Power plant bidding analysis method and system based on deep learning

Abstract

This invention discloses a deep learning-based bidding analysis method and system for thermal power plants, belonging to the field of bidding analysis technology. The method includes: collecting a set of production parameters and a set of bidding parameters; analyzing and obtaining consistency parameters; classifying and obtaining a set of reliable production parameters and a set of risky production parameters; evaluating and screening reliable production parameters and bidding parameters with a correlation greater than a correlation threshold as a first set of reliable parameters, and obtaining a set of risky production parameters and a set of risky bidding parameters; adding risky production parameters and risky bidding parameters with an integrity greater than a integrity threshold to a second set of reliable parameters; constructing a reliable analysis model to perform reliable correction on risky production parameters and risky bidding parameters with an integrity less than or equal to a integrity threshold, as a third set of reliable parameters; and classifying and storing the data to complete the bidding data analysis and management of thermal power plants. This invention solves the technical problem of poor bidding analysis and management effects in existing technologies for thermal power plants.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, specifically to a method and system for competitive bidding analysis of thermal power plants based on deep learning. Background Technology

[0002] In the electricity market bidding process, thermal power plants rely on a large amount of operational and market data, such as fuel consumption, power generation, and electricity prices, to formulate bidding strategies. Existing data management methods typically only perform simple format cleaning of the collected parameters, lacking a systematic assessment of the internal consistency and logical correlation of the data. Due to the complexity of power plant operating conditions, sensor drift, or human recording bias, inconsistencies often arise within the production parameter set, such as mismatches between fluctuations in fuel consumption and power generation. Summary of the Invention

[0003] This application provides a deep learning-based bidding analysis method and system for thermal power plants, which addresses the technical problem of poor data analysis and management of bidding in thermal power plants in the prior art.

[0004] In view of the above problems, this application provides a deep learning-based bidding analysis method and system for thermal power plants.

[0005] In a first aspect, this application provides a deep learning-based bidding analysis method for thermal power plants, the method comprising: Collect the production parameter set and bidding parameter set of the thermal power plant, and perform consistency analysis on the production parameter set to obtain consistent parameters; The production parameter set is classified based on the consistency parameters to obtain a reliable production parameter set and a risky production parameter set; The correlation between the trusted production parameter set and the bidding parameter set is evaluated. Trusted production parameters and bidding parameters with a correlation greater than the correlation threshold are selected as the first trusted parameter group. Trusted production parameters with a correlation less than or equal to the correlation threshold are added to the risk production parameter set, and bidding parameters with a correlation less than or equal to the correlation threshold are added to the risk bidding parameter set. Perform an integrity assessment on the risk production parameter set and the risk bidding parameter set, and add the risk production parameters and risk bidding parameters whose integrity is greater than the integrity threshold to the second reliable parameter group; Based on the first set of trusted parameters, a trusted analysis model is constructed to perform trusted corrections on the risk production parameters and risk bidding parameters whose integrity is less than or equal to the integrity threshold, and to obtain the corrected risk production parameters and corrected risk bidding parameters as the third set of trusted parameters. The first, second, and third trusted parameter groups are classified and stored to complete the data analysis and management of bidding for thermal power plants.

[0006] Secondly, this application provides a deep learning-based bidding analysis system for thermal power plants, including: The consistency analysis module is used to collect the production parameter set and bidding parameter set of thermal power plants, and to perform consistency analysis on the production parameter set to obtain consistent parameters. The production parameter classification module is used to classify the production parameter set based on the consistency parameters to obtain a reliable production parameter set and a risky production parameter set. The correlation filtering module is used to evaluate the correlation between the trusted production parameter set and the bidding parameter set, filter trusted production parameters and bidding parameters with a correlation greater than the correlation threshold as the first trusted parameter group, and add trusted production parameters with a correlation less than or equal to the correlation threshold to the risk production parameter set, and add bidding parameters with a correlation less than or equal to the correlation threshold to the risk bidding parameter set. The integrity assessment module is used to assess the integrity of the risk production parameter set and the risk bidding parameter set, obtain the risk production parameters and the risk bidding parameters whose integrity is greater than the integrity threshold, and add them to the second reliable parameter group. The credible correction module is used to construct a credible analysis model based on the first credible parameter group, perform credible correction on the risk production parameters and the risk bidding parameters whose integrity is less than or equal to the integrity threshold, and obtain the corrected risk production parameters and the corrected risk bidding parameters as the third credible parameter group. The data management module is used to classify and store the first trusted parameter group, the second trusted parameter group and the third trusted parameter group to complete the data analysis and management of bidding for thermal power plants.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application proposes a deep learning-based bidding analysis method and system for thermal power plants. By evaluating and classifying production and bidding parameters layer by layer from three dimensions—consistency, correlation, and completeness—it significantly improves the overall quality and reliability of bidding data for thermal power plants. Compared to traditional methods, the technical solution provided in this application significantly enhances the ability to identify anomalous data. It can effectively filter out production parameters with inconsistent internal fluctuations, parameters with weak correlation to electricity price trends, and parameters with missing or incomplete data types. Furthermore, it uses a deep learning model to specifically correct low-completeness-risk data based on a high-reliability parameter set, thereby avoiding information distortion caused by direct discarding or simple interpolation. This application achieves the technical effect of transforming raw, heterogeneous, multi-source, and inconsistent quality data into a clearly layered, traceable, and correctable reliable parameter set, providing a stable data foundation for subsequent bidding decisions. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 A flowchart illustrating the deep learning-based bidding analysis method for thermal power plants provided in this application embodiment.

[0010] Figure 2 A schematic diagram of the structure of a deep learning-based bidding analysis system for thermal power plants provided in an embodiment of this application.

[0011] The components represented by each number in the attached diagram are explained below: Consistency analysis module 100, production parameter classification module 200, correlation screening module 300, integrity assessment module 400, reliability correction module 500, data management module 600. Detailed Implementation

[0012] This application provides a deep learning-based bidding analysis method and system for thermal power plants, which addresses the technical problem of poor data analysis and management of bidding in existing thermal power plants.

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0014] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.

[0015] Example 1, as Figure 1 As shown, this application provides a deep learning-based bidding analysis method for thermal power plants, wherein the method includes: S10: Collect the production parameter set and bidding parameter set of the thermal power plant, and perform consistency analysis on the production parameter set to obtain consistent parameters.

[0016] In the management of bidding data for thermal power plants, inconsistencies may exist within the production parameter set due to measurement errors, equipment drift, or sudden changes in operating conditions. Traditional methods often directly use the raw parameters, making it difficult to automatically identify the degree of matching between fuel consumption and power generation fluctuations, resulting in unreliable analysis results based on these parameters.

[0017] Step S10 in the method provided in this application embodiment includes: Collect production parameter sets and bidding parameter sets from thermal power plants, wherein the production parameter set includes fuel consumption and power generation, and the bidding parameter set includes electricity price data; Calculate the Pearson correlation coefficient between the fuel consumption and power generation of the production parameter set to obtain consistency parameters.

[0018] In this embodiment of the application, the production parameter set and bidding parameter set of the thermal power plant are collected, and the consistency analysis of the production parameter set is performed to obtain the consistency parameters.

[0019] Specifically, firstly, a set of production parameters and a set of bidding parameters for the thermal power plant are collected. The production parameter set includes fuel consumption and power generation, and the bidding parameter set includes electricity price data. For example, firstly, through the distributed control system and electricity market trading platform of the thermal power plant, hourly fuel consumption sequences (unit: tons) and power generation sequences (unit: megawatt-hours) are collected as the production parameter set during continuous operation, and corresponding electricity price data (unit: yuan per megawatt-hour) are collected as the bidding parameter set.

[0020] Further, the Pearson correlation coefficients of fuel consumption and power generation in the production parameter set are calculated to obtain consistency parameters. For example, for a fuel consumption sequence F=[f1,f2,...,f24] and a power generation sequence G=[g1,g2,...,g24] of 24 points on a certain day, the correlation coefficient ρ=cov(F,G) / (σ) is calculated. F ·σ G Where F is the fuel consumption series, G is the power generation series, cov(F,G) is the covariance of F and G, reflecting the degree of synchronization of the changing trends of the two series, and σ F Let σ be the standard deviation of the fuel consumption series. G ρ represents the standard deviation of the power generation series, which measures the dispersion of the series. If ρ = 0.85, it indicates a high positive correlation and good consistency; if ρ = 0.12, it indicates almost no correlation and abnormal fluctuations. The calculated consistency is used to characterize the degree of fluctuation matching within the production parameters. Production parameters with a high degree of fluctuation matching are more reliable and have higher utilization value.

[0021] By collecting production parameter sets and bidding parameter sets, and performing consistency analysis on the production parameter sets, consistency parameters reflecting the degree of coordination among various production parameters can be automatically output. This step effectively achieves preliminary quality screening of raw production data, transforming invisible data conflicts into quantifiable indicators, and providing an objective basis for subsequent classification and screening.

[0022] S20: Classify the production parameter set based on the consistency parameters to obtain a reliable production parameter set and a risky production parameter set.

[0023] Current technologies lack clear classification criteria to distinguish which production parameters can be directly used for bidding analysis and which require further review or processing. If all parameters are used indiscriminately, inconsistent parameters will contaminate the analysis results; if all are manually checked, it will be inefficient and inconsistent in standards.

[0024] Step S20 in the method provided in this application embodiment includes: Based on the statistical distribution of consistency parameters of the sample production parameter set, obtain the consistency threshold; Production parameters with consistency parameters greater than the consistency threshold are added to the trusted production parameter set, while production parameters with consistency parameters less than or equal to the consistency threshold are added to the risky production parameter set.

[0025] In this embodiment of the application, the production parameter set is classified based on the consistency parameter to obtain a reliable production parameter set and a risky production parameter set.

[0026] Specifically, firstly, a consistency threshold is obtained based on the statistical distribution of consistency parameters in the sample production parameter set. For example, the consistency parameters of daily production parameters under normal operating conditions at a thermal power plant over the past month can be collected, and their average value minus the standard deviation can be calculated as the consistency threshold. Consistency threshold = Mean of consistency parameters over the past month - Standard deviation of consistency parameters. Subtracting the standard deviation from the mean ensures that the consistency threshold encompasses data with fluctuations within a normal range, making the consistency threshold more valuable for reference.

[0027] Further, production parameters with consistency parameters greater than the consistency threshold are added to the set of trusted production parameters, while those with consistency parameters less than or equal to the consistency threshold are added to the set of risky production parameters. Specifically, the consistency parameters are compared with the consistency threshold. If the consistency parameter is greater than the threshold, it indicates a high degree of matching in the fluctuations within the group of production parameters, and it is added to the set of trusted production parameters. If the consistency parameter is less than or equal to the threshold, it indicates abnormal fluctuations or logical conflicts between fuel consumption and power generation, and it is added to the set of risky production parameters. For example, assuming the average value of the consistency parameter in a certain month is 0.7 and the standard deviation is 0.01, then the consistency threshold = 0.7 - 0.01 = 0.69. For example, one group of production parameters has a consistency parameter of 0.85, which is greater than 0.69, so it is included in the set of trusted production parameters; another group of production parameters has a consistency parameter of 0.58, which is less than 0.69, so it is included in the set of risky production parameters.

[0028] By classifying production parameter sets using consistency parameters, reliable and risky production parameter sets are automatically obtained. This step achieves efficient data routing, allowing parameters with high internal consistency to enter the reliable channel, while parameters with low consistency are marked as risk data for subsequent targeted processing, thereby improving the granularity and automation of data management.

[0029] S30: Evaluate the correlation between the trusted production parameter set and the bidding parameter set, select trusted production parameters and bidding parameters with a correlation greater than the correlation threshold as the first trusted parameter group, add trusted production parameters with a correlation less than or equal to the correlation threshold to the risk production parameter set, and add bidding parameters with a correlation less than or equal to the correlation threshold to the risk bidding parameter set.

[0030] Even if production parameters are consistent, their correlation with bidding parameters (such as electricity prices) may not be reasonable. In the actual operation of thermal power plants, fuel consumption and power generation should show a certain trend correlation with changes in electricity prices.

[0031] Step S30 in the method provided in this application embodiment includes: Alignment processing is performed based on the timestamps of the trusted production parameter set and the bidding parameter set to obtain corresponding parameter groups, wherein each corresponding parameter group includes a set of trusted production parameters and a set of bidding parameters. Calculate the correlation coefficient between the changing trends of the reliable production parameters and the bidding parameters within the corresponding parameter group, and use it as the correlation degree parameter. The correlation coefficient is obtained by calculating the trend covariance of the reliable production parameters and the bidding parameters and dividing it by the product of the trend standard deviation of the reliable production parameters and the trend standard deviation of the bidding parameters. Select the parameter groups with a correlation degree greater than the correlation threshold as the first reliable parameter groups; Add trusted production parameters with a correlation degree less than or equal to the correlation threshold to the risk production parameter set, and add bidding parameters with a correlation degree less than or equal to the correlation threshold to the risk bidding parameter set.

[0032] In this embodiment of the application, the set of trusted production parameters and the set of bidding parameters are evaluated for correlation. Trusted production parameters and bidding parameters with a correlation greater than the correlation threshold are selected as the first set of trusted parameters. Trusted production parameters with a correlation less than or equal to the correlation threshold are added to the set of risky production parameters, and bidding parameters with a correlation less than or equal to the correlation threshold are added to the set of risky bidding parameters.

[0033] Specifically, firstly, alignment processing is performed based on the timestamps of the trusted production parameter set and the bidding parameter set to obtain corresponding parameter groups. Each corresponding parameter group includes a set of trusted production parameters and a set of bidding parameters. For example, a corresponding parameter group for a certain thermal power plant on a certain day includes: hour 1: fuel consumption 48.2 tons, power generation 290 MWh, electricity price 352 yuan; hour 2: fuel consumption 47.5 tons, power generation 285 MWh, electricity price 350 yuan; hour 3: fuel consumption 46.8 tons, power generation 281 MWh, electricity price 348 yuan; hour 4: fuel consumption 47.0 tons, power generation 282 MWh, electricity price 349 yuan.

[0034] Further, the correlation coefficient between the changing trends of reliable production parameters and bidding parameters within the corresponding parameter group is calculated as a correlation degree parameter. This correlation coefficient is obtained by calculating the trend covariance of the reliable production parameters and the bidding parameters, and dividing it by the product of the trend standard deviation of the reliable production parameters and the trend standard deviation of the bidding parameters. For example, the change in adjacent time points for each parameter sequence is first calculated. For instance, for the fuel consumption sequence F=[f1,f2,...,fn], its trend sequence ΔF=[f2-f1,f3-f2,...,fn-f_{n-1}] is calculated; similarly, the trend sequence ΔP=[p2-p1, p3-p2,...,pn-p_{n-1}] is calculated for the electricity price sequence P. The trend covariance is the covariance of ΔF and ΔP, and the trend standard deviations are the standard deviations of ΔF and ΔP, respectively. For example, for each corresponding parameter group, the Pearson correlation coefficient between the changing trends of reliable production parameters and electricity prices is calculated. In the specific calculation, first calculate the Pearson correlation coefficient between the fuel consumption trend series and the electricity price trend series, then calculate the Pearson correlation coefficient between the power generation trend series and the electricity price trend series, and finally take the arithmetic mean of the two as the comprehensive correlation parameter of this set of parameters. Taking the correlation calculation of fuel consumption and electricity price as an example, first calculate the change series of fuel consumption in adjacent hours, for example, the change of the 2nd hour minus the 1st hour is -0.7 tons, the change of the 3rd hour minus the 2nd hour is -0.7 tons, the change of the 4th hour minus the 3rd hour is +0.2 tons, and the change series of electricity price, for example, -2 yuan, -2 yuan, +1 yuan. Further, calculate the covariance of these two change series: covariance = [(-0.7 - mean change) × (-2 - mean change of electricity price) + (-0.7 - mean change) × (-2 - mean change of electricity price) + (0.2 - mean change) × (1 - mean change of electricity price)] / (3-1). Next, calculate the standard deviations of the fuel consumption change series and the electricity price change series. Divide the covariance by the product of the two standard deviations to obtain the correlation coefficient between the fuel consumption trend and the electricity price trend. Similarly, calculate the correlation coefficient between the power generation trend and the electricity price trend. Finally, calculate the mean of the correlation coefficients between the fuel consumption trend and the electricity price trend, and the correlation coefficients between the power generation trend and the electricity price trend, to obtain the comprehensive correlation parameter.

[0035] Furthermore, parameter groups with a correlation degree greater than a correlation threshold are selected as the first reliable parameter group. For example, based on the average correlation degree of multiple corresponding parameter groups, the correlation threshold is obtained as 0.7. In practical applications, the correlation degree distribution of multiple corresponding parameter groups can also be evaluated by experts to obtain and adjust the threshold. Selecting parameter groups with a correlation degree greater than the correlation threshold as the first reliable parameter group ensures a stable and reasonable logical relationship between the production parameters and bidding parameters within that parameter group.

[0036] Furthermore, trusted production parameters with a correlation degree less than or equal to the correlation threshold are added to the risk production parameter set, and bidding parameters with a correlation degree less than or equal to the correlation threshold are added to the risk bidding parameter set. The reason for adding trusted production parameters and bidding parameters with a correlation degree less than or equal to the threshold to the risk production parameter set and risk bidding parameter set respectively is that although these parameters meet the consistency requirements individually, they lack an effective correlation with the electricity price fluctuation trend of the corresponding period, and direct use would introduce noise. Including them in the risk set avoids contamination of the first trusted parameter group and preserves the data foundation for subsequent integrity assessment and trusted correction steps to explore their potential usable value. Through alignment and correlation calculations, a quantitative assessment of the logical correlation between trusted production parameters and bidding parameters is achieved, and the extraction of the first trusted parameter group and the supplementation of the risk set are completed.

[0037] By evaluating the correlation, reliable production parameters and bidding parameters with a correlation higher than the threshold are selected as the first reliable parameter group. At the same time, production parameters and bidding parameters with low correlation are classified into risk sets, ensuring that there is a reasonable business correlation among the parameters within the first reliable parameter group.

[0038] S40: Perform an integrity assessment on the risk production parameter set and the risk bidding parameter set, obtain the risk production parameters and the risk bidding parameters whose integrity is greater than the integrity threshold, and add them to the second reliable parameter group.

[0039] In the risk production parameter set and risk bidding parameter set, some data may lack consistency or correlation, but their data completeness is relatively high. Directly discarding them would result in information waste.

[0040] Step S40 in the method provided in this application embodiment includes: The percentage of valid data in the risk production parameter set and the risk bidding parameter set are obtained respectively, and used as a data integrity indicator. The data type completeness of the risk production parameter set and the risk bidding parameter set is identified and denoted as the data type completeness index. Calculate the weighted sum of the data volume integrity index and the data type integrity index, and use it as the overall integrity parameter; Filter out risk production parameters and risk bidding parameters whose overall integrity parameters are greater than the integrity threshold, and add them to the second trustworthy parameter group.

[0041] In this embodiment of the application, the integrity of the risk production parameter set and the risk bidding parameter set is evaluated, and the risk production parameters and the risk bidding parameters with integrity greater than the integrity threshold are obtained and added to the second trusted parameter group.

[0042] Specifically, firstly, the percentage of valid data in the risk production parameter set and the risk bidding parameter set is obtained separately as the data integrity index. For example, taking the daily risk production parameters of a thermal power plant as an example, this set of parameters should include two data fields: fuel consumption and power generation. Each field should contain 24 records over 24 hours. In the actual collected data, the fuel consumption field is missing 3 records (e.g., no data in hours 5, 12, and 18), and the power generation field is missing 1 record (e.g., no data in hour 20). The total number of records that should be in all fields is 2 fields × 24 hours = 48 records, and the actual number of valid records is (24-3) + (24-1) = 44 records. Therefore, the data integrity index = actual valid records / total number of records that should be in the set = 44 / 48 = 0.917. For the risk bidding parameter set, taking electricity price data as an example, it should contain 24 records, but 1 record is missing. Therefore, the data integrity index = (24-1) / 24 ≈ 0.958. Furthermore, the data type completeness of the risk production parameter set and the risk bidding parameter set is identified and denoted as the data type completeness index. For example, in the risk production parameter set, there should be two field types: fuel consumption and power generation. However, only the field type fuel consumption actually appears. Therefore, the data type completeness index = 1 ÷ 2 = 0.50. In the risk bidding parameter set, there should be only one field type: electricity price. Since only one field type appears, the data type completeness index = 1 ÷ 1 = 1.00.

[0043] Furthermore, the weighted sum of the data volume integrity index and the data type integrity index is calculated as the overall integrity parameter. For example, the weight of the data volume integrity index is set to 0.60, and the weight of the data type integrity index is set to 0.40, to calculate the overall integrity parameter. The overall integrity parameter for the risk production parameter = 0.917 × 0.60 + 0.40 × 0.50 = 0.79. The overall integrity parameter for the risk bidding parameter = 0.60 × 0.958 + 0.40 × 1.00 = 0.975. For example, a pre-set integrity threshold of 0.85 is set based on historical overall integrity parameters. This threshold can be obtained based on the average of historical overall integrity parameters. In practical applications, the integrity threshold can also be obtained and adjusted based on expert evaluation of historical overall integrity parameters. Since the overall integrity parameter for the risk production parameter (0.79) is less than 0.85, it is not added; since the overall integrity parameter for the risk bidding parameter (0.975) is greater than 0.85, the electricity price data for that day is added to the second reliable parameter group. This step allows for the selection of parameters with high data integrity from the risk set, which are then used as a second set of reliable parameters, thus enabling further recycling and utilization of available data.

[0044] By assessing the completeness of the risk production parameter set and the risk bidding parameter set, parameters with completeness exceeding a threshold are added to the second set of reliable parameters. This step enables refined reuse of risk data, expands the sources of reliable data, and improves overall data utilization.

[0045] S50: Based on the first set of trusted parameters, construct a trusted analysis model, perform trusted correction on the risk production parameters and risk bidding parameters whose integrity is less than or equal to the integrity threshold, and obtain the corrected risk production parameters and corrected risk bidding parameters as the third set of trusted parameters.

[0046] For risk parameters with low integrity, although they cannot be directly used for bidding analysis, simply discarding them may lead to a waste of data.

[0047] Step S50 in the method provided in this application embodiment includes: Based on the first set of reliable parameters, a missing data prediction model based on deep learning is constructed. The missing data prediction model includes three prediction branches, each of which is used to predict a class of missing reference parameters. Among them, based on the first set of reliable parameters, a deep learning-based missing data prediction model is constructed, including: The first set of reliable parameters is sampled with replacement to obtain three sets of missing sample datasets. The first set of missing sample datasets uses reliable fuel consumption and reliable electricity generation as input and reliable electricity price data as output. The second set of missing sample datasets uses reliable fuel consumption and reliable electricity price data as input and reliable electricity generation data as output. The third set of missing sample datasets uses reliable electricity price data and reliable electricity generation data as input and reliable fuel consumption data as output. Based on deep learning, a missing data prediction model is constructed. The missing data prediction model includes three prediction branches: the first prediction branch is used to predict electricity price data, the second prediction branch is used to predict power generation, and the third prediction branch is used to predict fuel consumption. The three prediction branches of the missing data prediction model are trained in a supervised manner using three sets of missing sample datasets until convergence, thus obtaining the trained missing data prediction model. Input the risk production parameters and risk bidding parameters with completeness less than or equal to the completeness threshold into the missing data prediction model to obtain the missing reference parameters predicted by the model; The missing reference parameters are fused with the existing valid parameter values ​​in the risk production parameters and the risk bidding parameters to obtain the corrected risk production parameters and the corrected risk bidding parameters, which serve as the third reliable parameter set.

[0048] In this embodiment of the application, a reliable analysis model is constructed based on the first reliable parameter group. The risk production parameters and risk bidding parameters whose integrity is less than or equal to the integrity threshold are reliably corrected to obtain the corrected risk production parameters and corrected risk bidding parameters as the third reliable parameter group.

[0049] Specifically, based on the first set of reliable parameters, a deep learning-based missing data prediction model is constructed, wherein the missing data prediction model includes three prediction branches, each of which is used to predict a class of missing reference parameters.

[0050] Among them, based on the first set of reliable parameters, a deep learning-based missing data prediction model is constructed, including: First, sampling with replacement is performed on the first reliable parameter group to obtain three sets of missing sample datasets. The first missing sample dataset uses reliable fuel consumption and reliable electricity generation as inputs and reliable electricity price data as outputs. The second missing sample dataset uses reliable fuel consumption and reliable electricity price as inputs and reliable electricity generation as outputs. The third missing sample dataset uses reliable electricity price data and reliable electricity generation as inputs and reliable fuel consumption as outputs. For example, the first reliable parameter group contains multiple sets of complete data, such as multiple fuel consumption sequences, electricity generation sequences, and electricity price sequences. Samples are first extracted from the first reliable parameter group using sampling with replacement to construct the three sets of missing sample datasets. The first missing sample dataset uses fuel consumption and electricity generation as inputs and electricity price as output. For example, 100 samples are extracted from the above three groups, with each input being (fuel consumption, electricity generation) and the output being the electricity price. The second missing sample dataset uses fuel consumption and electricity price as inputs and electricity generation as output, also with 100 samples extracted. The third missing sample dataset: 100 samples were drawn with electricity price and power generation as input and fuel consumption as output.

[0051] Furthermore, based on deep learning, a missing data prediction model is constructed. This model includes three prediction branches: a first branch predicts electricity price data, a second branch predicts power generation, and a third branch predicts fuel consumption. For example, three independent fully connected neural networks are used as the three prediction branches. To address the issue of missing values ​​in the input sequence, the missing positions in each input sequence are temporarily padded. For example, each missing value is padded with the mean of the sequence, and a missing mask vector of the same length as the input sequence is generated, with a mask value of 1 for missing positions and 0 for non-missing positions. The padded sequence and the mask vector are concatenated along the feature dimension as the model input. To predict missing values ​​at any time point, the model uses a bidirectional long short-term memory network. The number of nodes in its input layer is the time window length multiplied by the number of input features multiplied by 2 to include the missing mask, i.e., 24 × 2 × 2 = 96 nodes. Specifically, the first prediction branch takes as input the padded 24-hour fuel consumption sequence, the padded 24-hour power generation sequence, and two corresponding mask vectors. After concatenation, the input dimension is 24 × (1 + 1 + 1 + 1) = 96 (each hour corresponds to a fuel consumption value, fuel consumption mask, power generation value, and power generation mask). The model includes a bidirectional LSTM layer with 64 hidden units. The forward and backward outputs are concatenated to obtain a 128-dimensional vector, followed by a fully connected output layer with an output dimension of 24. The activation function is a linear function. Each time step in the output sequence corresponds to the predicted electricity price for that hour. During training, the loss function only calculates the mean squared error of missing bits (i.e., mask values ​​of 1) in the original input; non-missing bits are not included in the loss calculation. The second prediction branch predicts power generation, taking as input the padded fuel consumption sequence, electricity price sequence, and corresponding masks, with the same structure. The third prediction branch predicts fuel consumption, taking as input the padded electricity price sequence, power generation sequence, and corresponding masks, with the same structure. Before inputting the data into the model, each feature sequence is subjected to min-max normalization. The model is trained using the Adam optimizer with a learning rate of 0.001, a batch size of 16, and 500 training epochs. The second prediction branch predicts power generation and has the same structure. The third prediction branch predicts fuel consumption and has the same structure.

[0052] Furthermore, the three prediction branches of the missing data prediction model are trained in a supervised manner using three sets of missing sample datasets until convergence, thus obtaining the trained missing data prediction model. For example, for the first prediction branch, complete 24-hour fuel consumption sequences, 24-hour power generation sequences, and corresponding 24-hour electricity price sequences are extracted from the first set of reliable parameters as samples. Several time points in the electricity price sequence are randomly masked, and the values ​​at the masked positions are filled with the mean of the sequence, generating corresponding mask vectors to represent the missing value positions. Masked positions are represented by 1, and non-masked positions by 0. The filled fuel consumption sequence, power generation sequence, and their respective mask vectors are concatenated as input features, and the original complete electricity price sequence is the output target. During training, the loss function only calculates the mean square error at the masked positions; non-masked positions are not included in the loss calculation. The Adam optimizer is used with a learning rate of 0.001, a batch size of 16, and 500 training epochs. The second prediction branch takes the padded fuel consumption sequence, electricity price sequence, and corresponding mask as input and outputs the complete power generation sequence; the third prediction branch takes the padded electricity price sequence, power generation sequence, and corresponding mask as input and outputs the complete fuel consumption sequence. Training samples are constructed and supervised training is performed until the loss function of each branch converges.

[0053] Further, risk production parameters and risk bidding parameters with completeness less than or equal to the completeness threshold are input into the missing data prediction model to obtain the missing reference parameters predicted by the model. For example, the risk production parameters and risk bidding parameters for a certain day are not added to the second reliable parameter group because their completeness is less than or equal to the completeness threshold. The risk production parameters and risk bidding parameters for a certain day are not added to the second reliable parameter group because their completeness is below the threshold. Assume that the fuel consumption sequence for that day is complete, the power generation sequence is missing at hours 5, 12, and 18, and the electricity price sequence is complete. First, the missing positions in the power generation sequence are temporarily filled with the mean of the sequence, and corresponding mask vectors are generated. The filled power generation sequence, the complete fuel consumption sequence, and the two mask vectors are concatenated to form an input feature tensor with a dimension of 24×4 (each hour corresponds to the filled fuel consumption, fuel consumption mask, filled power generation, and power generation mask). This tensor is input into the second prediction branch, and the bidirectional LSTM network outputs a complete 24-hour power generation prediction sequence. The predicted values ​​corresponding to the mask positions are extracted from the output sequence as missing reference parameters. If only one valid data item remains for a certain set of risk production parameters and risk bidding parameters, such as only the fuel consumption sequence while power generation and electricity price are missing, the missing data prediction model cannot provide enough input information to make effective predictions for any branch. Therefore, the data set is discarded and not included in the third credible parameter set.

[0054] Furthermore, the missing reference parameter is fused with the existing valid parameter values ​​from the risk production parameters and the risk bidding parameters to obtain corrected risk production parameters and corrected risk bidding parameters, which serve as the third reliable parameter set. Specifically, this predicted value is used as the missing reference parameter and fused with the existing fuel consumption of 48.5 tons and electricity price of 352 yuan per MWh to obtain complete corrected risk production parameters and corrected risk bidding parameters, namely fuel consumption of 48.5 tons, power generation of 291 MWh, and electricity price of 352 yuan per MWh, which are then used as the third reliable parameter set. Through this step, intelligent completion of low-completeness risk data is achieved, improving data availability.

[0055] A reliable analysis model is constructed based on the first set of reliable parameters. Risk parameters with low integrity are then reliably corrected to obtain the corrected parameters, which serve as the third set of reliable parameters. This step utilizes a deep learning model to learn the mapping patterns within the high-reliability parameter set, thereby providing logically sound supplementary values ​​for missing or anomalous risk data. This effectively reduces the adverse impact of missing data on bidding management.

[0056] S60: Classify and store the first trusted parameter group, the second trusted parameter group and the third trusted parameter group to complete the bidding data analysis and management of thermal power plants.

[0057] After the aforementioned multi-step processing, the three sets of trusted parameters have different levels of trust. If they are stored together, their priorities and applicable scenarios cannot be distinguished during subsequent use.

[0058] Step S60 in the method provided in this application embodiment includes: The corresponding data collection time identifiers, integrity assessment results, and correction records associated with each trusted parameter group are integrated to establish a unified data index; The data is stored in the first trusted parameter group as the highest priority, the second trusted parameter group as the middle priority, and the third trusted parameter group as the basic priority. This classification and storage method completes the data analysis and management of bidding for thermal power plants.

[0059] In this embodiment of the application, the first trusted parameter group, the second trusted parameter group and the third trusted parameter group are classified and stored to complete the bidding data analysis and management of thermal power plants.

[0060] Specifically, firstly, the data collection time identifier, integrity assessment results, and correction records associated with each trusted parameter group are integrated to establish a unified data index. For example, the following metadata is associated with each data group: data collection time identifier (e.g., March 6, 2026, 12:00), integrity assessment results (e.g., integrity parameter of the first trusted parameter group is 1.00, the second trusted parameter group is 0.85, and the third trusted parameter group is 0.56), and correction records (e.g., the correction record for the third trusted parameter group is "Power generation is supplemented by the second prediction branch"). After integrating this information, a unified data index is established using a relational database, with each record assigned a unique index number.

[0061] Furthermore, the first trusted parameter group is stored as the highest priority data, the second trusted parameter group as the middle priority data, and the third trusted parameter group as the basic priority data, thus completing the bidding data analysis and management for thermal power plants. For example, the first trusted parameter group, as the highest priority data, is allocated a 50-gigabyte high-speed solid-state drive array with a read / write latency of less than 1 millisecond; the second trusted parameter group, as the middle priority data, is allocated a 100-gigabyte ordinary solid-state drive with a read / write latency of approximately 5 milliseconds; and the third trusted parameter group, as the basic priority data, is allocated a 200-gigabyte mechanical hard drive with a read / write latency of approximately 20 milliseconds. After the classified storage is completed, the bidding data analysis and management process for thermal power plants ends. This step achieves hierarchical and classified storage of the three trusted parameter groups, ensuring rapid access to high-trust data and low-cost long-term preservation of low-trust data.

[0062] The first, second, and third trusted parameter groups are classified and stored, and associated with data such as collection time stamps, integrity assessment results, and correction records. This enables the orderly management of bidding data, allowing parameters of different trust levels to be used as needed in subsequent queries and analyses. At the same time, complete processing traceability information is retained, significantly improving the standardization and traceability of data management.

[0063] Example 2, as Figure 2 As shown, based on the same inventive concept as the deep learning-based bidding analysis method for thermal power plants provided in Embodiment 1, this embodiment of the invention also provides a deep learning-based bidding analysis system for thermal power plants, including: The consistency analysis module 100 is used to collect the production parameter set and bidding parameter set of the thermal power plant, and to perform consistency analysis on the production parameter set to obtain consistent parameters. The production parameter classification module 200 is used to classify the production parameter set based on the consistency parameters to obtain a reliable production parameter set and a risky production parameter set. The correlation filtering module 300 is used to evaluate the correlation between the trusted production parameter set and the bidding parameter set, filter trusted production parameters and bidding parameters with a correlation greater than the correlation threshold as the first trusted parameter group, and add trusted production parameters with a correlation less than or equal to the correlation threshold to the risk production parameter set, and add bidding parameters with a correlation less than or equal to the correlation threshold to the risk bidding parameter set. The integrity assessment module 400 is used to assess the integrity of the risk production parameter set and the risk bidding parameter set, obtain the risk production parameters and the risk bidding parameters whose integrity is greater than the integrity threshold, and add them to the second trusted parameter group. The credible correction module 500 is used to construct a credible analysis model based on the first credible parameter group, perform credible correction on the risk production parameters and the risk bidding parameters whose integrity is less than or equal to the integrity threshold, and obtain the corrected risk production parameters and the corrected risk bidding parameters as the third credible parameter group. The data management module 600 is used to classify and store the first trusted parameter group, the second trusted parameter group and the third trusted parameter group to complete the bidding data analysis and management of thermal power plants.

[0064] In one embodiment, the consistency analysis module 100 is further configured to: Collect production parameter sets and bidding parameter sets from thermal power plants, wherein the production parameter set includes fuel consumption and power generation, and the bidding parameter set includes electricity price data; Calculate the Pearson correlation coefficient between the fuel consumption and power generation of the production parameter set to obtain consistency parameters.

[0065] In one embodiment, the production parameter classification module 200 is further configured to: Based on the statistical distribution of consistency parameters of the sample production parameter set, obtain the consistency threshold; Production parameters with consistency parameters greater than the consistency threshold are added to the trusted production parameter set, while production parameters with consistency parameters less than or equal to the consistency threshold are added to the risky production parameter set.

[0066] In one embodiment, the correlation filtering module 300 is further configured to: Alignment processing is performed based on the timestamps of the trusted production parameter set and the bidding parameter set to obtain corresponding parameter groups, wherein each corresponding parameter group includes a set of trusted production parameters and a set of bidding parameters. Calculate the correlation coefficient between the changing trends of the reliable production parameters and the bidding parameters within the corresponding parameter group, and use it as the correlation degree parameter. The correlation coefficient is obtained by calculating the trend covariance of the reliable production parameters and the bidding parameters and dividing it by the product of the trend standard deviation of the reliable production parameters and the trend standard deviation of the bidding parameters. Select the parameter groups with a correlation degree greater than the correlation threshold as the first reliable parameter groups; Add trusted production parameters with a correlation degree less than or equal to the correlation threshold to the risk production parameter set, and add bidding parameters with a correlation degree less than or equal to the correlation threshold to the risk bidding parameter set.

[0067] In one embodiment, the integrity assessment module 400 is further configured to: The percentage of valid data in the risk production parameter set and the risk bidding parameter set are obtained respectively, and used as a data integrity indicator. The data type completeness of the risk production parameter set and the risk bidding parameter set is identified and denoted as the data type completeness index. Calculate the weighted sum of the data volume integrity index and the data type integrity index, and use it as the overall integrity parameter; Filter out risk production parameters and risk bidding parameters whose overall integrity parameters are greater than the integrity threshold, and add them to the second trustworthy parameter group.

[0068] In one embodiment, the trusted correction module 500 is further configured to: Based on the first set of reliable parameters, a missing data prediction model based on deep learning is constructed. The missing data prediction model includes three prediction branches, each of which is used to predict a class of missing reference parameters. Among them, based on the first set of reliable parameters, a deep learning-based missing data prediction model is constructed, including: The first set of reliable parameters is sampled with replacement to obtain three sets of missing sample datasets. The first set of missing sample datasets uses reliable fuel consumption and reliable electricity generation as input and reliable electricity price data as output. The second set of missing sample datasets uses reliable fuel consumption and reliable electricity price data as input and reliable electricity generation data as output. The third set of missing sample datasets uses reliable electricity price data and reliable electricity generation data as input and reliable fuel consumption data as output. Based on deep learning, a missing data prediction model is constructed. The missing data prediction model includes three prediction branches: the first prediction branch is used to predict electricity price data, the second prediction branch is used to predict power generation, and the third prediction branch is used to predict fuel consumption. The three prediction branches of the missing data prediction model are trained in a supervised manner using three sets of missing sample datasets until convergence, thus obtaining the trained missing data prediction model. Input the risk production parameters and risk bidding parameters with completeness less than or equal to the completeness threshold into the missing data prediction model to obtain the missing reference parameters predicted by the model; The missing reference parameters are fused with the existing valid parameter values ​​in the risk production parameters and the risk bidding parameters to obtain the corrected risk production parameters and the corrected risk bidding parameters, which serve as the third reliable parameter set.

[0069] In one embodiment, the data management module 600 is further configured to: The corresponding data collection time identifiers, integrity assessment results, and correction records associated with each trusted parameter group are integrated to establish a unified data index; The data is stored in the first trusted parameter group as the highest priority, the second trusted parameter group as the middle priority, and the third trusted parameter group as the basic priority. This classification and storage method completes the data analysis and management of bidding for thermal power plants.

[0070] In summary, the embodiments of this application have at least the following technical effects: This application proposes a deep learning-based bidding analysis method and system for thermal power plants. By evaluating and classifying production and bidding parameters layer by layer from three dimensions—consistency, correlation, and completeness—it significantly improves the overall quality and reliability of bidding data for thermal power plants. Compared to traditional methods, the technical solution provided in this application significantly enhances the ability to identify anomalous data. It can effectively filter out production parameters with inconsistent internal fluctuations, parameters with weak correlation to electricity price trends, and parameters with missing or incomplete data types. Furthermore, it uses a deep learning model to specifically correct low-completeness-risk data based on a high-reliability parameter set, thereby avoiding information distortion caused by direct discarding or simple interpolation. This application achieves the technical effect of transforming raw, heterogeneous, multi-source, and inconsistent quality data into a clearly layered, traceable, and correctable reliable parameter set, providing a stable data foundation for subsequent bidding decisions.

[0071] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0072] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0073] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A deep learning-based bidding analysis method for thermal power plants, characterized in that, include: Collect the production parameter set and bidding parameter set of the thermal power plant, and perform consistency analysis on the production parameter set to obtain consistent parameters; The production parameter set is classified based on the consistency parameters to obtain a reliable production parameter set and a risky production parameter set; The correlation between the trusted production parameter set and the bidding parameter set is evaluated. Trusted production parameters and bidding parameters with a correlation greater than the correlation threshold are selected as the first trusted parameter group. Trusted production parameters with a correlation less than or equal to the correlation threshold are added to the risk production parameter set, and bidding parameters with a correlation less than or equal to the correlation threshold are added to the risk bidding parameter set. Perform an integrity assessment on the risk production parameter set and the risk bidding parameter set, and add the risk production parameters and risk bidding parameters whose integrity is greater than the integrity threshold to the second reliable parameter group; Based on the first set of trusted parameters, a trusted analysis model is constructed to perform trusted corrections on the risk production parameters and risk bidding parameters whose integrity is less than or equal to the integrity threshold, and to obtain the corrected risk production parameters and corrected risk bidding parameters as the third set of trusted parameters. The first, second, and third trusted parameter groups are classified and stored to complete the data analysis and management of bidding for thermal power plants.

2. The deep learning-based bidding analysis method for thermal power plants according to claim 1, characterized in that, Collect the production parameter set and bidding parameter set of the thermal power plant, and perform consistency analysis on the production parameter set to obtain consistent parameters; Collect production parameter sets and bidding parameter sets from thermal power plants, wherein the production parameter set includes fuel consumption and power generation, and the bidding parameter set includes electricity price data; Calculate the Pearson correlation coefficient between the fuel consumption and power generation of the production parameter set to obtain consistency parameters.

3. The deep learning-based bidding analysis method for thermal power plants according to claim 1, characterized in that, Based on the consistency parameters, the production parameter set is classified to obtain a reliable production parameter set and a risky production parameter set, including: Based on the statistical distribution of consistency parameters of the sample production parameter set, obtain the consistency threshold; Production parameters with consistency parameters greater than the consistency threshold are added to the trusted production parameter set, while production parameters with consistency parameters less than or equal to the consistency threshold are added to the risky production parameter set.

4. The deep learning-based bidding analysis method for thermal power plants according to claim 1, characterized in that, The trusted production parameter set and the bidding parameter set are evaluated for correlation. Trusted production parameters and bidding parameters with a correlation greater than a correlation threshold are selected as the first trusted parameter group. Trusted production parameters with a correlation less than or equal to the correlation threshold are added to the risk production parameter set, and bidding parameters with a correlation less than or equal to the correlation threshold are added to the risk bidding parameter set. This includes: Alignment processing is performed based on the timestamps of the trusted production parameter set and the bidding parameter set to obtain corresponding parameter groups, wherein each corresponding parameter group includes a set of trusted production parameters and a set of bidding parameters. Calculate the correlation coefficient between the changing trends of the reliable production parameters and the bidding parameters within the corresponding parameter group, and use it as the correlation degree parameter. The correlation coefficient is obtained by calculating the trend covariance of the reliable production parameters and the bidding parameters and dividing it by the product of the trend standard deviation of the reliable production parameters and the trend standard deviation of the bidding parameters. Select the parameter groups with a correlation degree greater than the correlation threshold as the first reliable parameter groups; Add trusted production parameters with a correlation degree less than or equal to the correlation threshold to the risk production parameter set, and add bidding parameters with a correlation degree less than or equal to the correlation threshold to the risk bidding parameter set.

5. The deep learning-based bidding analysis method for thermal power plants according to claim 1, characterized in that, The integrity of the risk production parameter set and the risk bidding parameter set is assessed. Risk production parameters and risk bidding parameters with integrity greater than a integrity threshold are obtained and added to the second reliable parameter group, including: The percentage of valid data in the risk production parameter set and the risk bidding parameter set are obtained respectively, and used as a data integrity indicator. The data type completeness of the risk production parameter set and the risk bidding parameter set is identified and denoted as the data type completeness index. Calculate the weighted sum of the data volume integrity index and the data type integrity index, and use it as the overall integrity parameter; Filter out risk production parameters and risk bidding parameters whose overall integrity parameters are greater than the integrity threshold, and add them to the second trustworthy parameter group.

6. The deep learning-based bidding analysis method for thermal power plants according to claim 1, characterized in that, Based on the first set of trusted parameters, a trusted analysis model is constructed. Trustworthy corrections are made to the risk production parameters and risk bidding parameters whose integrity is less than or equal to the integrity threshold. Corrected risk production parameters and corrected risk bidding parameters are obtained as the third set of trusted parameters, including: Based on the first set of reliable parameters, a missing data prediction model based on deep learning is constructed. The missing data prediction model includes three prediction branches, each of which is used to predict a class of missing reference parameters. Input the risk production parameters and risk bidding parameters with completeness less than or equal to the completeness threshold into the missing data prediction model to obtain the missing reference parameters predicted by the model; The missing reference parameters are fused with the existing valid parameter values ​​in the risk production parameters and the risk bidding parameters to obtain the corrected risk production parameters and the corrected risk bidding parameters, which serve as the third reliable parameter set.

7. The deep learning-based bidding analysis method for thermal power plants according to claim 6, characterized in that, Based on the first set of reliable parameters, a deep learning-based missing data prediction model is constructed, including: Sampling with replacement is performed on the first set of trusted parameters to obtain three sets of missing sample datasets. Based on deep learning, a missing data prediction model is constructed. The missing data prediction model includes three prediction branches: the first prediction branch is used to predict electricity price data, the second prediction branch is used to predict power generation, and the third prediction branch is used to predict fuel consumption. The three prediction branches of the missing data prediction model are trained in a supervised manner using three sets of missing sample datasets until convergence, thus obtaining the trained missing data prediction model.

8. The deep learning-based bidding analysis method for thermal power plants according to claim 7, characterized in that, The first set of reliable parameters is sampled with replacement to obtain three sets of missing sample datasets. The first set of missing sample datasets uses reliable fuel consumption and reliable electricity generation as input and reliable electricity price data as output. The second set of missing sample datasets uses reliable fuel consumption and reliable electricity price data as input and reliable electricity generation data as output. The third set of missing sample datasets uses reliable electricity price data and reliable electricity generation data as input and reliable fuel consumption data as output.

9. The deep learning-based bidding analysis method for thermal power plants according to claim 1, characterized in that, The first, second, and third trusted parameter groups are categorized and stored to complete the data analysis and management of bidding for thermal power plants, including: The data index is established by integrating the corresponding collection time identifier, integrity assessment results and correction records of each trusted parameter group; The data is stored in the first trusted parameter group as the highest priority, the second trusted parameter group as the middle priority, and the third trusted parameter group as the basic priority. This classification and storage method completes the data analysis and management of bidding for thermal power plants.

10. A deep learning-based bidding analysis system for thermal power plants, characterized in that, The system is used to implement the deep learning-based bidding analysis method for thermal power plants according to any one of claims 1-9, the system comprising: The consistency analysis module is used to collect the production parameter set and bidding parameter set of thermal power plants, and to perform consistency analysis on the production parameter set to obtain consistent parameters. The production parameter classification module is used to classify the production parameter set based on the consistency parameters to obtain a reliable production parameter set and a risky production parameter set. The correlation filtering module is used to evaluate the correlation between the trusted production parameter set and the bidding parameter set, filter trusted production parameters and bidding parameters with a correlation greater than the correlation threshold as the first trusted parameter group, and add trusted production parameters with a correlation less than or equal to the correlation threshold to the risk production parameter set, and add bidding parameters with a correlation less than or equal to the correlation threshold to the risk bidding parameter set. The integrity assessment module is used to assess the integrity of the risk production parameter set and the risk bidding parameter set, obtain the risk production parameters and the risk bidding parameters whose integrity is greater than the integrity threshold, and add them to the second reliable parameter group. The credible correction module is used to construct a credible analysis model based on the first credible parameter group, perform credible correction on the risk production parameters and the risk bidding parameters whose integrity is less than or equal to the integrity threshold, and obtain the corrected risk production parameters and the corrected risk bidding parameters as the third credible parameter group. The data management module is used to classify and store the first trusted parameter group, the second trusted parameter group and the third trusted parameter group to complete the data analysis and management of bidding for thermal power plants.

Images