Boiler water quality prediction analysis method and system based on big data

By constructing a sequence of control behavior numbers and a set of spliced ​​water quality change fragments, combined with changes in heat load, the problem of lag in boiler water quality prediction and analysis in existing technologies has been solved, enabling accurate prediction and timely control of boiler water quality trends.

CN122024893AInactive Publication Date: 2026-05-12NANJING HAIBANG ENVIRONMENTAL PROTECTION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202610090136.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-02-10
Publication Date
2026-05-12
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Existing boiler water quality prediction and analysis methods are unable to capture continuous changes in water quality over short periods, resulting in missing key nodes in condition monitoring, untimely system response, adjustment errors, or delayed early warnings, which reduces the accuracy and timeliness of water quality control.

Method used

By acquiring the action signals of the blowdown and chemical dosing control devices during boiler operation, a sequence of control behavior numbers is constructed. Combined with conductivity and dissolved oxygen data recorded by boiler water monitoring, a set of water quality change splicing segments is established to map the relationship between control behavior and water quality changes. By introducing heat load change records, a boiler water quality trend prediction sequence is formed.

Benefits of technology

It realizes the correspondence between continuous recording of boiler water quality changes and control behavior, distinguishes the water quality response characteristics caused by different control operations, provides load background support for water quality changes, and forms a water quality trend prediction sequence corresponding to control behavior and load changes, thereby improving the accuracy and timeliness of water quality regulation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122024893A_ABST
    Figure CN122024893A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of boiler water quality analysis, in particular to a boiler water quality prediction analysis method and system based on big data, and the method comprises the steps: obtaining pollution discharge and dosing action signals, constructing a behavior number sequence, extracting a corresponding water quality measurement fragment, calibrating a change trend, classifying response behaviors, and synchronizing thermal load data to form a linkage track. And summarizing the trend and outputting a prediction sequence. According to the method, sewage discharge and dosing signals are identified, a control behavior number sequence is constructed, various operations are continuously recorded according to a time sequence, continuous measurement data segments of conductivity and dissolved oxygen are extracted in combination with the sequence, water quality changes and control actions are associated and presented, water quality response types are distinguished on the basis of the corresponding relation between trends and behavior types, and the water quality response accuracy is improved. A thermal load change record is introduced, time connection is carried out, a water quality change track with a load background is formed, and a water quality trend prediction sequence associated with a control behavior and load change is constructed by summarizing trend directions of all stages.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of boiler water quality analysis technology, and in particular to a method and system for predictive analysis of boiler water quality based on big data. Background Technology

[0002] The field of boiler water quality analysis technology involves the sampling, testing, and analysis of water used in boiler operation. Core aspects include water sample collection, methods for detecting ions and impurities in the water, data recording and processing, and water quality adjustment procedures based on the test results. It is widely used in industries such as power and chemicals to prevent scaling, corrosion, and other malfunctions caused by water quality issues, ensuring the safety and efficiency of equipment operation.

[0003] Traditional boiler water quality prediction and analysis methods and systems refer to approaches that predict water quality trends based on historical water quality data and operating parameters, using empirical judgment, statistical analysis, or regression calculations. The general process includes: periodically collecting water samples, laboratory testing of water components, summarizing historical and current data, setting early warning thresholds, and using simple models to compare and judge water quality changes. These methods are weak at identifying correlations between multiple variables and are difficult to adapt to the prediction needs under complex operating conditions.

[0004] Existing methods rely on interval sampling and static detection, which struggles to capture continuous water quality changes within short cycles during boiler operation, resulting in missing key nodes in condition monitoring. Relying on static data to determine trends lacks a mechanism for identifying water quality responses triggered by operational behavior, failing to pinpoint the triggering factors for these changes. In scenarios with frequent load fluctuations, fluctuations in parameters such as conductivity and dissolved oxygen are easily misinterpreted, leading to delayed system responses, adjustment errors, or delayed warnings, thus reducing the accuracy and timeliness of water quality control. Summary of the Invention

[0005] To address the technical problems existing in the prior art, this invention provides a method and system for predicting and analyzing boiler water quality based on big data. The technical solution is as follows: A method for predicting and analyzing boiler water quality based on big data includes the following steps: S1: Obtain the action signals of the blowdown control device and the dosing control device during boiler operation, extract the start-up changes of the blowdown valve and the changes in the dosing drive state, match each signal with the actual behavior, classify and number the behaviors according to the time sequence, and obtain the boiler control behavior number sequence. S2: Using the time of each behavior number in the boiler control behavior number sequence, extract the measurement content of conductivity and dissolved oxygen from the boiler water monitoring record, organize the relevant measurement values ​​before and after the behavior into time segments, associate the water quality change data according to the behavior number, and obtain a set of water quality change splicing segments; S3: Based on the spliced ​​water quality change segments, focus on the data trends of conductivity and dissolved oxygen, connect the parts where the change trend changes with the corresponding control behaviors, link each data segment to the behavior number and action type, and obtain the behavior response water quality change mapping table; S4: Based on the data segments aligned in the behavior response water quality change mapping table, extract the corresponding heat load data from the steam output record, connect the heat load change and water quality segment sequentially with time points as clues, and perform unified processing in combination with behavior identifiers to obtain the heat load guided jump linkage trajectory group. S5: Call the data and control behaviors of different boiler operation stages in the heat load guided jump linkage trajectory group one by one, collect the continuity direction of water quality measurement trend, and output it in linkage with control behavior. Use continuous data description method to show the change of boiler water quality status and obtain the boiler water quality trend prediction sequence.

[0006] As a further embodiment of the present invention, the boiler control behavior numbering sequence includes a number identifier, behavior type label, and behavior time sequence record; the water quality change splicing fragment set includes conductivity change curve, dissolved oxygen fluctuation segment, and water quality measurement time axis; the behavior response water quality change mapping table includes behavior response type, water quality change pattern, and water quality fluctuation intensity classification; the heat load guided jump linkage trajectory group includes heat load change segment, control behavior correspondence, and time linkage index; and the boiler water quality trend prediction sequence includes water quality change trend line, water quality state continuity characteristics, and control behavior association label.

[0007] As a further aspect of the present invention, the step of obtaining the boiler control behavior number sequence is as follows: S101: Acquire the action signals of the blowdown control device and the dosing control device during boiler operation, match the acquisition time with the signal status one by one, distinguish the state change process according to the signal source, extract the time point when the state of the blowdown valve changes and the time point when the dosing action status switches, and obtain the switching time of the blowdown and dosing actions. S102: Based on the switching time of the sewage discharge and chemical dosing actions, call the corresponding time period content in the boiler behavior data, extract the sewage discharge and chemical dosing action content that is close to the above time, confirm the correspondence by whether the time is close, and obtain the action time correspondence sequence. S103: Call the action time correspondence sequence, number them sequentially according to time, organize the numbers and corresponding sewage discharge or chemical dosing actions in time order, output the number sequence, and obtain the boiler control action sequence number set.

[0008] As a further aspect of the present invention, the step of obtaining the water quality change splicing fragment set is as follows: S201: Using the action time corresponding to each number in the boiler control action sequence number set, call up the conductivity and dissolved oxygen sequence recorded in the boiler water monitoring device at minute intervals, take fixed intervals forward and backward according to time, arrange the conductivity and dissolved oxygen data in time order, and number them to obtain a water quality sequence fragment set; S202: Based on the water quality sequence fragment set, call the number sequence in the boiler control action sequence number set, organize the conductivity and dissolved oxygen data of the corresponding actions, list them in parallel as time segments under the same number, filter out missing data items, and obtain the water quality fragment group corresponding to the action. S203: For each time segment in the water quality segment group corresponding to the action, arrange them in the order of number, and splice the two types of data, conductivity and dissolved oxygen, in chronological order, and use the action number as the connection index position to obtain the spliced ​​water quality change segment set.

[0009] As a further aspect of the present invention, the step of obtaining the behavior response water quality change mapping table is as follows: S301: Based on the conductivity and dissolved oxygen measurements corresponding to each group number in the water quality change splicing segment set, sequentially check the time series, compare the change direction between adjacent measurement results, identify the time position where the continuous trend first switches, and record the time and corresponding number to obtain the set of starting time of the change trend; S302: Call the set of starting times of the change trend, for each time position corresponding to the number item, search the action category corresponding to the number in the set of boiler control action sequence numbers, match each time with the action category, and merge them into a list arranged in order of number to obtain a list of categories corresponding to action numbers; S303: Based on the list of categories corresponding to the action numbers, the start time of the change trend corresponding to each action category is grouped together with the number into a unified table, the data positions are arranged according to the number order, the corresponding trend time points are sorted by behavior category, and a behavior response water quality change mapping table is obtained.

[0010] As a further aspect of the present invention, the step of obtaining the heat load-guided jump linkage trajectory group is as follows: S401: Based on the data segment corresponding to each number in the behavior response water quality change mapping table, call the boiler load data in the steam output record, arrange the heat load data of fixed intervals before and after the action time in chronological order, and import them into the list in numerical order to obtain the heat load corresponding time segment set; S402: Call the set of time segments corresponding to the heat load, and perform time alignment processing on the heat load time segment in each numbered item and its water quality data in the water quality change splicing segment set. The alignment method uses the same start time as the reference standard. Arrange the data of the two time segments in order of time to obtain the heat load and water quality joint segment group. S403: Based on the heat load and water quality joint segment group, the continuous jump positions of the heat load data in each group are matched with the change trend in the water quality data. The time points of the two types of data are jointly indexed and summarized in order of number to obtain the heat load guided jump linkage trajectory group.

[0011] As a further aspect of the present invention, the step of obtaining the boiler water quality trend prediction sequence is as follows: S501: Based on the water quality measurement trend of each numbered item in the heat load-guided jump linkage trajectory group, extract the continuous change direction of conductivity and dissolved oxygen, record the continuity range of the trend direction in the time series, match the corresponding number with the control action one by one, and obtain the water quality trend continuity segment set. S502: Call the set of water quality trend continuation segments, attach the control action category corresponding to each number to the position after the corresponding water quality continuation direction, form a water quality trend and action joint sequence divided by number, and organize them according to the number order to obtain the water quality corresponding sequence of control behavior. S503: Based on the water quality corresponding sequence of the control behavior, extract the continuous time period of water quality continuity direction in each sequence, classify them according to the control action category, and string together all the trend segments corresponding to the actions into a complete change direction path to obtain the boiler water quality trend prediction sequence.

[0012] A boiler water quality prediction and analysis system based on big data, the system comprising: The control behavior extraction module acquires the valve status signals of the boiler blowdown control device and the drive signals of the chemical dosing control device. By identifying the start and end times of the status changes, it locates the change point of each operation. Based on the sequence of signals, it numbers the operation behaviors and associates the numbers with the control actions to obtain the boiler control behavior number sequence. The water quality data splicing module calls up the behavior numbers and corresponding time points recorded in the boiler control behavior number sequence, extracts the continuous measurement records of conductivity and dissolved oxygen data from the boiler water monitoring device, and pairs the measurement content with the control behavior number according to the time relationship to obtain a set of spliced ​​water quality change segments. The water quality trend attribution module extracts the continuous change sequence of conductivity and the continuous change sequence of dissolved oxygen based on the measurement content in the water quality change splicing fragment set. It compares the data segments in the two types of measurement data that show directional changes, identifies the starting point of the change and the direction of trend continuation, and associates the corresponding change trend with the control behavior number to obtain the behavior response water quality change mapping table. The heat load linkage module calls the control behavior number information in the behavior response water quality change mapping table, retrieves the steam pressure value and steam flow rate value from the steam output related records, matches the data time period corresponding to the number, and merges the pressure and flow rate change direction with the corresponding water quality change trend according to the number to obtain the heat load guided jump linkage trajectory group. The trend prediction generation module extracts the conductivity and dissolved oxygen change trends under each number based on the water quality change trend and load change direction merged in the heat load-guided jump linkage trajectory group. It compares these trends with the load change direction and outputs the trend type under each numbered behavior in a centralized manner through the trend direction attribution method, thus obtaining the boiler water quality trend prediction sequence.

[0013] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: In this invention, by identifying sewage discharge and chemical dosing action signals and constructing a sequence of control behavior numbers, various control operations during boiler operation are continuously recorded and maintain a temporal correspondence. This sequence is then used to extract continuous measurement data segments of boiler water conductivity and dissolved oxygen, allowing water quality changes to be presented in relation to control actions within the same timeframe. Based on the correspondence between water quality trend changes and behavior categories, the water quality response characteristics triggered by different control operations can be distinguished. Furthermore, records of heat load changes on the steam output side are incorporated and temporally linked to the water quality measurement segments, providing load background support for water quality changes. By aggregating the water quality continuity direction at each operational stage, a water quality trend prediction sequence corresponding to control actions and load changes is formed. Attached Figure Description

[0014] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a flowchart illustrating the process of obtaining the boiler control behavior number sequence of the present invention. Figure 3 This is a flowchart illustrating the process of obtaining the water quality change splicing fragment set according to the present invention. Figure 4 This is a flowchart illustrating the process of obtaining the behavioral response water quality change mapping table of the present invention. Figure 5 This is a flowchart illustrating the acquisition of the heat load-guided jump linkage trajectory group of the present invention. Figure 6 This is a flowchart of the process for obtaining the boiler water quality trend prediction sequence of the present invention. Detailed Implementation

[0015] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0016] refer to Figures 1 to 6 A method for predicting and analyzing boiler water quality based on big data includes the following steps: S1: Acquire the action signals of the blowdown control device and the dosing control device during boiler operation, identify the start-up change point of the blowdown valve and the moment of change of the driving state of the dosing action, correspond each signal change with the actual behavior, list the numbers in the order of time, and obtain the boiler control behavior number sequence. S2: Using the behavior time corresponding to each number in the boiler control behavior number sequence, extract the continuous measurement content of conductivity and dissolved oxygen from the boiler water monitoring device, arrange the relevant measurement values ​​before and after the behavior into time segment, and link each behavior event with the water quality measurement segment in sequence to obtain a set of spliced ​​water quality change segments. S3: Based on the spliced ​​segments of water quality changes, examine the trend of the measured data one by one, observe the continuous change trend of conductivity and dissolved oxygen alternately, align the data segments where the change trend begins with the corresponding control behavior, use the number and behavior category to assign the data, and obtain the behavior response water quality change mapping table. S4: Based on the data segments aligned to the behavior response water quality change mapping table, extract relevant data on boiler load changes from the records of the steam output link, perform time synchronization processing on the heat load records corresponding to the control behavior, connect the heat load records with the water quality measurement segment in time, and obtain the heat load guidance jump linkage trajectory group. S5: Based on the data content and corresponding control behaviors of different boiler operation stages in the heat load-guided jump linkage trajectory group, process them one by one, collect the continuity direction of all water quality measurement trends, and output them together with the control behaviors. Use continuous data description to show the trend of boiler water quality status and obtain the boiler water quality trend prediction sequence.

[0017] The boiler control behavior number sequence includes a number identifier, behavior type label, and behavior time sequence record. The water quality change splicing fragment set includes conductivity change curve, dissolved oxygen fluctuation segment, and water quality measurement time axis. The behavior response water quality change mapping table includes behavior response type, water quality change pattern, and water quality fluctuation intensity classification. The heat load guidance jump linkage trajectory group includes heat load change segment, control behavior correspondence, and time linkage index. The boiler water quality trend prediction sequence includes water quality change trend line, water quality status continuity characteristics, and control behavior association label.

[0018] Please see Figure 2 The steps for obtaining the boiler control behavior number sequence are as follows: S101: Acquire the action signals of the blowdown control device and the dosing control device during boiler operation, match the acquisition time with the signal status one by one, distinguish the state change process according to the signal source, extract the time point when the state of the blowdown valve changes and the time point when the dosing action status switches, and obtain the switching time of the blowdown and dosing actions. For a gas-fired steam boiler (model SZS20-1.25-Q) in an industrial setting, a high-frequency data acquisition channel was built based on a distributed control system to achieve precise control behavior capture. The sampling frequency was set to 10Hz, meaning the status of the underlying equipment was polled and read every 100 milliseconds. The main data acquisition targets included the opening feedback signal of the surface drain valve PV-101 (responsible for water quality regulation) and the operating status feedback signal of the phosphate dosing metering pump P-201. The system standardized the analog or switch signals transmitted from these two physical devices into binary status values, with logic 0 representing the drain valve being fully closed or the dosing pump being stopped, and logic 1 representing the drain valve being open or the dosing pump being running. To capture transient changes, a rolling time window was created in the memory buffer to store and refresh the signal sequence within the most recent period in real time. During data processing, the program used a point-by-point scanning method to logically compare the signal status values ​​at two adjacent sampling times. Specifically, the occurrence of an action was identified by determining whether the status values ​​at two different time points were different. Once a millisecond-level state transition from 0 to 1 or from 1 to 0 is detected, the system immediately identifies that moment as the critical point for action switching and simultaneously parses the direction attribute of whether the action is on or off. For example, in the monitoring record of March 15, 2024, the system keenly captured the state jump of the drain valve signal at 10:15:05, and the switching of the dosing pump signal at 10:20:10. The system then stores these two absolute timestamps, accurate to the second, and their corresponding device tag numbers in a temporary cache array. Through this deep analysis and extraction of the underlying signal source and temporal state changes, the system obtains the switching moments of the drain and dosing actions, including the precise time points and the direction of the state jump.

[0019] S102: Based on the switching time of the sewage discharge and chemical dosing actions, call the corresponding time period content in the boiler behavior data, extract the sewage discharge and chemical dosing action content that is close to the above time, confirm the correspondence by whether the time is close, and obtain the action time correspondence sequence. Based on the switching times of blowdown and chemical dosing actions, the system accesses the operator event log table in the boiler's historical operation database via a data interface. This table details the issuance time and specific type of all manual operation commands or automatic logic trigger commands in the DCS system. Considering the physical delay between signal transmission from the control room to the field actuator and the return feedback signal, as well as slight clock deviations between different subsystems, the system sets a 5-second time matching tolerance. During execution, the system iterates through the switching time of each physical signal and searches the massive log table for all records within a 5-second range before and after that moment. It performs correlation matching by calculating the absolute difference between the log recording time and the signal switching time. If the difference between the time of a retrieved command record and the signal action time is the smallest and within the tolerance range, a correspondence is determined between the two. For example, for the blowdown action recorded at 10:15:05, the system retrieves the "Surface Blowdown Valve Open" command issued at 10:15:03 in the log, with only a 2-second difference. The system then strongly binds the semantic description in the log to the physical time point. If multiple similar records exist, the system selects the optimal match based on the principle of minimum time difference. Through this two-way verification mechanism, the system successfully transforms monotonous physical switch signals into operational events with clear business meanings, thereby establishing the instruction context for each action and obtaining the sequence of action-time correspondences.

[0020] S103: Call the action time correspondence sequence, number them sequentially according to time, organize the numbers and corresponding sewage discharge or chemical dosing actions in chronological order, output the number sequence, and obtain the boiler control action sequence number set.

[0021] The system invokes the matched and confirmed action time correspondence sequence, and strictly sorts all action items in ascending order according to the timestamp sequence to ensure that the timing logic of the control behavior is completely consistent with the actual operating sequence. A number generator is initialized, assigning a unique identifier to each sorted action item. The number format is set to a standardized string containing the date and serial number, such as ACT-20240315-001. In this process, the system does not simply assign numbers, but uses these numbers as a core index key to structurally integrate the corresponding sewage discharge or chemical dosing action type, the specific action direction (such as on or off), and the verified precise execution time, forming an ordered list structure led by the numbers. For example, the system marks the surface sewage discharge start operation executed at 10:15:05 as the first action in the sequence, the phosphate chemical dosing start operation executed at 10:20:10 as the second action, and so on, constructing a clear timeline of control behavior. Through this systematic organization and standardized numbering process in chronological order, the system outputs a set of number sequences containing complete control logic chains, thereby obtaining the boiler control action sequence number set.

[0022] Please see Figure 3 The steps for obtaining the water quality change splicing fragment set are as follows: S201: Using the action time corresponding to each number in the boiler control action sequence number set, call up the conductivity and dissolved oxygen sequence recorded in the boiler water monitoring device at minute intervals, take fixed intervals forward and backward according to time, arrange the conductivity and dissolved oxygen data in time sequence, and number them to obtain a water quality sequence fragment set; Using the execution time of each action number in the boiler control action sequence number set as the core benchmark, the system establishes a communication link with online water quality monitoring instruments through an industrial standard OPC interface, retrieving historical data recorded at minute intervals. For each independent action number, the system, according to a preset analysis window strategy, retrieves 10 minutes of background data backward from the action time as zero point and 30 minutes of response data forward, thus constructing a data interval covering the complete dynamic process before and after the action. Within this interval, the system focuses on retrieving the conductivity value of boiler water conductivity monitoring point AT-101 and the dissolved oxygen value of dissolved oxygen monitoring point AT-102. For example, for the blowdown action numbered ACT-001 with an execution time of 10:15:05, the system will automatically capture all water quality monitoring records from 10:05:00 to 10:45:00. The system strictly arranges the captured raw data according to the timestamp order and establishes separate storage structures for conductivity and dissolved oxygen sequences to ensure that the temporal sequence of the data is not destroyed, thereby obtaining a set of water quality sequence fragments containing the original fluctuation characteristics for each action number.

[0023] S202: Based on the water quality sequence fragment set, call the number sequence of the boiler control action sequence number set, organize the conductivity and dissolved oxygen data of the corresponding actions, list them in parallel as time segments under the same number, filter out missing data items, and obtain the water quality fragment group corresponding to the action. Based on the acquired water quality sequence fragments, the system initiates a data cleaning and preprocessing procedure to eliminate interference from instrument malfunctions or packet loss during transmission. The system scans each data point in the conductivity and dissolved oxygen sequences, checking for null values, non-numeric symbols, or obvious outliers. For cases with a small number of consecutive missing points (e.g., no more than three), the system uses linear interpolation logic to calculate the theoretical value at the intermediate time using valid values ​​before and after the missing segment. For example, it uses known values ​​from 10:24:00 and 10:28:00 to extrapolate the conductivity at the intermediate time. For fragments with severe consecutive missing points that prevent data reconstruction, the system implements a strict rejection strategy, marking the data corresponding to that action number as invalid. After completing all necessary data repair work, the system aligns the conductivity and dissolved oxygen data columns along a unified time axis and arranges them side-by-side in a matrix format. This ensures that each number contains two sets of time-synchronized and highly complete water quality parameters, thereby filtering out missing data items and obtaining the water quality fragment group corresponding to the action.

[0024] S203: For each time segment in the water quality segment group corresponding to the action, arrange them in the order of number, and splice the two types of data, conductivity and dissolved oxygen, in chronological order, and use the action number as the connection index position to obtain the spliced ​​water quality change segment set.

[0025] For each effective time segment in the water quality segment group corresponding to an action, after cleaning and processing, the system uniformly reassembles them globally according to the order of the action numbers. During this process, the system concatenates the conductivity and dissolved oxygen data matrices belonging to different action numbers sequentially, forming a long-range time series stream. To distinguish the influence range of different actions after concatenation, the system embeds each action number and its corresponding start and end time index as metadata into the connection position of the concatenated sequence, or retains segment markers in the data structure. Through this concatenation process, the originally discrete short-term water quality segments distributed before and after each action are linked into a continuous comprehensive curve that reflects the entire process of water quality evolution under multiple control operations of the boiler. This dataset not only completely preserves the relative temporal position and numerical fluctuation pattern within each independent segment, but also establishes the logical order between segments, thus obtaining a set of concatenated water quality change segments that facilitates long-term trend analysis.

[0026] Please see Figure 4 The steps for obtaining the behavioral response water quality change mapping table are as follows: S301: Based on the conductivity and dissolved oxygen measurements corresponding to each group number in the water quality change splicing segment set, check the time series in sequence, compare the change direction between adjacent measurement results, identify the time position where the continuous trend first switches, and record the time and corresponding number to obtain the set of starting time of the change trend; Based on the conductivity and dissolved oxygen measurements corresponding to each group of numbers in the water quality change splicing segment set, the system executes a refined trend reversal identification algorithm. The program sequentially traverses each data point in the time series, calculating the slope or difference between adjacent measurement results to quantify the rate of change of water quality parameters. The system sets specific rate of change thresholds as judgment criteria; for example, a change in conductivity exceeding a certain number of microsiemens per minute or a change in dissolved oxygen exceeding a certain number of micrograms per minute is considered a significant fluctuation. The system scans point by point from the moment the action is executed. When it finds that the rate of change of multiple consecutive points exceeds the set threshold and the direction is consistent, it determines that this moment is the key position where the continuous trend first changes. For example, the system analysis finds that 6 minutes after a sewage discharge action, the rate of decrease in conductivity data suddenly increases and remains stable; the system then identifies this 6th minute as the trend starting point. The system records these identified specific time points and their corresponding action numbers to obtain the set of trend start times.

[0027] S302: Call the set of starting times of the change trend, for each time position corresponding to the number item, search for the action category corresponding to the number in the set of boiler control action sequence numbers, match each time with the action category, and import them into a list arranged in order of number to obtain a list of categories corresponding to action numbers; The system retrieves the set of trend start times and uses the unique index number to search backwards within the boiler control action sequence number set for the specific action category corresponding to each time point. The system logically pairs each recorded trend start time with the action type (e.g., surface blowdown, phosphate dosing) to determine which physical operation triggered the water quality fluctuation. For example, the system associates the identified response start point "6 minutes after the action" with the "surface blowdown" action, and the response start point "15 minutes after the action" with the "phosphate dosing" action, while simultaneously recording the corresponding water quality index type. By combining abstract time point data with the specific meaning of process actions and integrating them into a unified structured list, the system constructs a detailed list containing action number, action category, response delay time, and corresponding index object—that is, the action number-corresponding category list.

[0028] S303: Based on the list of categories corresponding to action numbers, the start time of the change trend corresponding to each action category is grouped together with the number into a unified table. The data positions are arranged in order of number, and the corresponding trend time points are sorted by behavior category to obtain the behavior response water quality change mapping table.

[0029] Based on the matching information in the list of categories corresponding to action numbers, the system summarizes and displays the start time of the change trend corresponding to each action category along with the number in a unified analysis table. During the construction of this table, the system strictly arranges the data rows according to the number order and classifies the data by behavior category, thus clearly presenting the time lag patterns and directions of change between different control behaviors and water quality responses. For example, the table data intuitively reveals that surface discharge typically causes a decrease in conductivity after about 6 minutes, while deoxygenation and dosing actions lead to a decrease in dissolved oxygen content after about 4 minutes. Through this systematic classification and mapping relationship construction, the system successfully quantifies the temporal characteristics of the impact of various control behaviors on water quality indicators and obtains a behavior-response water quality change mapping table.

[0030] Please see Figure 5 The steps for obtaining the heat load-guided jump linkage trajectory group are as follows: S401: Based on the data segment corresponding to each number in the behavior response water quality change mapping table, call the boiler load data in the steam output record, arrange the heat load data of fixed intervals before and after the action time in chronological order, import them into the list in numerical order, and obtain the set of time segments corresponding to the heat load. Based on the time period information corresponding to each number in the behavior response water quality change mapping table, the system accesses the historical database of the DCS system and retrieves the boiler load data recorded by the main steam flow transmitter. Following the same time window principle as the water quality data processing, the system takes the heat load values ​​of fixed intervals before and after each action's execution time (e.g., 10 minutes before the action to 30 minutes after), and organizes them into a sequence in chronological order. These collected minute-by-minute steam flow data are then imported into a data list in numerical order, establishing a one-to-one mapping with the previously acquired water quality data segments. Through this step, the system places the simple water quality change data within the context of specific boiler operating conditions, ensuring that subsequent analysis can fully consider the potential impact of heat load fluctuations on water quality, thereby obtaining a set of time segments corresponding to the heat load.

[0031] S402: Call the time segment set corresponding to the heat load, and perform time alignment processing on the heat load time segment in each numbered item and its water quality data in the water quality change splicing segment set. The alignment method uses the consistency of the start time as the reference standard. Arrange the data of the two time segments in order of time to obtain the heat load and water quality joint segment group. The system calls upon the time segment set corresponding to the heat load and performs strict time axis alignment processing on the heat load time period in each numbered item with its water quality data in the water quality change splicing segment set. To eliminate the differences in the absolute time of different actions, the system adopts a relative time standardization method, uniformly resetting the execution time of all segments to a relative zero point. Under this unified time coordinate system, the system arranges the heat load data column, conductivity, and dissolved oxygen data columns side by side according to time index, constructing a multi-dimensional attribute joint data structure in which each row contains "time-heat load-conductivity-dissolved oxygen". This fusion processing allows the system to intuitively compare the synchronicity of changes in operating load and fluctuations in water quality parameters at the same micro-time scale, thereby obtaining the heat load and water quality joint segment group.

[0032] S403: Based on the heat load and water quality joint segment group, the continuous jump positions of the heat load data in each group are matched with the change trend in the water quality data. The time points of the two types of data are jointly indexed and summarized in order of number to obtain the heat load guided jump linkage trajectory group.

[0033] Based on the combined heat load and water quality data segments, the system performs multivariate linkage analysis on each data set, focusing on the correlation between continuous jumps in heat load data and the trends in water quality data. The system sets a load change threshold (e.g., 5 tons / hour) and automatically detects any abrupt changes exceeding this threshold in the load sequence. Once a load jump is detected, the system immediately calculates the time difference between the jump moment and the start of the water quality trend in the same data set. If the time interval is very short (e.g., less than 3 minutes), the system determines, based on causal logic, that the abnormal water quality change is a linkage effect guided by a sudden change in heat load, rather than simply caused by chemical dosing or sewage discharge. The system records these special operating conditions, verified by location pairing and joint indexing, and summarizes them by number, thereby identifying which water quality fluctuations are dominated by load disturbances and obtaining heat load-guided jump linkage trajectory sets.

[0034] Please see Figure 6 The steps for obtaining the boiler water quality trend prediction sequence are as follows: S501: Based on the water quality measurement trend of each numbered item in the heat load-guided jump linkage trajectory group, extract the continuous change direction of conductivity and dissolved oxygen, record the continuity range of the trend direction in the time series, match the corresponding number with the control action one by one, and obtain the water quality trend continuity segment set. Based on the water quality measurement trends of each numbered item in the heat load-guided jump linkage trajectory group, the system further quantifies and extracts the dynamic morphology of water quality parameters, focusing on analyzing the continuous change direction and amplitude characteristics of conductivity and dissolved oxygen. The system calculates the linear fitting slope of the data within the response interval. If the slope value is positive and exceeds a set threshold, it is defined as an upward segment; if it is negative and exceeds the threshold, it is defined as a downward segment; otherwise, it is considered a stable segment. Simultaneously, the system records the duration and start-end range of this specific trend direction in the time series. For example, it identifies a continuous 14-minute downward trend in conductivity after a certain action. The system maps these quantified morphological feature data to the corresponding action numbers and specific control action types, constructing a dataset that finely describes the dynamic response characteristics of water quality, thereby obtaining a set of water quality trend continuation segments.

[0035] S502: Call the water quality trend continuation segment set, attach the control action category corresponding to each number to the position after the corresponding water quality continuation direction, form a water quality trend and action joint sequence divided by number, and organize it according to the number order to obtain the water quality corresponding sequence of control behavior. The system retrieves the set of water quality trend continuation segments and performs semantic tag fusion, appending the control action category corresponding to each number as a prefix tag to the end of the description of the corresponding water quality continuation direction. Specifically, if a synchronous jump in heat load is detected in the preceding disturbance identification step, the system will simultaneously label it with "load disturbance" information to distinguish between normal and disturbance responses. In this way, the system generates descriptors containing complete causal logic, such as "surface discharge leads to a decrease in conductivity" or "phosphate dosing leads to an increase in conductivity fluctuations under load disturbances." The system organizes and sorts these descriptors containing logical causal relationships according to their numbering order, transforming the numerical analysis results into a highly readable logical sequence, thereby obtaining the water quality corresponding sequence of control behaviors.

[0036] S503: Based on the water quality corresponding sequence of the control behavior, extract the continuous time period of water quality continuity direction in each sequence, classify them according to the control action category, and string together all the trend segments corresponding to the actions into a complete change direction path to obtain the boiler water quality trend prediction sequence.

[0037] Based on the water quality corresponding sequences of control actions, the system extracts continuous time periods of water quality continuity in each sequence and categorizes and reconstructs paths according to the type of control action. The system employs a path concatenation algorithm, focusing on trend segments generated by similar actions (such as multiple consecutive discharges). If the ending state of the previous segment is numerically and temporally continuous with the starting state of the next segment, the system connects these segments end-to-end, piecing them together into a complete evolution trajectory. In this way, the system integrates originally discrete short-term response segments into a complete path reflecting long-term trends. This path characterizes the expected evolution of boiler water quality over a future period under the continuous action of a specific control strategy. This output set of sequence data containing predictive information—the boiler water quality trend prediction sequence—provides a reliable benchmark reference for subsequent model predictive control.

[0038] A boiler water quality prediction and analysis system based on big data, the system includes: The control behavior extraction module acquires the valve status signals of the boiler blowdown control device and the drive signals of the chemical dosing control device. By identifying the start and end times of the status changes, it locates the change point of each operation. Based on the sequence of signals, it numbers the operation behaviors and associates the numbers with the control actions to obtain the boiler control behavior number sequence. The water quality data splicing module calls up the behavior numbers and corresponding time points recorded in the boiler control behavior number sequence, extracts the continuous measurement records of conductivity and dissolved oxygen data from the boiler water monitoring device, and pairs the measurement content with the control behavior number according to the time relationship to obtain a set of spliced ​​water quality change segments. The water quality trend attribution module extracts the continuous change sequence of conductivity and the continuous change sequence of dissolved oxygen based on the measurement content in the water quality change splicing fragment set. It compares the data segments in the two types of measurement data that show directional changes, identifies the starting point of the change and the direction of trend continuation, and associates the corresponding change trend with the control behavior number to obtain the behavior response water quality change mapping table. The heat load linkage module calls the control behavior number information in the behavior response water quality change mapping table, retrieves the steam pressure value and steam flow rate value from the steam output related records, matches the data time period corresponding to the number, and merges the pressure and flow rate change direction with the corresponding water quality change trend according to the number to obtain the heat load guided jump linkage trajectory group. The trend prediction generation module extracts the conductivity and dissolved oxygen change trends under each number based on the water quality change trend and load change direction merged in the heat load-guided jump linkage trajectory group. It compares these trends with the load change direction and outputs the trend type under each numbered behavior in a centralized manner through the trend direction attribution method, thus obtaining the boiler water quality trend prediction sequence.

[0039] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for predicting and analyzing boiler water quality based on big data, characterized in that, Includes the following steps: S1: Obtain the action signals of the blowdown control device and the dosing control device during boiler operation, extract the start-up changes of the blowdown valve and the changes in the dosing drive state, match each signal with the actual behavior, classify and number the behaviors according to the time sequence, and obtain the boiler control behavior number sequence. S2: Using the time of each behavior number in the boiler control behavior number sequence, extract the measurement content of conductivity and dissolved oxygen from the boiler water monitoring record, organize the relevant measurement values ​​before and after the behavior into time segments, associate the water quality change data according to the behavior number, and obtain a set of water quality change splicing segments; S3: Based on the spliced ​​water quality change segments, focus on the data trends of conductivity and dissolved oxygen, connect the parts where the change trend changes with the corresponding control behaviors, link each data segment to the behavior number and action type, and obtain the behavior response water quality change mapping table; S4: Based on the data segments aligned in the behavior response water quality change mapping table, extract the corresponding heat load data from the steam output record, connect the heat load change and water quality segment sequentially with time points as clues, and perform unified processing in conjunction with behavior identifiers to obtain the heat load-guided jump linkage trajectory group.

2. The boiler water quality prediction and analysis method based on big data according to claim 1, characterized in that: The boiler control behavior number sequence includes a number identifier, behavior type label, and behavior time sequence record; the water quality change splicing fragment set includes conductivity change curve, dissolved oxygen fluctuation segment, and water quality measurement time axis; the behavior response water quality change mapping table includes behavior response type, water quality change mode, and water quality fluctuation intensity classification; and the heat load guided jump linkage trajectory group includes heat load change segment, control behavior correspondence, and time linkage index.

3. The boiler water quality prediction and analysis method based on big data according to claim 1, characterized in that: The steps for obtaining the boiler control behavior number sequence are as follows: S101: Acquire the action signals of the blowdown control device and the dosing control device during boiler operation, match the acquisition time with the signal status one by one, distinguish the state change process according to the signal source, extract the time point when the state of the blowdown valve changes and the time point when the dosing action status switches, and obtain the switching time of the blowdown and dosing actions. S102: Based on the switching time of the sewage discharge and chemical dosing actions, call the corresponding time period content in the boiler behavior data, extract the sewage discharge and chemical dosing action content that is close to the above time, confirm the correspondence by whether the time is close, and obtain the action time correspondence sequence. S103: Call the action time correspondence sequence, number them sequentially according to time, organize the numbers and corresponding sewage discharge or chemical dosing actions in time order, output the number sequence, and obtain the boiler control action sequence number set.

4. The boiler water quality prediction and analysis method based on big data according to claim 1, characterized in that: The steps for obtaining the water quality change splicing fragment set are as follows: S201: Using the action time corresponding to each number in the boiler control action sequence number set, call up the conductivity and dissolved oxygen sequence recorded in the boiler water monitoring device at minute intervals, take fixed intervals forward and backward according to time, arrange the conductivity and dissolved oxygen data in time order, and number them to obtain a water quality sequence fragment set; S202: Based on the water quality sequence fragment set, call the number sequence in the boiler control action sequence number set, organize the conductivity and dissolved oxygen data of the corresponding actions, list them in parallel as time segments under the same number, filter out missing data items, and obtain the water quality fragment group corresponding to the action. S203: For each time segment in the water quality segment group corresponding to the action, arrange them in the order of number, and splice the two types of data, conductivity and dissolved oxygen, in chronological order, and use the action number as the connection index position to obtain the spliced ​​water quality change segment set.

5. The boiler water quality prediction and analysis method based on big data according to claim 1, characterized in that: The steps for obtaining the behavior response water quality change mapping table are as follows: S301: Based on the conductivity and dissolved oxygen measurements corresponding to each group number in the water quality change splicing segment set, sequentially check the time series, compare the change direction between adjacent measurement results, identify the time position where the continuous trend first switches, and record the time and corresponding number to obtain the set of starting time of the change trend; S302: Call the set of starting times of the change trend, for each time position corresponding to the number item, search the action category corresponding to the number in the set of boiler control action sequence numbers, match each time with the action category, and merge them into a list arranged in order of number to obtain a list of categories corresponding to action numbers; S303: Based on the list of categories corresponding to the action numbers, the start time of the change trend corresponding to each action category is grouped together with the number into a unified table, the data positions are arranged according to the number order, the corresponding trend time points are sorted by behavior category, and a behavior response water quality change mapping table is obtained.

6. The boiler water quality prediction and analysis method based on big data according to claim 1, characterized in that: The steps for obtaining the heat load-guided jump linkage trajectory group are as follows: S401: Based on the data segment corresponding to each number in the behavior response water quality change mapping table, call the boiler load data in the steam output record, arrange the heat load data of fixed intervals before and after the action time in chronological order, and import them into the list in numerical order to obtain the heat load corresponding time segment set; S402: Call the set of time segments corresponding to the heat load, and perform time alignment processing on the heat load time segment in each numbered item and its water quality data in the water quality change splicing segment set. The alignment method uses the same start time as the reference standard. Arrange the data of the two time segments in order of time to obtain the heat load and water quality joint segment group. S403: Based on the heat load and water quality joint segment group, the continuous jump positions of the heat load data in each group are matched with the change trend in the water quality data. The time points of the two types of data are jointly indexed and summarized in order of number to obtain the heat load guided jump linkage trajectory group.

7. The boiler water quality prediction and analysis method based on big data according to claim 1, characterized in that, The method further includes step S5: S5: Call the data and control behaviors of different boiler operation stages in the heat load guided jump linkage trajectory group one by one, collect the continuity direction of water quality measurement trend, output in linkage with control behavior, use continuous data description method to show the change of boiler water quality status, and obtain the boiler water quality trend prediction sequence. The boiler water quality trend prediction sequence includes a water quality change trend line, water quality status continuity characteristics, and control behavior association labels.

8. The boiler water quality prediction and analysis method based on big data according to claim 7, characterized in that: The steps for obtaining the boiler water quality trend prediction sequence are as follows: S501: Based on the water quality measurement trend of each numbered item in the heat load-guided jump linkage trajectory group, extract the continuous change direction of conductivity and dissolved oxygen, record the continuity range of the trend direction in the time series, match the corresponding number with the control action one by one, and obtain the water quality trend continuity segment set. S502: Call the set of water quality trend continuation segments, attach the control action category corresponding to each number to the position after the corresponding water quality continuation direction, form a water quality trend and action joint sequence divided by number, and organize them according to the number order to obtain the water quality corresponding sequence of control behavior. S503: Based on the water quality corresponding sequence of the control behavior, extract the continuous time period of water quality continuity direction in each sequence, classify them according to the control action category, and string together all the trend segments corresponding to the actions into a complete change direction path to obtain the boiler water quality trend prediction sequence.

9. A boiler water quality prediction and analysis system based on big data, characterized in that, The system is used in the boiler water quality prediction and analysis method based on big data as described in any one of claims 1-8, and the system comprises: The control behavior extraction module acquires the valve status signals of the boiler blowdown control device and the drive signals of the chemical dosing control device. By identifying the start and end times of the status changes, it locates the change point of each operation. Based on the sequence of signals, it numbers the operation behaviors and associates the numbers with the control actions to obtain the boiler control behavior number sequence. The water quality data splicing module calls up the behavior numbers and corresponding time points recorded in the boiler control behavior number sequence, extracts the continuous measurement records of conductivity data and dissolved oxygen data from the boiler water monitoring device, and pairs the measurement content with the control behavior number according to the time relationship to obtain a set of spliced ​​water quality change segments. The water quality trend attribution module extracts the continuous conductivity variation sequence and the continuous dissolved oxygen variation sequence based on the measurement content in the water quality change splicing fragment set. It compares the data segments in the two types of measurement data that show directional changes, identifies the starting point of the change and the direction of trend continuation, and associates the corresponding change trend with the control behavior number to obtain the behavior response water quality change mapping table. The heat load linkage module calls the control behavior number information in the behavior response water quality change mapping table, retrieves the steam pressure value and steam flow rate value from the steam output related records, matches the data time period corresponding to the number, and merges the pressure and flow rate change direction with the corresponding water quality change trend according to the number to obtain the heat load guided jump linkage trajectory group. The trend prediction generation module extracts the conductivity and dissolved oxygen change trends under each number based on the water quality change trend and load change direction merged in the heat load-guided jump linkage trajectory group. It compares these trends with the load change direction and outputs the trend type under each numbered behavior in a centralized manner through the trend direction attribution method, thus obtaining the boiler water quality trend prediction sequence.