Abnormal time period data processing method for performance assessment of thermal power generating unit

Abstract

The invention discloses a thermal power generating unit performance assessment-oriented abnormal time period data processing method. The method comprises the following steps of S1, data structure expansion; s2, abnormal time period configuration, wherein parameter configuration comprises specific data point selection, time period setting and processing mode and parameter setting; s3, writing and marking data, and after configuration is completed, writing an extended field value into the time sequence database according to a set rule; and S4, performance calculation logic integration, wherein performance calculation tasks for the corresponding sets in the specific assessment time period are triggered in a timed manner. The method at least has the following beneficial effects: by identifying and rejecting or replacing the data in the abnormal time period, the interference of data deviation caused by non-human factors such as equipment failure, control switching and test operation on the performance result is effectively avoided, and unfair scoring on teams or personnel is avoided; and the objectivity and the credibility of performance assessment are obviously improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of data processing, and particularly relates to an abnormal period data processing method for performance evaluation of a thermal power generating unit. BACKGROUND

[0002] In the process of operation management and performance evaluation of a thermal power generating unit, time series data collected by sensors, such as main steam temperature, coal consumption, power load and other key operation indexes, are widely used as evaluation basis. However, in actual operation environment, due to sudden faults, equipment maintenance, system switching or external operation, the data in some periods may deviate from the real working condition, and may appear severe fluctuation, obvious abnormality or even data loss.

[0003] The current mainstream evaluation system (such as the performance evaluation module based on the extension of DCS, SIS and the like) usually adopts a simple data aggregation algorithm (such as average value, integral value) to calculate the score, and defaults that the data in all time periods are valid, and lacks the identification and processing mechanism for abnormal periods. Such processing method has the following obvious problems:

[0004] 1. Abnormal data affects the fairness of evaluation

[0005] In the case of abnormal situations such as unplanned shutdown of the unit, switching of the control system, parameter mutation and the like, the data collected by the sensors may be temporarily distorted, and if directly involved in performance statistics, it will seriously interfere with the evaluation results, causing unfair evaluation of the unit or personnel.

[0006] 2. Unable to flexibly respond to emergencies

[0007] The existing system generally lacks a flexible abnormal processing rule configuration mechanism, and once a special event (such as abnormal coal quality, unit test and the like) occurs, the administrator usually needs to manually exclude the relevant data segment, which is low in efficiency and lacks audit traces.

[0008] 3. Split from performance evaluation logic

[0009] Abnormal data exclusion is usually completed independently by a data cleaning system, and cannot be linked with performance rules, making it difficult to realize the scene of "exemption based on evaluation logic", such as: a team should be exempted from responsibility during non-human fault of the equipment, but the system still includes the abnormal data in the statistics, which violates the actual business requirements. SUMMARY

[0010] The application provides an abnormal period data processing method for performance evaluation of a thermal power generating unit, aiming to solve the problems of abnormal data affecting the fairness of evaluation, inability to flexibly respond to emergencies and split from performance evaluation logic in the prior art.

[0011] An abnormal period data processing method for thermal power unit performance assessment, comprising the following steps:

[0012] S1: Data structure expansion, based on the point table structure of the database storing the thermal power unit operation time series data, a set of extension fields with "#application" as the suffix is added, which is used to store the abnormal processing related information of the point in the specified time period;

[0013] S2: Abnormal time period configuration, define abnormal data processing rules according to configuration parameters, configuration parameters include selecting specific data points, setting time period, and setting processing mode and parameters;

[0014] S3: Data writing and marking, according to the configuration parameters, according to the abnormal data processing rules, write the extension field value into the time series database;

[0015] S4: Performance calculation logic integration, when performing assessment calculation, query the corresponding extension field at the same time of obtaining original data from the time series database, and apply preprocessing rules for real-time preprocessing, the performance calculation engine uses the obtained data set to execute performance calculation formula, and outputs the assessment result.

[0016] Optionally, the extension field includes:

[0017] A. <point_name>#application_valid: indicates the validity of the specified time period data, the default value is 1, indicating that the point value data corresponding to the specified time period is valid and participates in the performance assessment calculation, if set to 0, it indicates that the point value data corresponding to the time stamp is invalid and is excluded in the performance assessment calculation;

[0018] B. <point_name>#application_mode: identifies the abnormal processing mode,

[0019] C. <point_name>#application_value: when the processing mode is fixed value filling or weight adjustment, store the replacement value or weight coefficient, and for other modes, the extension field is empty.

[0020] Optionally, the value of the abnormal processing identifier is as follows:

[0021] a. 0: not processed, the original point value will be directly used for calculation;

[0022] b. 1: reject, ignore the original point value of the specified time period;

[0023] c. 2: fixed value filling, use the value stored in the {point_name}#application_value field to replace the original point value for calculation;

[0024] d.3: Mean fill, which calculates a replacement value according to preset rules and uses it for calculation. The calculation result is pre-calculated and stored in #application_value;

[0025] e.4: Weight adjustment will read the weight coefficient stored in the {point_name}#application_value field, multiply the original point value by the weight, and then participate in the performance calculation formula.

[0026] Optionally, the execution process of S3 includes:

[0027] S3.1: Command generation. Based on the start and end times set by the user, generate a sequence of timestamps for all data to be written within the specified time period, generate one or more data records for each timestamp, and generate batch write commands.

[0028] S3.2: Data writing, execute the above instructions through the database batch write statement;

[0029] S3.3: Marking Results. After successful writing, at the database level, every data record for the target location within the specified time period has been marked with the extended field:

[0030] S3.4: For technical considerations, batch writing is adopted for rules with long time periods;

[0031] Conflict handling: If the same point in time is covered by multiple rules, a clear priority is defined and handled during writing.

[0032] Optionally, step S4 specifically includes the following steps:

[0033] S4.1: Performance Calculation Engine Construction;

[0034] S4.2: Calculation execution, performance calculation engine calculates the performance score = F(A,B,C) value within the time period [T_start,T_end].

[0035] Optionally, the process of S4.1 includes:

[0036] S4.1.1: Data acquisition, querying all raw data point values ​​required for calculation from the time series database;

[0037] S4.1.2: Data preprocessing, querying the extended field values ​​corresponding to these data points;

[0038] S4.1.3: Logical judgment and execution: Based on the value of the extended field, apply the defined rules to preprocess the original data;

[0039] S4.1.4: Calculation, using the preprocessed data to execute the performance calculation formula.

[0040] Optionally, S4.2 specifically includes:

[0041] S4.2.1: Construct the query; the engine sends a query request to the time-series database.

[0042] S4.2.2: Integrated Query and Preprocessing: The query logic embeds preprocessing rules, and the engine constructs a virtual value for each data point before actual calculation;

[0043] S4.2.3: The performance calculation engine uses the virtual values ​​A, B, and C generated by the above rules to substitute into the performance calculation formula to obtain the final score.

[0044] Optionally, the preprocessing rules include:

[0045] IFA#application_valid==0THEN

[0046] For this point in time, it is considered that there is no valid data at point A, and this point is ignored in aggregation calculations (such as summation and averaging).

[0047] ELSEIFA#application_mode==2THEN

[0048] Replace A.value with A#application_value, that is, the virtual value A = A#application_value;

[0049] ELSEIFA#application_mode==3THEN

[0050] Based on the preset mean calculation strategy, a value is calculated in real time to replace A.value, that is, the virtual value A = mean_value;

[0051] ELSEIFA#application_mode==4THEN

[0052] Adjust the weights of the original values, i.e., virtual value A = A.value * A#application_value;

[0053] ELSE (Mode is 0 or not configured)

[0054] Virtual value A = A.value / / Using the original value.

[0055] Compared with the prior art, this application has at least the following beneficial effects:

[0056] 1. Improve the accuracy and fairness of performance appraisals.

[0057] This application effectively avoids interference with performance results caused by data deviations due to non-human factors such as equipment failure, control switching, and experimental operations by identifying, eliminating, or replacing data during abnormal periods. This prevents unfair scoring of work teams or individuals and improves the objectivity and credibility of performance appraisals.

[0058] 2. Supports refined anomaly handling based on location and time period.

[0059] Compared to the traditional "one-size-fits-all" data cleaning approach, this application allows users to select specific sensor points and time intervals as needed, flexibly configure processing strategies, achieve precise "point-to-point and segment-to-segment" control, and meet the differentiated data correction needs under complex operating scenarios.

[0060] 3. Provides diverse anomaly data handling strategies to adapt to various business scenarios.

[0061] This application incorporates multiple anomaly handling modes, including elimination, fixed value filling, mean replacement, and weight adjustment, which can be flexibly selected according to different assessment indicators and business needs. This not only preserves the continuity of data analysis but also avoids the misleading nature of outliers, making the assessment results more business-adaptable.

[0062] 4. Achieve deep integration of anomaly data processing and performance scoring.

[0063] This application embeds abnormal data processing logic into the performance calculation process to ensure that all data involved in the scoring has been preprocessed, avoiding omissions or logical errors caused by the disconnect between data processing and scoring, achieving true "data quality-driven scoring", and enhancing the system's integration and closed-loop capabilities. Attached Figure Description

[0064] Figure 1 This is a flowchart illustrating a method for processing abnormal time period data in thermal power unit performance evaluation, as provided in one embodiment of this application. Detailed Implementation

[0065] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments.

[0066] This application provides a method for processing abnormal time period data for thermal power unit performance evaluation, including the following steps:

[0067] S1: Data Structure Expansion

[0068] Based on the point table structure in the database storing thermal power unit operation sequence data, a new set of extended fields with the suffix "#application" is added to store information related to anomaly handling for that point within a specified time period. These extended fields are stored in parallel with the original point value data, enabling the marking and configuration of handling strategies for anomaly data.

[0069] The specific fields include:

[0070] A.<point_name> #application_valid: Indicates the validity of the data at this moment. The default value is 1 (valid), which means that the point value data corresponding to this timestamp is valid and should be included in the performance evaluation calculation. If it is set to 0 (invalid), it means that the point value data corresponding to this timestamp is invalid and should be excluded from the performance evaluation calculation.

[0071] B.<point_name> #application_mode: Identifies the exception handling mode, with the following possible values:

[0072] a.0: No processing (default), the original point values ​​will be used directly for calculation.

[0073] b.1: Remove, which means ignoring the original value of that time point, equivalent to the data point not existing at that moment.

[0074] c.2: Fixed value filling, the value stored in the {point_name}#application_value field will be used to replace the original point value for calculation.

[0075] d.3: Mean imputation. A replacement value will be calculated based on preset rules (such as the mean of a time window before and after, or the historical mean of the same time period on the same day), and used for calculation. The calculation result can be pre-calculated and stored in #application_value, or calculated in real time by the performance calculation engine.

[0076] e.4: Weight adjustment will read the weight coefficient (usually a floating-point number between 0 and 1) stored in the {point_name}#application_value field, multiply the original point value by this weight, and then participate in the performance calculation formula.

[0077] C.<point_name> #application_value: Stores an alternative value or weight coefficient when the processing mode is fixed value fill or weight adjustment. For other modes, this field can be empty (NULL) or not used.

[0078] Specifically, when #application_mode = 2 (fixed value filling), this field stores the specified alternative value;

[0079] When #application_mode = 4 (weight adjustment), this field stores the weight coefficient;

[0080] When #application_mode=3 (mean fill), this field can store a pre-calculated mean result, or serve as a flag indicating that the calculation engine needs to call the corresponding mean calculation algorithm; it can also be empty, and the system will calculate it automatically.

[0081] Other modes mainly refer to: Mode 0: Do not process (default), Mode 1: Remove.

[0082] Specifically, Mode 0: No processing, which is the default state for all data. This indicates that no exception handling rules have been applied to this data point during this time period, and the original value should be used directly. To save storage space, the system may choose not to create records for these extended fields.

[0083] Field status:

[0084] #application_valid: Logically considered as 1 (valid), but physically this field can be empty (NULL) or does not exist at all. When the performance calculation engine finds that this field does not exist or is NULL during the query, it should interpret it as the default value of 1.

[0085] #application_mode: Logically, it is considered 0 (not processed), but physically this field can be empty (NULL) or does not exist at all. The engine needs to interpret it as the default value 0.

[0086] #application_value: must be NULL or does not exist at all. This mode does not require any parameters.

[0087] Mode 1: Elimination, explicitly indicating that the data point is invalid at this moment and should be excluded from the calculation. This mode requires explicit marking;

[0088] Field status:

[0089] #application_valid: must be explicitly set to 0, cannot be empty or not exist;

[0090] #application_mode: Must be explicitly set to 1 to specify the processing method;

[0091] #application_value: must be NULL or does not exist at all; the removal operation does not require an additional parameter value.

[0092] These extended fields share the same timestamp sequence as the original point values. Together, "point name" and "timestamp" form a composite primary key, ensuring that each original data point corresponds to a set of exception handling states. All fields (including original value fields and extended fields) together form a complete data record, which can be uniformly queried and processed by the subsequent performance calculation engine.

[0093] S2: Abnormal Time Period Configuration

[0094] The system provides a visual configuration interface, allowing users to define rules for handling abnormal data according to configuration parameters:

[0095] Configuration parameters include:

[0096] a. Select specific data points;

[0097] b. Set a time period;

[0098] c. Set the processing mode and parameters (such as fixed value, substitution logic, weight coefficient, etc.);

[0099] S3: Data Writing and Tagging

[0100] After configuration, the system will write extended field values ​​to the time-series database according to the set rules. Within the specified time period, the #application_* fields of the target locations will be updated to the set content to ensure accurate identification and processing in subsequent performance calculations.

[0101] The execution process of S3 includes:

[0102] S3.1: Instruction generation. The rule processor generates a timestamp sequence of all data to be written within the specified time period based on the start and end times set by the user (usually in the smallest time granularity supported by the database, such as 1 second or 1 minute).

[0103] Generate one or more data records for each timestamp, with the goal of updating the extended fields of the corresponding point name, and will generate batch write instructions;

[0104] S3.2: Data Writing. The system executes the above commands through the database's batch write API or efficient SQL statements. Batch writing is a standard operation for high-performance scenarios in time-series databases, which can greatly improve data update efficiency.

[0105] The write operation only modifies extended fields such as #application_*, and does not overwrite or change the original point_value field. This ensures the integrity of the original data and facilitates subsequent auditing and traceability; S3.3: Marking Results. After a successful write, at the database level, every data record of the target point within the specified time period has been marked with a "tag":

[0106] #application_valid is set to the value specified by the rule (usually 1 for valid and 0 for invalid);

[0107] #application_mode is set to the selected processing mode code (e.g., 2);

[0108] #application_value is set to the specified parameter (e.g., 100).

[0109] These tags, as metadata, are closely linked to the original data and await reading in the performance calculation process.

[0110] S3.4: Technical considerations: For rules that last for a long period of time (such as several days), batch writing must be used instead of single-point updates to ensure that writing a rule is an atomic operation, either all succeed or all fail, to avoid data inconsistency.

[0111] Conflict handling: If multiple rules cover the same location at the same time, the system needs to define clear priorities (such as the last configured rule taking precedence) and handle them properly during writing.

[0112] S4: Performance calculation logic integration. When performing assessment calculations, while retrieving raw data from the time-series database, the corresponding extended fields are queried simultaneously, and preprocessing rules are applied for real-time preprocessing. The performance calculation engine uses the obtained data set to execute the performance calculation formula and output the assessment results.

[0113] S4 specifically includes the following steps:

[0114] S4.1: Performance Calculation Engine Construction. The performance calculation engine is the core program for performing performance calculations, and its process is as follows:

[0115] S4.1.1: Data Acquisition: Query the time series database for all raw data point values ​​required for calculation;

[0116] S4.1.2: Data preprocessing: Synchronously or in advance query the extended field values ​​corresponding to these data points;

[0117] S4.1.3: Logical Judgment and Execution: Based on the value of the extended field, the rules defined in this patent are applied to preprocess the original data;

[0118] S4.1.4: Calculation: Perform the performance calculation formula using the preprocessed "clean" data;

[0119] S4.2: Calculation execution. The performance calculation engine needs to calculate the value of the performance score = F(A,B,C) within the time period [T_start,T_end].

[0120] S4.2 specifically includes:

[0121] S4.2.1: Constructing the query: The engine sends a query request to the time-series database.

[0122] S4.2.2: Integrated Query and Preprocessing: Preprocessing rules are embedded in the query logic. Before actual calculation, the engine constructs a "virtual value" for each data point, according to the following rules:

[0123] IFA#application_valid==0THEN

[0124] For this point in time, point A is considered to have no valid data and is ignored in aggregation calculations (such as summation and averaging). F(A,B,C) may become F(B,C).

[0125] ELSEIFA#application_mode==2THEN

[0126] Replace A.value with A#application_value, that is, the virtual value A = A#application_value.

[0127] ELSEIFA#application_mode==3THEN

[0128] Based on a preset mean calculation strategy (such as taking the average of the previous and next hour), a value is calculated in real time to replace A.value, that is, the virtual value A = mean_value.

[0129] ELSEIFA#application_mode==4THEN

[0130] The original value is weighted and adjusted, i.e., the virtual value A = A.value * A#application_value.

[0131] ELSE (Mode is 0 or not configured)

[0132] Dummy value A = A.value. / / Using primitive value

[0133] S4.2.3: The performance calculation engine uses the virtual values ​​A, B, and C generated by the above rules to substitute into the performance calculation formula F() to calculate the final score.

[0134] It should be noted that this application does not limit the specific content of the performance calculation formula. Each power plant can independently set its own performance calculation model for work teams or individuals (such as weighted average, points, over-limit penalties, etc.) based on its own management requirements, equipment operation characteristics, and assessment rules. The core innovation of this application lies in providing a configurable and traceable abnormal data processing mechanism for these performance calculations, ensuring that abnormal data involved in the calculation has been removed or corrected, thereby improving the fairness and accuracy of the assessment without changing the original scoring logic and weighting system of each plant.

[0135] Here is a specific application example:

[0136] Remove outliers during boiler trips

[0137] A power plant boiler tripped between 14:00 and 16:00 on September 12, 2024, causing the temperature sensor FURNACE_TEMP_1 to collect an abnormal value. The user configured the anomaly handling strategy for this time period in the system as follows:

[0138] a.FURNACE_TEMP_1#application_valid is set to 0 during this time period;

[0139] b. Set FURNACE_TEMP_1#application_mode to 1 (reject);

[0140] When the system calculates performance, the temperature values ​​during this time period will not be included in the total score.

[0141] Here is a specific application example:

[0142] Use fixed values ​​to fill false alarms in the desulfurization system

[0143] During the operation of the desulfurization system, the SO2_LEVEL_3 sensor experienced drastic fluctuations in data over a certain period due to false alarms. To avoid misjudgments, the user configured a handling strategy to fill the data with a fixed value:

[0144] a. Set SO2_LEVEL_3#application_mode = 2;

[0145] b. Set SO2_LEVEL_3#application_value = 80 (estimated based on the average value of the last 7 days);

[0146] The system replaced the original value with 80 during this period and continued to use it for performance scoring.

[0147] Here is a specific application example:

[0148] Weight adjustment to address poor sensor stability

[0149] Although the O2_SENSOR_5 data at a certain point is not completely invalid, its stability is poor, and the user has decided to reduce its influence weight in performance evaluation.

[0150] a. Set O2_SENSOR_5#application_mode = 4;

[0151] b. Set O2_SENSOR_5#application_value = 0.5;

[0152] In the performance formula, this point value is multiplied by 0.5 before being used in the calculation.

[0153] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A method for processing abnormal time period data for performance evaluation of thermal power units, characterized in that, Includes the following steps: S1: Data structure extension. Based on the point table structure in the database that stores the operating sequence data of thermal power units, a new set of extended fields with the suffix "#application" is added to store information related to the abnormal handling of the point within a specified time period. S2: Abnormal Time Period Configuration. Define abnormal data processing rules according to configuration parameters. Configuration parameters include selecting specific data points, setting time periods, and setting processing modes and parameters. S3: Data writing and tagging: Write extended field values ​​to the time series database according to configuration parameters and abnormal data handling rules; S4: Performance calculation logic integration. When performing assessment calculations, while retrieving raw data from the time-series database, it simultaneously queries the corresponding extended fields and applies preprocessing rules for real-time preprocessing. The performance calculation engine uses the obtained data set to execute the performance calculation formula and output the assessment results.

2. The method for processing abnormal time period data for performance evaluation of thermal power units according to claim 1, characterized in that, The extended fields include: A.<point_name> #application_valid: Indicates the validity of data for a specified time period. The default value is 1, which means that the point value data corresponding to the specified time period is valid and can be included in the performance evaluation calculation. If it is set to 0, it means that the point value data corresponding to the timestamp is invalid and will be excluded from the performance evaluation calculation. B.<point_name> #application_mode: Indicates the exception handling mode; C.<point_name> #application_value: Stores an alternative value or weight coefficient when the processing mode is fixed value fill or weight adjustment. For other modes, the extended field is empty.

3. The method for processing abnormal time period data for performance evaluation of thermal power units according to claim 2, characterized in that, The values ​​for the anomaly handling identifier are as follows: a.0: No processing is performed; the original point values ​​will be used directly for calculation. b.1: Remove, which will ignore the original point values ​​for the specified time period; c.2: Fixed value filling, the value stored in the {point_name}#application_value field will be used to replace the original point value for calculation; d.3: Mean fill, which calculates a replacement value according to preset rules and uses it for calculation. The calculation result is pre-calculated and stored in #application_value; e.4: Weight adjustment will read the weight coefficient stored in the {point_name}#application_value field, multiply the original point value by the weight, and then participate in the performance calculation formula.

4. The method for processing abnormal time period data for performance evaluation of thermal power units according to claim 1, characterized in that, The execution process of S3 includes: S3.1: Command generation. Based on the start and end times set by the user, generate a sequence of timestamps for all data to be written within the specified time period, generate one or more data records for each timestamp, and generate batch write commands. S3.2: Data writing, execute the above instructions through the database batch write statement; S3.3: Marking Results. After successful writing, at the database level, every data record for the target location within the specified time period has been marked with the extended field: S3.4: For technical considerations, batch writing is adopted for rules with long time periods; Conflict handling: If the same point in time is covered by multiple rules, a clear priority is defined and handled during writing.

5. The method for processing abnormal time period data for performance evaluation of thermal power units according to claim 1, characterized in that, S4 specifically includes the following steps: S4.1: Performance Calculation Engine Construction; S4.2: Calculation execution, performance calculation engine calculates the performance score = F(A,B,C) value within the time period [T_start,T_end].

6. The method for processing abnormal time period data for performance evaluation of thermal power units according to claim 1, characterized in that, The process in S4.1 includes: S4.1.1: Data acquisition, querying all raw data point values ​​required for calculation from the time series database; S4.1.2: Data preprocessing, querying the extended field values ​​corresponding to these data points; S4.1.3: Logical judgment and execution: Based on the value of the extended field, apply the defined rules to preprocess the original data; S4.1.4: Calculation, using the preprocessed data to execute the performance calculation formula.

7. The method for processing abnormal time period data for performance evaluation of thermal power units according to claim 1, characterized in that, S4.2 specifically includes: S4.2.1: Construct the query; the engine sends a query request to the time-series database. S4.2.2: Integrated Query and Preprocessing: The query logic embeds preprocessing rules, and the engine constructs a virtual value for each data point before actual calculation; S4.2.3: The performance calculation engine uses the virtual values ​​A, B, and C generated by the above rules to substitute into the performance calculation formula to obtain the final score.

8. The method for processing abnormal time period data for performance evaluation of thermal power units according to claim 7, characterized in that, The preprocessing rules include: IFA#application_valid==0THEN For this point in time, it is considered that point A does not have valid data, and this point is ignored in the aggregation calculation; ELSEIFA#application_mode==2THEN Replace A.value with A#application_value, that is, the virtual value A = A#application_value; ELSEIFA#application_mode==3THEN Based on the preset mean calculation strategy, a value is calculated in real time to replace A.value, that is, the virtual value A = mean_value; ELSEIFA#application_mode==4THEN Adjust the weights of the original values, i.e., virtual value A = A.value * A#application_value; ELSE (Mode is 0 or not configured) Virtual value A = A.value / / Using the original value.

Images