Dynamic multi-dimensional quality evaluation system and method for thermal power plant intelligent disk monitoring model

Abstract

The invention relates to a dynamic multi-dimensional quality evaluation system and method for a thermal power plant intelligent inventory supervision model. Real-time data flow of the DCS is collected and compared with the model prediction value, and a prediction error is generated; based on the load rate, the fuel characteristic and other characteristics, the current operation state is matched to a preset working condition library, and historical data is called for index calculation; the method comprises the following steps of: according to stability, reusability and economic indexes such as a variable coefficient of a prediction error, a model calling frequency, economic benefit influence and the like, normalizing the indexes by adopting a range method, and calculating the weight of each index by utilizing an entropy weight method; constructing a weighted standardization matrix, defining a dynamic ideal solution, calculating a close degree index, and outputting a comprehensive evaluation result; according to the scoring result, an optimization instruction is automatically pushed, and model parameter adjustment or data calibration is carried out, a set of dynamic and multi-dimensional thermal power plant supervision model evaluation system is constructed, the model quality can be objectively and comprehensively evaluated, and a closed-loop optimization mechanism is provided.

Description

Technical Field

[0001] The present invention relates to the technical field of quality evaluation of power plant monitoring models, and particularly to a dynamic multi-dimensional quality evaluation system and method for a smart monitoring model of a thermal power plant. Background Art

[0002] The technical background of power plant monitoring model evaluation stems from the sharp increase in the complexity of unit operating conditions under the background of high penetration of new energy, and the large-scale application of intelligent models in thermal power plants brought about by the development of artificial intelligence technology. The necessity of model evaluation technology lies in the lack of an objective and comprehensive evaluation system in the existing monitoring system, which mostly relies on single indicators and static weight allocation, and cannot correctly measure the quality and health of the model, resulting in low model quality and long update cycles, and it is difficult to adapt to dynamic conditions such as deep peak shaving.

[0003] The current evaluation methods are mainly based on expert experience and single-dimensional error indicators, and have three defects: First, the dimension is single, only considering the accuracy of model prediction, ignoring the economic value, reusability, stability, etc. brought by the model, and unable to evaluate the actual use value of the model; Second, the efficiency of manual judgment is low, relying on expert experience to judge the model regularly, with a long cycle, heavy workload and strong subjectivity; Third, there is a lack of evaluation mechanism and closed-loop feedback mechanism for the model under dynamic conditions. Under different operating conditions of thermal power units, the model evaluation criteria should be dynamically adjusted and matched. For example, under low-load conditions, the reliability requirements of the model are increased and the economic indicators are appropriately reduced; under high-load conditions, the focus is on the economy of operation. The existing technical system cannot meet the requirements of dynamic adjustment of evaluation indicators and cannot guide the real-time adjustment of the model. These defects make it difficult for the existing evaluation system to support intelligent services such as unit economic operation, fault warning, and closed-loop optimal control, and restrict the realization of the goals of less personnel on duty and energy efficiency optimization in thermal power plants.

[0004] Therefore, the present invention proposes a dynamic multi-dimensional quality evaluation system and method for a smart monitoring model of a thermal power plant. Summary of the Invention

[0005] The present invention aims at the technical problems existing in the prior art and provides a dynamic multi-dimensional quality evaluation system and method for a smart monitoring model of a thermal power plant.

[0006] The technical solution of the present invention to solve the above technical problems is as follows: A dynamic multi-dimensional quality evaluation system and method for a smart monitoring model of a thermal power plant; including the following steps:

[0007] Data input and model processing: Collect the real-time data stream of the DCS system, synchronously access the output result of the monitoring model, the model receives the production flow data and outputs a predicted value, compares it with the actual measured value, and generates a model prediction error value;

[0008] Operating condition matching and historical data retrieval: Based on the characteristics of production flow data, classify the current operating state into a preset operating condition library. After successful matching, retrieve the historical operating data, model statistical indicators, and model economic data under this operating condition from the historical database to participate in indicator calculation;

[0009] Multi-dimensional indicator calculation: Calculate stability indicators, reusability indicators, and economic indicators;

[0010] Data standardization and weight assignment: Use the range method to perform dimensionless processing on multiple indicators, and calculate the information entropy of each indicator through the entropy weight method to determine the weights;

[0011] TOPSIS comprehensive evaluation: Construct a weighted standardized matrix. According to the current operating condition type, calculate the positive / negative ideal solutions through historical data statistics, and use the Euclidean distance to calculate the relative closeness of each indicator to the ideal solution to generate a comprehensive evaluation index;

[0012] Score output and model optimization closed-loop: Output the comprehensive evaluation index and display a radar chart through a visualization interface. If the score is lower than the threshold, automatically push an optimization instruction to the training platform for model optimization.

[0013] Furthermore, the calculation of the model prediction error value includes:

[0014] By comparing the prediction result P(t) of the model with the actual measurement value A(t), the error e(t) is:

[0015] e(t) = |P(t) - A(t)|.

[0016] Furthermore, in the step of classifying the current operating state into a preset operating condition library based on the characteristics of production flow data, it includes selecting a similarity calculation method to match the preset operating condition library:

[0017]

[0018] x i and y i are the characteristic values of the current operating condition x and the operating condition y in the preset operating condition library respectively. n is the number of characteristics, and d represents the Euclidean distance between the two. The smaller the distance, the higher the similarity between the operating condition x and the operating condition y.

[0019] Furthermore, in the step of calculating stability indicators, reusability indicators, and economic indicators, it includes:

[0020] The stability indicator is evaluated by calculating the coefficient of variation of the prediction error within the sliding window, and the reusability indicator;

[0021] The stability indicator is measured by the coefficient of variation. CV > 15% is set as the threshold for instability. If CV exceeds this value, a warning is triggered;

[0022]

[0023] Among them, σ(e) is the standard deviation of the error, representing the degree of error fluctuation, and μ(e) is the mean value of the error, representing the central position of the error;

[0024] The reusability index is evaluated by the call frequency and cross-condition adaptability rate of the statistical model under the same type of working conditions:

[0025] Suppose there are N historical working conditions, and the effective operation times of the model are R i , R i is the effective operation times of the model under the working condition i, and the effective operation rate P of the model in the total working conditions is calculated by the formula:

[0026] N represents the total number of working conditions;

[0027] The economy index is evaluated by quantifying the impact of the model prediction on the actual economic benefit:

[0028] L = e(t) * T * S; where L is the economic loss, e(t) is the error between the model prediction value and the actual value, T is the operation duration, and S is the actual cost of the production data.

[0029] Furthermore, the steps of calculating the information entropy of each index by the entropy weight method include:

[0030] Calculate the entropy value of the j-th index:

[0031] Among them

[0032] x aj is the value of the j-th index under the a-th working condition;

[0033] Calculate the weight of the j-th index:

[0034]

[0035] Furthermore, the weighted standardization matrix includes:

[0036] Weight each standardized data to obtain the element v in the weighted matrix aj ;

[0037] V = [w j * x′ aj ,

[0038] V is the weighted matrix, and x′ aj is the standardized value of the j-th index under the a-th working condition.

[0039] For each indicator, calculate the ideal solution and the negative ideal solution:

[0040] The ideal solution is the maximum value of all indicators, and the negative ideal solution is the minimum value of all indicators;

[0041]

[0042] Calculate the proximity of each indicator to the ideal solution or the negative ideal solution:

[0043]

[0044] is the distance from the ideal solution under the ath working condition, the distance from the negative ideal solution under the th working condition.

[0045] A dynamic multi-dimensional quality evaluation system for a smart monitoring panel model in a thermal power plant, which is applied to a dynamic multi-dimensional quality evaluation method for a smart monitoring panel model in a thermal power plant. The system includes:

[0046] Data acquisition module: Real-time collect the production data stream of the thermal power plant and the output results of the monitoring panel model;

[0047] Monitoring panel model processing module: Process the input production flow data by the model, output the predicted value, and compare the predicted value with the actual measured value to generate an error value;

[0048] Working condition matching and historical data calling module: By analyzing the current operating condition characteristics, classify the current operating state into a preset working condition library; Under the matched working condition, retrieve relevant historical data from the historical database;

[0049] Multi-dimensional index calculation module: Calculate evaluation indexes of multiple dimensions;

[0050] Data standardization and weight assignment module: Use the range method to standardize each index to make it dimensionless, and calculate the information entropy of each index by the entropy weight method, and then assign weights;

[0051] TOPSIS comprehensive evaluation module: According to the standardized data and the weights calculated by the entropy weight method, construct a weighted standardized matrix, and calculate the distances of each solution from the ideal solution and the negative ideal solution through the TOPSIS model, and finally calculate the comprehensive evaluation index of each solution;

[0052] Scoring output and model optimization module: According to the TOPSIS comprehensive evaluation results, output the comprehensive score of each solution, and display it through a visual interface; If the score is lower than the set threshold, automatically push an optimization instruction to the model training platform for optimization.

[0053] The beneficial effects of the present invention are:

[0054] Multi - dimensional evaluation: The existing evaluation of the monitoring model relies on a single static index and cannot quantify the composite performance such as the stability and economy of the model under multiple working conditions. By introducing a multi - attribute decision - making method (such as the TOPSIS model based on the entropy weight method), the present invention realizes the multi - dimensional evaluation of the model, which can more comprehensively measure the quality and health of the model.

[0055] Objective quantification: The existing evaluation methods rely on expert scoring or qualitative analysis, lacking objective quantification criteria and being easily affected by human factors. The present invention provides objective quantification evaluation criteria through data - driven and automated calculations, reducing subjectivity and making the evaluation results more scientific and reliable.

[0056] Accurate optimization basis: The existing technology cannot accurately locate model defects through systematic evaluation results, resulting in the difficulty of using evaluation results to guide model iteration and optimization. Through multi - dimensional index calculation and comprehensive evaluation, the present invention can accurately locate model defects and provide a scientific basis for model optimization.

[0057] Dynamic evaluation and closed - loop feedback: The existing technology lacks an evaluation mechanism and a closed - loop feedback mechanism for the model under dynamic working conditions. The present invention can dynamically adjust the evaluation criteria according to different operating conditions of thermal power units and form a closed - loop feedback mechanism to guide the real - time adjustment of the model, improving the adaptability and performance of the model under different working conditions. Description of the Drawings

[0058] Figure 1 It is a schematic flow chart of a dynamic multi - dimensional quality evaluation method for an intelligent monitoring model in a thermal power plant.

[0059] Figure 2 It is a schematic flow chart of the technical process in an embodiment. Detailed Embodiments

[0060] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0061] In the description of the present application, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of the described features. In the description of the present application, "a plurality" means two or more unless otherwise specifically defined.

[0062] In the description of the present application, the term "for example" is used to mean "serving as an example, illustration, or explanation". Any embodiment described in the present application as "for example" is not necessarily to be construed as more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to implement and use the present invention. In the following description, details are set forth for purposes of explanation. It should be understood that those of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other instances, well-known structures and processes are not elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but rather to be in line with the broadest scope consistent with the principles and features disclosed in the present application.

[0063] In one embodiment, as Figure 1-2 shown: A dynamic multi-dimensional quality evaluation method for the intelligent monitoring panel model of a thermal power plant; comprising the following steps:

[0064] S1: Data input and model processing: Collect the real-time data stream of the DCS system, synchronously access the output result of the monitoring panel model, the model receives the production flow data and outputs a predicted value, compares it with the actual measured value, and generates a model prediction error value;

[0065] S2: Operating condition matching and historical data calling: Based on the characteristics of the production flow data, classify the current operating state into a preset operating condition library. After successful matching, retrieve the historical operating data, model statistical indicators, and model economic data under this operating condition from the historical database to participate in the index calculation;

[0066] S3: Multi-dimensional index calculation: Calculate the stability index, reusability index, and economic index;

[0067] S4: Data standardization and weight assignment: Use the range method to perform dimensionless processing on multiple indicators, and calculate the information entropy of each indicator through the entropy weight method to determine the weight;

[0068] S5: TOPSIS comprehensive evaluation: Construct a weighted standardized matrix, calculate the positive / negative ideal solutions through historical data statistics according to the current operating condition type, and use the Euclidean distance to calculate the relative closeness of each indicator to the ideal solution to generate a comprehensive evaluation index;

[0069] S6: Score output and model optimization closed-loop: Output the comprehensive evaluation index, and display a radar chart through a visual interface. If the score is lower than the threshold, automatically push an optimization instruction to the training platform to optimize the model.

[0070] In this embodiment, taking coal consumption data as an example, the DCS system is a distributed control system that collects data from real-time data streams, such as main steam pressure, flue gas temperature, denitration efficiency, etc. The sampling frequency is 1 Hz. The data is used for the analysis of the monitoring panel model. The monitoring panel model accepts and processes this data, generates a predicted value (taking coal consumption prediction as an example), and compares it with the actual measured value (coal consumption meter data); then, calculates the prediction error value to provide basic data for subsequent evaluation; according to the current operating conditions (such as load rate, fuel type, ambient temperature, etc.), the system classifies the current working condition into a preset working condition library (such as working condition 1, working condition 2, etc.). Assume that the current working condition is working condition 3; retrieves relevant data (such as historical coal consumption data, model reuse times, economic data, etc.) from the historical database of working condition 3 to participate in subsequent index calculations; uses a 1-hour window. If the standard deviation of the error within this window is 10 tons and the mean is 50 tons, counts the call frequency of the statistical model under the same type of working conditions. Under working condition 3, the model runs effectively 50 times, and the cross-working condition adaptation rate is 80%. Quantifies the impact of coal consumption deviation on economic benefits, and dimensionless processes various indicators (such as stability, reusability, economy) to make them unified in the [0,1] interval. Assume that the standardized value of the stability index is 0.8, the reusability is 0.6, and the economy is 0.9. Through the entropy weight method, the system assigns a weight of 40% to the stability index, and 30% to the reusability and economy respectively. The comprehensive evaluation index of the current coal consumption prediction model is 0.6, which is lower than the threshold of 0.7. The system automatically triggers model optimization and pushes an optimization instruction to the training platform, requiring the model to be improved.

[0071] In one embodiment, the calculation of the model prediction error value includes:

[0072] By comparing the prediction result P(t) of the model with the actual measured value A(t), the error e(t) is:

[0073] e(t) = |P(t) - A(t)|.

[0074] Assume that in a thermal power plant, the goal of the monitoring panel model to predict the coal consumption rate is to predict the coal consumption in the next 15 minutes, and the actual measured coal consumption data is obtained through real-time measurement by a coal consumption meter.

[0075] Model prediction value (for example, coal consumption prediction value): The model outputs the predicted coal consumption in the next 15 minutes. For example, the model predicts that the coal consumption in the next 15 minutes is 500 tons.

[0076] Actual measured value (for example, real-time coal consumption data): The actual coal consumption is obtained through real-time measurement. For example, the real-time measured coal consumption is 480 tons.

[0077] It shows that the error value between the predicted value obtained by the monitoring panel model and the actual value is 20 tons.

[0078] Further, in the step of classifying the current operating state into a preset working condition library based on the characteristics of production flow data, the method includes selecting a similarity calculation method to match the preset working condition library:

[0079]

[0080] x i and y i are the characteristic values of the current working condition x and the working condition y in the preset working condition library respectively, n is the number of characteristics, d represents the Euclidean distance between the two, and the smaller the distance, the higher the similarity between the working condition x and the working condition y.

[0081] First, define the comparison characteristics between the current working condition and the preset working condition. For example, load rate, fuel type, environmental temperature, etc. Compare the characteristics of the current working condition with those of each historical working condition, calculate the Euclidean distance, compare the calculated Euclidean distance with the distances of all historical working conditions, and select the historical working condition with the smallest Euclidean distance from the current working condition as the matching result. The smaller the distance, the higher the similarity between the two.

[0082] If the calculated distance between the current working condition and working condition 3 is the smallest, the system will select working condition 3 as the matching historical working condition, and retrieve information such as the historical operation data and error records of the model from the historical database of working condition 3.

[0083] Further, the steps of calculating the stability index, reusability index, and economic index include:

[0084] The stability index is evaluated by calculating the coefficient of variation of the prediction error within the sliding window, and the reusability index;

[0085] The stability index is measured by the coefficient of variation, and CV > 15% is set as the threshold for instability. If CV exceeds this value, an early warning will be triggered;

[0086]

[0087] Among them, σ(e) is the standard deviation of the error, indicating the degree of error fluctuation, and μ(e) is the mean of the error, indicating the central position of the error;

[0088] Suppose that within the past 1 hour, the error data of the model is: [10, 15, 20, 10, 25] tons (the unit is coal consumption error). The standard deviation and mean of the error can be calculated, and the reusability index is evaluated by statistically analyzing the call frequency of the model under the same type of working conditions and the cross-working condition adaptation rate:

[0089] Mean:

[0090] Standard deviation:

[0091]

[0092] Coefficient of variation:

[0093] If the set threshold is 15%, since CV > 15%, a warning of model instability is triggered, indicating that the prediction error of the model fluctuates greatly (due to the small amount of data used in this example, the coefficient of variation is large. Under normal circumstances, the standard deviation of historical data is around 15).

[0094] Suppose there are N historical working conditions, and the effective operation times of the model is R i , R i is the effective operation times of the model under working condition i. The effective operation rate P of the model under the total working conditions is calculated by the formula:

[0095] N represents the total number of working conditions;

[0096] The economic index is evaluated by quantifying the impact of the model prediction on the actual economic benefit:

[0097] L = e(t) * T * S; where L is the economic loss, e(t) is the error between the model prediction value and the actual value, T is the operation duration, and S is the actual cost of production data.

[0098] Taking the coal consumption error of 20 tons mentioned above as an example:

[0099] Economic loss: L = 20 * 24 * 300 = 144000;

[0100] The economic loss caused by the coal consumption error of the monitoring model is 144000 yuan. Therefore, it is necessary to optimize and adjust the monitoring model.

[0101] Furthermore, in the step of calculating the information entropy of each index by the entropy weight method, it includes:

[0102] Calculating the entropy value of the j-th index:

[0103] where

[0104] x aj is the value of the j-th index under the a-th working condition;

[0105] Calculating the weight of the j-th index:

[0106]

[0107] Once the entropy value of each index is calculated, the weight of each index can be calculated according to the entropy value. The weight reflects the influence of this index on the final comprehensive evaluation. The smaller the entropy value of an index, the greater its weight, indicating that it contributes more to the evaluation result.

[0108] Further, the weighted normalization matrix includes:

[0109] Each normalized data is weighted to obtain the element v in the weighted matrix aj ;

[0110] V = [w j * x′ aj ,

[0111] V is the weighted matrix, and x′ aj is the normalized value of the j-th index under the a-th working condition.

[0112] For each index, the ideal solution and the negative ideal solution are:

[0113] The ideal solution is the maximum value of all indices, and the negative ideal solution is the minimum value of all indices;

[0114]

[0115] Calculate the proximity of each index to the ideal solution or the negative ideal solution:

[0116]

[0117] is the distance from the a-th working condition to the ideal solution, the distance from the a-th working condition to the negative ideal solution.

[0118] By combining each normalized index value with the weights calculated by the entropy weight method, a weighted normalization matrix is generated. This matrix is used to further calculate the proximity to the ideal solution and the negative ideal solution. The weighted normalization matrix assigns appropriate weights to the normalized data to ensure a reasonable contribution ratio of different indices to the final evaluation.

[0119] Suppose there are the following data on coal consumption error:

[0120] The normalized coal consumption errors (for working conditions 1, 2, and 3) are respectively:

[0121] Working condition 1: 0.48

[0122] Working condition 2: 0.42

[0123] Working condition 3: 0.54

[0124] Suppose the ideal solution and the negative ideal solution are as follows:

[0125] Ideal solution: Stability is 0.54, and economy is 0.28.

[0126] Negative ideal solution: Stability is 0.42, and economy is 0.20.

[0127] Then, calculate the proximity of each working condition:

[0128] For working condition 1: The Euclidean distances from the ideal solution and the negative ideal solution are respectively:

[0129]

[0130] Proximity = 0.5

[0131] For working conditions 2 and 3, they can also be calculated in this way, and finally the corresponding proximity indices are obtained.

[0132] The closer the proximity index is to 1, the closer the model performance under the working condition is to the ideal solution, indicating that the coal consumption prediction error is smaller and the model performance is better.

[0133] The comprehensive evaluation index obtained through TOPSIS comprehensive evaluation can show the model performance under each working condition. According to the comprehensive score, we can decide whether to optimize the current model. If the performance of the model under certain working conditions is lower than the set threshold (for example, the proximity index is lower than 0.7), the optimization mechanism is triggered to improve the model.

[0134] Suppose the following comprehensive scores are obtained:

[0135] Proximity index of working condition 1: 0.5

[0136] Proximity index of working condition 2: 0.7

[0137] Proximity index of working condition 3: 0.8

[0138] Threshold setting: If the proximity is lower than 0.7, the system will trigger optimization. The score of working condition 1 is lower than 0.7, so optimization is required.

[0139] Model optimization may include:

[0140] Insufficient stability: Increase adversarial training samples or adjust model parameters to make it more stable under different working conditions.

[0141] Economic deviation: Recalibrate the input data, especially the coal consumption prediction under high-load and low-load working conditions.

[0142] Low reusability: Expand the working condition library and increase cross-working-condition training samples.

[0143] These optimization measures help the model to continuously adjust to improve the accuracy and reliability of prediction.

[0144] A dynamic multi-dimensional quality evaluation system for a smart monitoring panel model of a thermal power plant, which is applied to a dynamic multi-dimensional quality evaluation method for a smart monitoring panel model of a thermal power plant. The system includes:

[0145] Data acquisition module: Real-time acquisition of the production data stream of the thermal power plant and the output results of the monitoring model;

[0146] Monitoring model processing module: The model processes the input production flow data, outputs the predicted value, compares the predicted value with the actual measured value, and generates an error value;

[0147] Operating condition matching and historical data calling module: By analyzing the characteristics of the current operating condition, classify the current operating state into a preset operating condition library; Under the matched operating condition, retrieve relevant historical data from the historical database;

[0148] Multi-dimensional index calculation module: Calculate evaluation indexes of multiple dimensions;

[0149] Data standardization and weight assignment module: Use the range method to standardize each index to make it dimensionless, and calculate the information entropy of each index through the entropy weight method, and then assign weights;

[0150] TOPSIS comprehensive evaluation module: According to the standardized data and the weights calculated by the entropy weight method, construct a weighted standardized matrix, and calculate the distances between each solution and the ideal solution and the negative ideal solution through the TOPSIS model, and finally calculate the comprehensive evaluation index of each solution;

[0151] Score output and model optimization module: According to the TOPSIS comprehensive evaluation result, output the comprehensive score of each solution and display it through the visualization interface; If the score is lower than the set threshold, automatically push the optimization instruction to the model training platform for optimization.

[0152] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concept. Therefore, the appended claims are intended to be construed to include the preferred embodiments and all changes and modifications falling within the scope of the present invention.

[0153] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A dynamic multi-dimensional quality evaluation method for the intelligent monitoring panel model of a thermal power plant, characterized in that, It includes the following steps: Data input and model processing: Collect the real-time data stream of the DCS system, synchronously access the output result of the monitoring model. The model receives the production flow data and outputs the predicted value, compares it with the actual measurement value, and generates the model prediction error value; Operating condition matching and historical data calling: Based on the characteristics of the production flow data, classify the current operating state into the preset operating condition library. After successful matching, retrieve the historical operating data, model statistical indicators, and model economic data under this operating condition from the historical database to participate in the index calculation; Multi-dimensional index calculation: Calculate the stability index, reusability index, and economic index; Data standardization and weight assignment: Use the range method to perform dimensionless processing on multiple indicators, and calculate the information entropy of each indicator through the entropy weight method to determine the weight; TOPSIS comprehensive evaluation: Construct a weighted standardized matrix, calculate the positive / negative ideal solutions through historical data statistics according to the current operating condition type, and use the Euclidean distance to calculate the relative closeness of each indicator to the ideal solution to generate the comprehensive evaluation index; Score output and model optimization closed-loop: Output the comprehensive evaluation index, and display the radar chart through the visualization interface. If the score is lower than the threshold, automatically push the optimization instruction to the training platform for model optimization.

2. The dynamic multi-dimensional quality evaluation method for the intelligent monitoring panel model of a thermal power plant according to claim 1, characterized in that, The calculation of the model prediction error value includes: By comparing the prediction result P(t) of the model with the actual measurement value A(t), the error e(t) is: e(t) = |P(t) - A(t)|.

3. The dynamic multi-dimensional quality evaluation method for the intelligent monitoring panel model of a thermal power plant according to claim 2, wherein In the step of classifying the current operating state into the preset operating condition library based on the characteristics of the production flow data, it includes selecting a similarity calculation method to match the preset operating condition library: x i and y i are the characteristic values of the current working condition x and the working condition y in the preset working condition library respectively. n is the number of characteristics, and d represents the Euclidean distance between the two. The smaller the distance, the higher the similarity between the working condition x and the working condition y.

4. The dynamic multi-dimensional quality evaluation method for the intelligent monitoring panel model of a thermal power plant according to claim 3, wherein, In the step of calculating the stability index, reusability index, and economic index, it includes: The stability index is evaluated by calculating the coefficient of variation of the prediction error within the sliding window, and the reusability index; The stability index is measured by the coefficient of variation. CV > 15% is set as the unstable threshold. If CV exceeds this value, an alarm is triggered; Among them, σ(e) is the standard deviation of the error, representing the degree of error fluctuation, and μ(e) is the mean value of the error, representing the central position of the error; The reusability index is evaluated by statistically counting the call frequency of the model under the same type of operating conditions and the cross-operating condition adaptation rate: Assume there are N historical operating conditions, and the effective operation times of the model is R i , R i is the effective operation times of the model under condition i. The effective operation rate P of the model under the total conditions is calculated by the formula as follows: N represents the total number of operating conditions; The economic index is evaluated by quantifying the impact of the model prediction on the actual economic benefit: L = e(t) * T * S; where L is the economic loss, e(t) is the error between the model prediction value and the actual value, T is the operating duration, and S is the actual cost of the production data.

5. The dynamic multi-dimensional quality evaluation method for the intelligent monitoring panel model of a thermal power plant according to claim 4, characterized in that In the step of calculating the information entropy of each indicator through the entropy weight method, it includes: Calculate the entropy value of the j-th indicator: wherein x aj is the value of the j-th index under the a-th operating condition; Calculate the weight of the j-th indicator:

6. The dynamic multi-dimensional quality evaluation method for the intelligent monitoring panel model of a thermal power plant according to claim 5, characterized in that, The weighted standardized matrix includes: Weight each standardized data to obtain the element v in the weighted matrix aj ; V is the weighted matrix, x′ aj is the standardized value of the j-th index under the a-th operating condition. For each indicator, there are ideal solution and negative ideal solution: The ideal solution is the maximum value of all indicators, and the negative ideal solution is the minimum value of all indicators; Calculate the closeness of each indicator to the ideal solution or negative ideal solution: is the distance from the ideal solution under the a-th working condition, and the distance from the negative ideal solution under the -th working condition.

7. A dynamic multi-dimensional quality evaluation system for the intelligent monitoring panel model of a thermal power plant, characterized in that, Applied to the dynamic multi-dimensional quality evaluation method of a smart monitoring model for a thermal power plant as described in any one of claims 1-6, the system includes: Data acquisition module: Real-time collect the production data stream of the thermal power plant and the output result of the monitoring model; Supervision and inventory model processing module: The module processes the input production flow data of the model, outputs the predicted value, compares the predicted value with the actual measured value, and generates the error value; Operating condition matching and historical data calling module: By analyzing the current operating condition characteristics, classify the current operating state into the preset operating condition library; Under the matched operating condition, retrieve relevant historical data from the historical database; Multi-dimensional index calculation module: Calculate evaluation indexes of multiple dimensions; Data standardization and weight assignment module: Use the range method to standardize each index to make it dimensionless, and calculate the information entropy of each index through the entropy weight method, and then assign weights; TOPSIS comprehensive evaluation module: According to the standardized data and the weights calculated by the entropy weight method, construct a weighted standardized matrix, calculate the distances of each solution from the ideal solution and the negative ideal solution through the TOPSIS model, and finally calculate the comprehensive evaluation index of each solution; Score output and model optimization module: According to the TOPSIS comprehensive evaluation result, output the comprehensive score of each solution and display it through the visualization interface; If the score is lower than the set threshold, automatically push the optimization instruction to the model training platform for optimization.

Images