal event j. This serves as the baseline adjustment value for the device to respond to such anomalies, while also calculating the average adjustment performance value of device i in anomaly type event j. To construct a two-dimensional feature vector [ ]; S223, K-Means clustering in two-dimensional space [ The system clusters all equipment into three groups: high-efficiency, normal, and low-efficiency. A comprehensive compensation coefficient is then calculated for each group. : ; in, These represent the reference adjustment values ​​for all devices within the group. The average value and the benchmark adjustment performance of all devices within the group The average value is recorded in the water vapor monitoring database as the comprehensive compensation coefficient for the corresponding equipment i under abnormal event j. .

[0009] In a preferred embodiment, step S3 includes the following steps: S31. Collect the dosage of medicine and standard equipment adjustment parameters under different types of water vapor anomalies in history. Construct a standard multimodal model based on reinforcement learning Q-learning. Obtain real-time water vapor data through the water vapor monitoring database. Select the optimal action based on the current type of water vapor anomaly and adjust the compensation coefficient in combination with the parameters of different grouped equipment to execute the output. Output the dosage of medicine and equipment adjustment parameters. S32. Collect water vapor data after the action is performed through sensors, and update the standard multimodal model based on the water vapor data under the new state.

[0010] In a preferred embodiment, S31 includes the following steps: S311. Construct a reinforcement learning model based on standard equipment, filter out similar anomaly events of standard equipment from historical data, and construct training samples for each event. The training samples include state S, action A, reward R, and new state S'. The Q-learning algorithm is used to learn the optimal action value function Q(S, A), which is the maximum expected cumulative reward that can be obtained after taking action A in state S. The Q value is iterated to gradually approach its true value in order to obtain a standard multimodal model. S312. Construct the current state vector using real-time water vapor data obtained from the water vapor monitoring database. , the current state In the input standard multimodal model, the model evaluates the Q-value of all actions and selects the action with the highest Q-value as the standard action. Standard actions Multiply by the comprehensive compensation coefficient under the current equipment group to output the final drug addition amount and equipment adjustment parameters under the final action.

[0011] In a preferred embodiment, S32 includes the following steps: S321. After the action is performed, the water vapor parameters are monitored by the sensor to obtain the new state. Calculate the actual reward obtained from this decision based on the actual results. ,Will The new sample is used as a new training sample to update the Q value through Q-learning.

[0012] A water and steam quality monitoring system for thermal power plants based on multimodal situational awareness technology includes a thermal power plant data acquisition module, an equipment characteristic analysis and compensation establishment module, an intelligent decision-making and closed-loop optimization module, and an effect evaluation and model update module. The thermal power plant data acquisition module, based on the thermal power plant thermal system design drawing, deploys sensors at the raw water, circulating water, demineralized water, condensate, boiler water, and steam of the thermal power plant to collect water and steam data, packages the collected data points into data frames and adds source address and timestamp information, and establishes a water and steam monitoring database for data recording. The equipment characteristic analysis and compensation establishment module collects historical equipment parameter adjustment values ​​under abnormal water and steam events for equipment involved in different water and steam collection locations in the current thermal power plant. Based on the standard equipment, it calculates the standard parameter adjustment value range to determine the standard equipment adjustment parameters. Based on the parameter adjustment values ​​of the historical equipment and the standard equipment adjustment parameters, it groups the equipment status and calculates the parameter adjustment compensation coefficients for different groups of equipment. The intelligent decision-making and closed-loop optimization module collects the amount of medicine added and the standard equipment adjustment parameters under different types of water vapor anomalies in history. It constructs a standard multimodal model based on reinforcement learning, and selects the optimal action based on the water vapor data collected in the water vapor monitoring database, according to the current type of water vapor anomaly event, and performs the output by adjusting the compensation coefficients in combination with the parameters of different grouped equipment. The effect evaluation and model update module collects data again from the sensors after the action is performed to obtain a new state and update the standard multimodal model.

[0013] The beneficial effects of this invention are as follows: 1. This invention collects historical data on water and steam anomalies in thermal power plants, and statistically analyzes historical adjustment data of standard groups and different operating equipment based on different types of water and steam anomaly events. Based on the standard group data, it classifies the current operating equipment of thermal power plants into states and establishes compensation coefficients, thereby enabling localized monitoring and adjustment of water and steam anomalies in thermal power plants. Based on the parameter adjustment values ​​of historical equipment and the adjustment parameters of standard equipment, it groups the equipment states and calculates the parameter adjustment compensation coefficients of different groups of equipment. This allows for a more accurate understanding of the state and response of different equipment under water and steam anomaly conditions, providing a basis for subsequent targeted adjustments and enhancing practicality. 2. By collecting historical adjustment data of standard groups and operating equipment under different types of water vapor anomaly events, this invention can fully take into account the differences in the actual operating conditions of various equipment in thermal power plants. Different equipment may behave differently when facing water vapor anomalies due to factors such as service life, production batch, and operating environment. Based on this, the operating equipment is classified into states and compensation coefficients are established, which can realize precise adjustment according to the individual characteristics of each piece of equipment, avoiding the problems of over-adjustment or under-adjustment that may be caused by a one-size-fits-all adjustment method. 3. This invention collects the dosage of medicines and the adjustment parameters of standard equipment under different types of historical water vapor anomalies. Through reinforcement learning models, it can learn the best handling strategies for different anomalies. It analyzes and makes decisions in real time based on data from the water vapor monitoring database, and combines the compensation coefficients of different equipment under different types of anomalies to output decisions adapted to the current equipment, thus enhancing functionality. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a flowchart of a method for monitoring the water and steam quality of a thermal power plant based on multimodal situational awareness technology according to an embodiment of the present invention; Figure 2 This is a block diagram of a water and steam quality monitoring system for thermal power plants based on multimodal situational awareness technology, according to an embodiment of the present invention. Detailed Implementation

[0016] To further illustrate the various embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. These drawings are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these drawings, those skilled in the art should be able to understand other possible implementation methods and the advantages of the present invention. The components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0017] According to an embodiment of the present invention, a method and system for monitoring the water and steam quality of thermal power plants based on multimodal situational awareness technology is provided.

[0018] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments: Example 1: like Figure 1 As shown, according to an embodiment of the present invention, a method for monitoring the steam and water quality of a thermal power plant based on multimodal situational awareness technology includes the following steps: S1. Based on the thermal power plant's thermal system design drawing, sensors are deployed at the raw water, circulating water, demineralized water, condensate, boiler water, and steam in the thermal power plant to collect water and steam data. The collected data points are packaged into data frames and the source address and timestamp information are added to establish a water and steam monitoring database for data recording. S11. Based on the current thermal power plant thermal system design drawings, deploy sensors at the raw water, circulating water, demineralized water, condensate, boiler water, and steam of the thermal power plant. The sensors include residual chlorine sensors, turbidity sensors, temperature sensors, pH sensors, acid-base meters, conductivity sensors, pressure sensors, dissolved oxygen sensors, silicate monitors, and phosphate analyzers. According to different water and steam monitoring requirements, water and steam data are collected through the sensors based on the set data acquisition frequency. It should be noted that different types of water vapor have different variation characteristics and monitoring requirements. Therefore, a reasonable data acquisition frequency needs to be set. Furthermore, based on the characteristics and monitoring needs of different types of water vapor, appropriate sensors need to be selected and deployed. Raw water is typically taken from the water source at the inlet of the power plant; circulating water is taken at the inlet and outlet of the cooling tower; demineralized water is taken at the inlet and outlet of the desalination equipment; condensate is taken at the condenser outlet; boiler water is taken inside the boiler drum; and steam is taken at the superheater outlet. The specific locations need to be determined based on the current thermal system design drawings of the power plant. See Table 1 for specific sensor selection. Table 1: List of Sampling Points for Monitoring ; ;

[0019] S12. Pack the data collected by the sensor into data frames, define the source address of the data collection according to the sensor number, record the data collection time as a timestamp, sort and combine the source address and timestamp in sequence to form a data header, establish a water vapor monitoring database through MySQL, create a file for each water vapor, and record the data collected by the corresponding sensor.

[0020] It should be noted that by packaging the collected data into data frames, the collected data can be easily transmitted. By receiving and parsing the data frames, the data collected by the sensor can be obtained.

[0021] S2. For the equipment involved in different water and steam collection locations in the current thermal power plant, collect the historical equipment parameter adjustment values ​​under water and steam anomaly events, calculate the standard parameter adjustment value range based on the standard equipment to determine the standard equipment adjustment parameters, group the equipment status based on the historical equipment parameter adjustment values ​​and the standard equipment adjustment parameters, and calculate the parameter adjustment compensation coefficient for different groups of equipment. S21. Statistically analyze the equipment involved in different water and steam collection locations in the current thermal power plant, record the equipment number, identify abnormal events of water and steam quality parameters, extract equipment adjustment records, and calculate the adjustment effect index of different equipment under different abnormal events. S211. Collect historical water and steam data from thermal power plants, filter out all periods when water and steam parameters exceed the standard from the historical data, and record the abnormal event type and start time. End time ; S212. Extract the start time for each type of abnormal event. Until the end time The adjustment values ​​of the relevant equipment parameters are used to calculate the adjustment efficiency value. ,in These represent the area of ​​the abnormal parameter curve deviating from the standard value before the adjustment action and the area of ​​the abnormal parameter curve deviating from the standard value after the adjustment action until the parameter returns to normal, respectively. This refers to the absolute value of the equipment parameter adjustment.

[0022] It should be noted that the recorded anomaly types, such as "boiler water pH is lower than the set threshold" and "steam temperature is higher than the set threshold," are recorded as parameter adjustment values ​​for the relevant equipment used to correct the anomaly. For example, for the "boiler water pH is too low" event, the frequency or stroke adjustment of the "phosphate dosing pump" during this period is recorded. A larger E value indicates a better corrective effect per unit adjustment, signifying higher equipment efficiency and better equipment condition. The area under the curve representing the abnormal parameter deviation from the standard value before the adjustment action is taken indicates the total cumulative damage caused to the system by the abnormality before the adjustment action was taken. It is the product of the degree of deviation and the duration of the abnormality, reflecting not only how far the abnormality deviates from the standard value, but also how long the abnormality lasts. It is the area of ​​the abnormal parameter curve that deviates from the standard value after the adjustment action until the parameter returns to normal. It measures the efficiency and quality of the adjustment action. When classifying abnormal events and collecting historical data, it is necessary to construct a historical dataset with analytical value based on the numerical value of the abnormality. It is necessary to select events with the same abnormality type and similar severity to form historical abnormal event data to ensure that all subsequent analyses are carried out within the same abnormality group and to ensure that subsequent compensation is effective.

[0023] S22. Establish the standard parameter adjustment value range through the abnormal adjustment parameters of standard equipment. Based on the parameter adjustment values ​​of historical equipment and the standard parameter adjustment value range, group the equipment status of the current thermal power plant operating equipment, and calculate the parameter adjustment compensation coefficient of different groups of equipment. S221. For each type of abnormal event j, calculate the adjustment amount of the standard equipment group in historical abnormal events. For the adjustment amount, the lower quartile, upper quartile, and interquartile range of the standard group adjustment amount are used to define the standard parameter adjustment value range. ,in These represent the lower quartile, upper quartile, and interquartile range of the standard group's regulation values, respectively. The average regulation efficiency value is calculated based on the regulation efficiency values ​​within the standard parameter regulation value range. and average adjustment amount , as standard equipment adjustment parameters; It should be noted that the standard equipment group consists of equipment that is stable in performance, recently put into operation or has undergone major overhaul and optimization, and is recognized as being in the best operating condition. The selection should be based on the different equipment in the current thermal power plant and the equipment manufacturers, and should be made by consulting experts in the field based on experience.

[0024] S222. For equipment i in the current thermal power plant, calculate the average value of the adjustment amount under abnormal event j. This serves as the baseline adjustment value for the device to respond to such anomalies, while also calculating the average adjustment performance value of device i in anomaly type event j. To construct a two-dimensional feature vector [ ]; S223, K-Means clustering in two-dimensional space [ The system clusters all equipment into three groups: high-efficiency, normal, and low-efficiency. A comprehensive compensation coefficient is then calculated for each group. : ; in, These represent the reference adjustment values ​​for all devices within the group. The average value and the benchmark adjustment performance of all devices within the group The average value is recorded in the water vapor monitoring database as the comprehensive compensation coefficient for the corresponding equipment i under abnormal event j. .

[0025] It should be noted that K-Means clustering is used to cluster all devices in a two-dimensional space, dividing the devices into high-efficiency, normal, and low-efficiency groups, with the high-efficiency group being a subset of the low-efficiency group. high This means that a small dosage yields good results, while the normal group is considered moderately effective. , The inefficient group is the high Low ,exist middle These are compensation based on adjustment amount and compensation based on performance. If the group of devices requires a larger adjustment amount on average, the command needs to be amplified so that the devices can achieve a better adjustment effect. Conversely, if the group of devices requires a smaller adjustment amount on average, the command needs to be reduced. The same applies to performance.

[0026] Example 2: S3. Collect the dosage of medicines and the adjustment parameters of standard equipment under different types of water vapor anomalies in history. Construct a standard multimodal model based on reinforcement learning. Based on the water vapor data collected in the water vapor monitoring database, select the optimal action according to the current type of water vapor anomaly event and adjust the compensation coefficients in combination with the parameters of different grouped equipment to execute the output. At the same time, collect data again from the sensors after the action is executed to obtain a new state to update the standard multimodal model. S31. Collect the dosage of medicine and standard equipment adjustment parameters under different types of water vapor anomalies in history. Construct a standard multimodal model based on reinforcement learning Q-learning. Obtain real-time water vapor data through the water vapor monitoring database. Select the optimal action based on the current type of water vapor anomaly and adjust the compensation coefficient in combination with the parameters of different grouped equipment to execute the output. Output the dosage of medicine and equipment adjustment parameters. S311. Construct a reinforcement learning model based on standard equipment, filter out similar anomaly events of standard equipment from historical data, and construct training samples for each event. The training samples include state S, action A, reward R, and new state S'. The Q-learning algorithm is used to learn the optimal action value function Q(S, A), which is the maximum expected cumulative reward that can be obtained after taking action A in state S. The Q value is iterated to gradually approach its true value in order to obtain a standard multimodal model. It should be noted that state S is the set of all relevant water vapor parameters at the time of the anomaly, forming a multimodal state vector. Action A represents the measures taken by operators historically, including the dosage of chemicals, equipment adjustment parameters, and rewards. The reward value is calculated after the fact, where These represent the reduction in abnormal area before and after the action, the drug cost of the action, and the equipment wear and tear, respectively. The quantification of wear and tear needs to be based on empirical methods or by using a machine learning model (LSTM) to predict the remaining useful life through continuous monitoring of the equipment's operating data. The weights are used to balance the influence of each parameter, and the cumulative value is 1. The new state S' is the new state of the water vapor system after the execution of action A.

[0027] It should be noted that the Q-table is updated using the Q=learning algorithm, and the specific formula is as follows: ; Here, α controls the rate at which new information overwrites old knowledge; it needs to be set by consulting experts in the field, and its range is 0 < α ≤ 1. γ measures the importance of future rewards; it also needs to be set by consulting experts in the field, and its range is 0 < γ < 1. Let Q be the maximum Q value among all possible actions in the next state S'.

[0028] S312. Construct the current state vector using real-time water vapor data obtained from the water vapor monitoring database. , the current state In the input standard multimodal model, the model evaluates the Q-value of all actions and selects the action with the highest Q-value as the standard action. Standard actions Multiply by the comprehensive compensation coefficient under the current equipment group to output the final drug addition amount and equipment adjustment parameters under the final action.

[0029] It should be noted that the comprehensive compensation coefficient needs to be selected in conjunction with the current abnormal event type. By determining which abnormal event type the current state belongs to, the comprehensive compensation coefficient of the current device under the current abnormal event type can be obtained. In other words, it determines which abnormal event type the abnormal value of the parameter in the current state falls within.

[0030] S32. Collect water vapor data after the action is performed using sensors, and update the standard multimodal model based on the water vapor data under the new state. S321. After the action is performed, the water vapor parameters are monitored by the sensor to obtain the new state. Calculate the actual reward obtained from this decision based on the actual results. ,Will The new sample is used as a new training sample to update the Q value through Q-learning.

[0031] It should be noted that the performance of each device needs to be reassessed periodically, and the comprehensive compensation coefficient needs to be recalculated based on the new operating data to ensure that it always accurately reflects the current status of the device. This is usually done every 3 months, but can also be adjusted according to the actual application environment.

[0032] Example 3: like Figure 2 As shown, a water and steam quality monitoring system for thermal power plants based on multimodal situational awareness technology includes a thermal power plant data acquisition module, an equipment characteristic analysis and compensation establishment module, an intelligent decision-making and closed-loop optimization module, and an effect evaluation and model update module.

[0033] The thermal power plant data acquisition module, based on the thermal power plant thermal system design drawing, deploys sensors at the raw water, circulating water, demineralized water, condensate, boiler water, and steam of the thermal power plant to collect water and steam data, packages the collected data points into data frames and adds source address and timestamp information, and establishes a water and steam monitoring database for data recording; The equipment characteristic analysis and compensation module collects historical equipment parameter adjustment values ​​under abnormal water and steam events for equipment involved in different water and steam collection locations in the current thermal power plant. Based on the standard equipment, it calculates the standard parameter adjustment value range to determine the standard equipment adjustment parameters. Based on the parameter adjustment values ​​of the historical equipment and the standard equipment adjustment parameters, it groups the equipment status and calculates the parameter adjustment compensation coefficients for different groups of equipment. The intelligent decision-making and closed-loop optimization module collects the dosage of medicines and the adjustment parameters of standard equipment under different types of water vapor anomalies in history. It constructs a standard multimodal model based on reinforcement learning. Based on the water vapor data collected in the water vapor monitoring database, it selects the optimal action according to the current type of water vapor anomaly and performs the execution output by adjusting the compensation coefficients in combination with the parameters of different grouped equipment. The effect evaluation and model update module collects data from the sensors again after the action is performed to obtain a new state and update the standard multimodal model.

[0034] In summary, this invention collects historical data on water and steam anomalies in thermal power plants, and statistically analyzes historical adjustment data of standard groups and different operating equipment based on different types of water and steam anomaly events. Based on the standard group data, it classifies the current operating equipment of the thermal power plant into states and establishes compensation coefficients, thereby enabling site-specific monitoring and adjustment of water and steam anomalies in thermal power plants. By grouping equipment states based on historical equipment parameter adjustment values ​​and standard equipment adjustment parameters, and calculating parameter adjustment compensation coefficients for different groups of equipment, this invention provides a more accurate understanding of the state and response of different equipment under water and steam anomaly conditions, providing a basis for subsequent targeted adjustments and enhancing practicality. By collecting historical adjustment data of standard groups and operating equipment under different types of water vapor anomalies, the differences in the actual operating conditions of various equipment in thermal power plants can be fully considered. Different equipment may behave differently when facing water vapor anomalies due to factors such as service life, production batch, and operating environment. Based on this, the operating equipment is classified into states and compensation coefficients are established, which can achieve precise adjustment according to the individual characteristics of each piece of equipment, avoiding the problems of over- or under-adjustment that may be caused by a one-size-fits-all approach. By collecting the dosage of chemicals and the adjustment parameters of standard equipment under different types of water vapor anomalies in history, the best handling strategy corresponding to different anomalies can be learned through reinforcement learning models. Real-time analysis and decision-making are performed based on data in the water vapor monitoring database, and combined with the compensation coefficients of different equipment under different types of anomalies, the decision adapted to the current equipment is output, enhancing functionality.

[0035] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for monitoring water-steam quality in a thermal power plant based on multi-modal situation awareness technology, characterized in that, The method comprises the following steps: S1, based on the thermal system design drawing of the thermal power plant, deploying sensors at the raw water, circulating water, desalted water, condensate water, boiler water and steam of the thermal power plant to collect data of the water and steam, packing the collected data points into data frames and adding source addresses and timestamp information, establishing a water and steam monitoring database for data recording; S2, collecting the historical device parameter adjustment values of the devices involved in different water and steam collection positions in the current thermal power plant under water and steam abnormal type events, calculating the standard parameter adjustment value interval based on the standard device to determine the standard device adjustment parameter, grouping the device states based on the parameter adjustment values of the historical devices based on the standard device adjustment parameter, and calculating the parameter adjustment compensation coefficients of different grouped devices; S3, collecting the drug addition amount under different water and steam abnormal type events and the standard device adjustment parameter, constructing a standard multi-modal model based on reinforcement learning, selecting the optimal action based on the collected water and steam data in the water and steam monitoring database according to the current water and steam abnormal type event, and executing and outputting in combination with the parameter adjustment compensation coefficients of different grouped devices, and then collecting data again based on the sensors after executing the action to obtain a new state to update the standard multi-modal model.

2. The method of claim 1, wherein the method is characterized by, The S1 comprises the following steps: S11, deploying sensors at the raw water, circulating water, desalted water, condensate water, boiler water and steam of the thermal power plant through the current thermal system design drawing of the thermal power plant, wherein the sensors include residual chlorine sensors, turbidity sensors, temperature sensors, pH sensors, acid-base meters, conductivity sensors, pressure sensors, dissolved oxygen sensors, silicate monitors and phosphate analyzers, collecting water and steam data through the sensors based on the set data collection frequency according to the monitoring requirements of different water and steam; S12, packing the data collected by the sensors into data frames, defining the source address of the collected data according to the number of the sensors, recording the time of data collection as the timestamp, sequentially sorting and combining the source address and the timestamp into a data header, establishing a water and steam monitoring database through MySQL, and establishing a file for each water and steam to record the data collected by the corresponding sensors.

3. The method of claim 2, wherein the method is characterized by, The S2 comprises the following steps: S21, counting the devices involved in different water and steam collection positions in the current thermal power plant, recording the device number to identify water quality parameter abnormal events and extracting device adjustment records, and calculating the adjustment effect index of different devices under different abnormal events; S22, determining the standard parameter adjustment value interval through the abnormal adjustment parameter of the standard device, grouping the device states of the running devices in the current thermal power plant based on the parameter adjustment values of the historical devices in combination with the standard parameter adjustment value interval, and calculating the parameter adjustment compensation coefficients of different grouped devices.

4. The method of claim 3, wherein the method is characterized by, The S21 comprises the following steps: S211, collect the water vapor historical data of the thermal power plant, filter out all the time periods in which the water vapor parameters exceed the standard from the historical data, record the abnormal type event, start time , end time ; S212, extracting the start time for each abnormal type event to the end time of the parameter adjustment value of the related equipment, and calculating an adjustment performance value wherein respectively represent the area of the abnormal parameter curve deviating from the standard value before the adjustment action and the area of the abnormal parameter curve deviating from the standard value after the adjustment action until the parameter returns to normal, is the absolute value of the equipment parameter adjustment amount.

5. The method of claim 4, wherein the method is characterized by, The S22 comprises the following steps: S221. For each type of abnormal event j, the standard group of devices is calculated in the amount of adjustment in the history of abnormal events For the amount of adjustment, the lower quartile, the upper quartile and the interquartile range of the standard group adjustment are defined as the standard parameter adjustment value interval wherein represent the lower quartile, the upper quartile and the interquartile range of the standard group adjustment, respectively, the average adjustment effectiveness value is calculated based on the adjustment effectiveness values in the standard parameter adjustment value interval and the average adjustment amount as the standard device adjustment parameter; S222、for the current thermal power plant equipment i, calculate the average value of the adjustment amount under the abnormal type event j As the equipment coping with such an abnormality, the average value of the adjustment amount of the equipment i under the abnormal type event j is calculated, and the average value of the adjustment performance value of the equipment i under the abnormal type event j is calculated To construct a two-dimensional feature vector ] S223, K-Means clustering in two-dimensional space [ The system clusters all equipment into three groups: high-efficiency, normal, and low-efficiency. A comprehensive compensation coefficient is then calculated for each group. : ； wherein, respectively represent the average value of all the reference adjustment amounts of the devices within the group and the average value of all the reference adjustment efficiencies of the devices within the group , and the comprehensive compensation coefficient under the abnormal type event j of the device i in the water vapor monitoring database record .​ 6. The method of claim 5, wherein the method is characterized by, The S3 comprises the following steps: S31, collect the drug adding amount and the standard equipment adjustment parameter under different water vapor anomaly type events in history, construct a standard multimodal model according to the reinforcement learning Q-leaning, obtain the real-time water vapor data collected through the water vapor monitoring database, select the optimal action according to the current water vapor anomaly type event, and execute and output by combining the parameter adjustment compensation coefficient of different grouped equipment, and output the drug adding amount and the equipment adjustment parameter; S32, collect the water vapor data after the execution of the action through the sensor, and update the standard multimodal model based on the water vapor data in the new state.

7. The method of claim 6, wherein the method is based on a multi-modal situation awareness technique. The S31 comprises the following steps: S311, based on the standard equipment to build reinforcement learning model, from the historical data of the same type of abnormal events of standard equipment, for each event to build a training sample , the training sample includes state S, action A, reward R and new state S', using Q-learning algorithm to learn the optimal action value function Q(S, A), that is, the maximum expected cumulative reward that can be obtained after taking action A in state S, iteration makes the Q value estimation gradually approach its true value to obtain the standard multi-modal model; S312, real-time water vapor data obtained through the water vapor monitoring database is used to construct a state vector at the current time , the current state The standard multi-modal model inputs the standard action , the model evaluates the Q value of all actions, and selects the action with the highest Q value as the standard action , and multiplies the comprehensive compensation coefficient under the current device group to output the drug addition amount and device adjustment parameter under the final action.

8. The method of claim 7, wherein the method is characterized by, The S32 comprises the following steps: S321. After the action is performed, the water vapor parameters are monitored by the sensor to obtain the new state. Calculate the actual reward obtained from this decision based on the actual results. ,Will The new sample is used as a new training sample to update the Q value through Q-learning.

9. A thermal power plant water vapor quality monitoring system based on multi-modal situation awareness technology, characterized in that, The system adopts a thermal power plant water vapor quality monitoring method based on the multimodal situation awareness technology according to any one of claims 1-8, comprising a thermal power plant data acquisition module, an equipment characteristic analysis and compensation establishment module, an intelligent decision and closed-loop optimization module, and an effect evaluation and model updating module.

10. The water-steam quality monitoring system for thermal power plants based on multi-modal situation awareness technology according to claim 9, characterized in that: The thermal power plant data acquisition module is based on the thermal system design drawing of the thermal power plant, and sensors are deployed at the raw water, circulating water, desalted water, condensate water, boiler water and steam in the thermal power plant to collect data of the water vapor, package the collected data points into data frames and add source addresses and timestamp information, and establish a water vapor monitoring database for data recording; The equipment characteristic analysis and compensation establishment module collects the equipment parameter adjustment numerical value of the historical equipment under the water vapor anomaly type event, determines the standard equipment adjustment parameter based on the standard equipment calculation standard parameter adjustment numerical value interval, groups the equipment state based on the standard equipment adjustment parameter according to the historical equipment parameter adjustment numerical value, and calculates the parameter adjustment compensation coefficient of different grouped equipment; The intelligent decision and closed-loop optimization module collects the drug adding amount and the standard equipment adjustment parameter under different water vapor anomaly type events in history, constructs a standard multimodal model according to the reinforcement learning, and based on the water vapor data collected in the water vapor monitoring database, selects the optimal action according to the current water vapor anomaly type event, and executes and outputs by combining the parameter adjustment compensation coefficient of different grouped equipment; The effect evaluation and model updating module collects data again through the sensor after the execution of the action, obtains a new state to update the standard multimodal model.

Images