Online evaluation methods, devices, equipment and media for primary frequency regulation capability of thermal power units

Abstract

This application discloses an online evaluation method, apparatus, equipment, and medium for the primary frequency regulation capability of thermal power units, relating to the field of thermal power unit frequency regulation. It employs the UMAP algorithm to perform feature dimensionality reduction and reconstruction on the high-dimensional operating dataset of the thermal power unit to be evaluated. Unlike the linear dimensionality reduction limitations of traditional PCA, UMAP can capture the manifold structure of the data while preserving local correlations and global distribution, significantly reducing the dimensionality of features. The obtained unit output feature set is then input into a pre-trained primary frequency regulation output prediction model for thermal power units to achieve online frequency regulation output prediction. In terms of evaluation dimensions, it breaks through the traditional framework that only focuses on frequency regulation performance. Based on the unit output prediction results and the high-dimensional operating dataset, the evaluation dimensions are quantified to construct a set of evaluation factors. Finally, the entropy-weighted TOPSIS method is used to calculate the weight allocation of the evaluation factor set, generating a comprehensive evaluation index for primary frequency regulation capability, which intuitively reflects the quality of the unit's frequency regulation capability and improves the accuracy of frequency regulation capability evaluation.

Description

Technical Field

[0001] This application relates to the field of frequency regulation technology for thermal power units, and in particular to a method, apparatus, equipment and medium for online evaluation of the primary frequency regulation capability of thermal power units. Background Technology

[0002] With the rapid depletion of fossil fuels and increasing environmental pollution, there is a growing demand for a clean, friendly, flexible, and safe new power system. As new energy sources are increasingly integrated into the grid, grid frequency fluctuations are becoming more frequent and complex, making the real-time response and dynamic assessment capabilities of primary frequency regulation for thermal power units increasingly urgent. Therefore, traditional coal-fired power units are rapidly transforming from primary power sources to regulating power sources, undertaking a heavier workload in grid frequency regulation.

[0003] In related technologies, the primary frequency regulation process of coal-fired power units exhibits nonlinear and time-varying characteristics. During unit operation, changes in operating conditions, different control parameters, and alterations in the inherent characteristics of the coal-fired power unit can all lead to a reduction in the primary frequency regulation capability. Furthermore, the applicant recognizes that existing evaluation systems rely on offline testing, which can only reflect the frequency regulation capability under specific static conditions, thus limiting the effectiveness of frequency regulation capability assessment. Moreover, these systems often focus on performance indicators such as response time and settling time, resulting in one-sided evaluations that fail to reflect the unit's comprehensive frequency regulation capability under varying operating conditions. Additionally, existing online evaluation systems cannot simulate dynamic disturbance processes, leading to large model prediction errors and significant deviations between the evaluation results and the actual frequency regulation capability. Summary of the Invention

[0004] In view of this, this application provides an online evaluation method, device, equipment and medium for the primary frequency regulation capability of thermal power units. The main purpose is to solve the limitations of the existing evaluation system, such as the lack of multi-dimensional evaluation elements and insufficient online evaluation capability. At the same time, it is also limited by the evaluation method that must be carried out under off-grid conditions, which restricts the effectiveness of frequency regulation capability evaluation.

[0005] According to a first aspect of this application, an online evaluation method for the primary frequency regulation capability of a thermal power unit is provided, the method comprising:

[0006] The real-time operating data of the thermal power unit to be evaluated is obtained, and the real-time operating data is preprocessed to obtain a high-dimensional operating dataset.

[0007] The UMAP algorithm is used to perform feature dimensionality reduction and reconstruction on the high-dimensional operating dataset to obtain the unit output feature set. The unit output feature set is then input into a pre-trained thermal power unit primary frequency regulation output prediction model to obtain the unit output prediction result.

[0008] Based on the unit output prediction results and the high-dimensional operation dataset, the evaluation dimensions of the thermal power unit to be evaluated are quantitatively calculated to obtain a set of evaluation factors.

[0009] The TOPSIS method of entropy weight is used to assign weights to the set of evaluation factors. Based on the weight assignment results, a weighted average is calculated on the set of evaluation factors to generate a comprehensive evaluation index of the primary frequency regulation capability of the thermal power unit to be evaluated.

[0010] According to a second aspect of this application, an online evaluation device for the primary frequency regulation capability of a thermal power unit is provided, the device comprising:

[0011] The preprocessing module is used to acquire real-time operating data of the thermal power unit to be evaluated, and to preprocess the real-time operating data to obtain a high-dimensional operating dataset.

[0012] The power output prediction module is used to perform feature dimensionality reduction and reconstruction processing on the high-dimensional operating dataset using the UMAP algorithm to obtain the unit power output feature set, and input the unit power output feature set into the pre-trained thermal power unit primary frequency regulation power output prediction model to obtain the unit power output prediction result.

[0013] The factor calculation module is used to perform quantitative calculation of the evaluation dimensions of the thermal power unit to be evaluated based on the unit output prediction results and the high-dimensional operation dataset, and obtain a set of evaluation factors.

[0014] The evaluation module is used to assign weights to the set of evaluation factors using the entropy-weighted TOPSIS method, and to calculate a weighted average of the set of evaluation factors based on the weight assignment results, thereby generating a comprehensive evaluation index of the primary frequency regulation capability of the thermal power unit to be evaluated.

[0015] According to a third aspect of this application, an apparatus is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the method described in any of the first aspects above.

[0016] According to a fourth aspect of this application, a medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any one of the first aspects above.

[0017] By employing the above technical solutions, the technical solutions provided in the embodiments of this application have at least the following advantages:

[0018] This application provides a method, apparatus, equipment, and medium for online evaluation of the primary frequency regulation capability of thermal power units. The application acquires real-time operating data of the thermal power unit to be evaluated, preprocesses the real-time operating data to obtain a high-dimensional operating dataset, and then uses the UMAP algorithm to perform feature dimensionality reduction and reconstruction on the high-dimensional operating dataset to obtain a unit output feature set. Unlike the linear dimensionality reduction limitations of the traditional PCA algorithm, the UMAP algorithm can capture the manifold structure of the data while preserving local correlations and global distribution, significantly reducing the dimensionality of the features and improving the accuracy of the model prediction results. Subsequently, the unit output feature set is input into a pre-trained primary frequency regulation output prediction model for thermal power units to obtain the unit output prediction results, achieving online real-time frequency regulation output prediction. In terms of evaluation dimensions, this application breaks through the traditional framework that only focuses on frequency regulation performance. Based on the unit output prediction results and the high-dimensional operating dataset, it quantifies the evaluation dimensions of the thermal power unit to be evaluated, constructing a set of evaluation factors from three dimensions: primary frequency regulation performance, safety, and economy. Finally, the entropy-weighted TOPSIS method is used to assign weights to the evaluation factor set. Based on the weight assignment results, a weighted average is calculated on the evaluation factor set to generate a comprehensive evaluation index for the primary frequency regulation capability of the thermal power unit to be evaluated. The entropy-weighted TOPSIS method can achieve objective weighting and quantify the gap between each evaluation factor and the optimal state using positive and negative ideal solutions. The final output of the comprehensive evaluation index can intuitively reflect the quality of the unit's frequency regulation capability and improve the accuracy of frequency regulation capability evaluation.

[0019] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0020] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0021] Figure 1 This paper illustrates a flowchart of a method for online evaluation of the primary frequency regulation capability of a thermal power unit according to an embodiment of this application.

[0022] Figure 2 This paper illustrates a flowchart of another method for online evaluation of the primary frequency regulation capability of thermal power units provided in an embodiment of this application.

[0023] Figure 3 A schematic diagram of a BiLSTM-Attention structure provided in an embodiment of this application is shown;

[0024] Figure 4 This paper illustrates a flowchart of an online evaluation method for the primary frequency regulation capability of a thermal power unit, provided in an embodiment of this application.

[0025] Figure 5 This illustration shows a structural schematic diagram of an online evaluation method for the primary frequency regulation capability of a thermal power unit, provided in an embodiment of this application.

[0026] Figure 6 A schematic diagram of the device structure of an embodiment of this application is shown. Detailed Implementation

[0027] In the description of this application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0028] Exemplary embodiments of the present application will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this application will be thorough and complete, and will fully convey the scope of the present application to those skilled in the art.

[0029] Existing methods for assessing primary frequency regulation capacity are insufficient in terms of safety and economy, resulting in incomplete comprehensive evaluation results. Furthermore, under the influence of boiler-side regulation characteristics, the accuracy of existing assessment methods for assessing primary frequency regulation capacity under deep peak-shaving conditions is also inadequate. To address this issue, this application proposes an online assessment method for the primary frequency regulation capacity of thermal power units. This method collects operating data from thermal power units, including grid frequency, AGC commands, valve commands, valve opening, main steam pressure, regulating stage steam pressure, reheat steam pressure, and real-time coal consumption. The UMPA algorithm is used to analyze and process the thermal power unit operating data, reducing the correlation of high-dimensional time-frequency features and replacing the pooling operation before the fully connected layer of the neural network. A neural network suitable for predicting the primary frequency regulation output of coal-fired units is established. After obtaining the predicted unit output, the primary frequency regulation performance, safety, and economic indicators of the unit are quantitatively calculated to obtain their respective evaluation factors. The weight of each evaluation factor is determined using the entropy-weighted TOPSIS method, and the weighted average is calculated as the final comprehensive evaluation index. The implementing entity of this application may be a primary frequency modulation capability online assessment system. The primary frequency modulation capability online assessment system provides services to users by relying on the computing power of the server. The server may be an independent server or a server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0030] This application provides an online evaluation method for the primary frequency regulation capability of thermal power units, such as... Figure 1 As shown, the method includes:

[0031] 101. Obtain the real-time operating data of the thermal power unit to be evaluated, preprocess the real-time operating data, and obtain a high-dimensional operating dataset.

[0032] In this embodiment, real-time operational data replaces traditional offline test data, overcoming the lag of static testing. Real-time data can track the current state of the unit, allowing the evaluation results to be synchronized with the current frequency regulation capability, supporting real-time frequency regulation decisions of the power grid. Real-time operational data consists of multi-dimensional operating parameters, covering grid frequency, AGC commands, valve commands, valve opening, main steam pressure, regulating stage steam pressure, reheat steam pressure, and unit heating output, avoiding evaluation biases caused by missing parameters. Since using raw data directly can lead to noise interfering with UMAP dimensionality reduction manifold structure capture, or causing slow model training convergence and large prediction errors, preprocessing methods such as denoising, missing value imputation, and normalization are used to clean up noise in high-dimensional data and unify data scale.

[0033] 102. The UMAP algorithm is used to perform feature dimensionality reduction and reconstruction on the high-dimensional operating dataset to obtain the unit output feature set. The unit output feature set is then input into the pre-trained thermal power unit primary frequency regulation output prediction model to obtain the unit output prediction result.

[0034] In this embodiment, the high-dimensional operating data of thermal power units naturally exhibits nonlinear correlations between the boiler and turbine. Traditional PCA dimensionality reduction methods can only handle linear relationships, easily disrupting the manifold structure of the data and leading to the loss of key correlations. UMAP, as a nonlinear dimensionality reduction algorithm, captures the manifold topology of the data by constructing a local fuzzy simplex set. This preserves local correlations, avoids feature logic breakage after dimensionality reduction, and restores the global distribution. The primary frequency regulation output prediction model for thermal power units uses a BiLSTM-Attention structure. BiLSTM can encode information sequentially from back to front, exhibiting stronger ability to capture bidirectional dependencies and effectively extracting information from time-series data. Attention, on the other hand, can focus limited attention on interpreting key information, thus acquiring information more efficiently. Therefore, combining Attention with a BiLSTM network can enable the network to capture useful information more efficiently and improve network performance.

[0035] 103. Based on the unit output prediction results and high-dimensional operation dataset, the evaluation dimensions of the thermal power units to be evaluated are quantitatively calculated to obtain a set of evaluation factors.

[0036] In this embodiment, the unit output prediction results can simulate the dynamic evolution of output during frequency regulation, while high-dimensional operating data can record the actual state of the equipment during frequency regulation. The linkage between the two can restore the dynamic details of the real frequency regulation scenario. Moreover, traditional primary frequency regulation assessments often focus on performance indicators such as response time and settling time, ignoring equipment safety boundaries (such as the risk of exceeding power limits) and economic costs (such as excessive coal consumption) during frequency regulation. Therefore, by linking the unit output prediction results with high-dimensional operating data, a three-dimensional evaluation system of performance, safety, and economy can be constructed, thereby achieving a full-scenario, quantifiable, and guiding assessment of primary frequency regulation capabilities.

[0037] 104. The entropy weight TOPSIS method is used to assign weights to the evaluation factor set. Based on the weight assignment results, the weighted average of the evaluation factor set is calculated to generate a comprehensive evaluation index of the primary frequency regulation capability of the thermal power unit to be evaluated.

[0038] In this embodiment, the entropy weight method objectively quantifies the weights of evaluation factors, avoiding subjective intervention. Combined with TOPSIS, it accurately determines the proximity of the evaluation object to the ideal solution. Finally, a weighted average is used to integrate multi-dimensional indicators, generating a scientific and systematic comprehensive evaluation result. This solves the problems of large biases in manual weighting, one-dimensional evaluation bias, and poor adaptability to operating conditions in traditional evaluations. The entropy weight method automatically assigns weights based on the dispersion of evaluation factors. If the grid frequency fluctuates drastically, the data differences of primary frequency regulation performance evaluation factors will significantly increase, and the entropy weight will automatically increase their weights to prioritize frequency regulation speed. If coal prices soar, the dispersion of primary frequency regulation economic evaluation factors increases, and the weights naturally tilt, avoiding excessive cost consumption for frequency regulation. This data-driven weighting mechanism allows the evaluation system to dynamically adapt to changes in grid demand and operating costs, making it more accurate and realistic than subjective assignment. The TOPSIS method transforms multi-dimensional evaluation factors into distances from the ideal state through positive and negative ideal solutions. This not only accurately exposes the comprehensive shortcomings of a single unit in terms of performance, safety, and economy, but also helps to promptly identify problems in the primary frequency regulation process by using a single quantitative value with weighted average output.

[0039] This application provides an online evaluation method for the primary frequency regulation capability of thermal power units. Compared with existing technologies, this application obtains real-time operating data of the thermal power unit to be evaluated, preprocesses the real-time operating data to obtain a high-dimensional operating dataset. Then, the UMAP algorithm is used to perform feature dimensionality reduction and reconstruction on the high-dimensional operating dataset to obtain a unit output feature set. Unlike the linear dimensionality reduction limitations of the traditional PCA algorithm, the UMAP algorithm can capture the manifold structure of the data while preserving local correlations and global distribution, significantly reducing the dimensionality of the features and improving the accuracy of the model prediction results. Subsequently, the unit output feature set is input into a pre-trained primary frequency regulation output prediction model for thermal power units to obtain the unit output prediction results, achieving online real-time frequency regulation output prediction. In terms of evaluation dimensions, this method breaks through the traditional framework that only focuses on frequency regulation performance. Based on the unit output prediction results and the high-dimensional operating dataset, the evaluation dimensions of the thermal power unit to be evaluated are quantitatively calculated, constructing a set of evaluation factors from three dimensions: primary frequency regulation performance, safety, and economy. Finally, the entropy-weighted TOPSIS method is used to assign weights to the evaluation factor set. Based on the weight assignment results, a weighted average is calculated on the evaluation factor set to generate a comprehensive evaluation index for the primary frequency regulation capability of the thermal power unit to be evaluated. The entropy-weighted TOPSIS method can achieve objective weighting and quantify the gap between each evaluation factor and the optimal state using positive and negative ideal solutions. The final output of the comprehensive evaluation index can intuitively reflect the quality of the unit's frequency regulation capability and improve the accuracy of frequency regulation capability evaluation.

[0040] Furthermore, as a refinement and extension of the specific implementation methods of the above embodiments, and in order to fully illustrate the specific implementation process of this embodiment, this application provides another method for online evaluation of the primary frequency regulation capability of thermal power units, such as... Figure 2 As shown, the method includes:

[0041] 201. Obtain the real-time operating data of the thermal power unit to be evaluated, and preprocess the real-time operating data to obtain a high-dimensional operating dataset.

[0042] In this embodiment of the application, in order to obtain more accurate and comprehensive frequency regulation capability evaluation factors based on accurate prediction of unit output, and then obtain more representative primary frequency regulation capability evaluation results after reasonable weight allocation, it is necessary to collect real-time operating data of the thermal power unit to be evaluated, including grid frequency, AGC commands, valve commands, valve opening degree, main steam pressure, regulating stage steam pressure, reheat steam pressure, and unit heat supply, etc., and then perform preprocessing in combination with unit characteristic parameters and grid connection standards.

[0043] 202. The UMAP algorithm is used to perform feature dimensionality reduction and reconstruction on the high-dimensional operating dataset to obtain the unit output feature set.

[0044] In this embodiment, the operating environment of thermal power units is highly variable. Furthermore, based on the physical principles of the boiler-generator coordination system, numerous parameters are involved in calculating the unit's output power during the primary frequency regulation process, and these operating parameters may still have certain coupling relationships. To obtain more representative features, it is necessary to reconstruct the high-dimensional operating dataset. The UMAP method is used to reduce the dimensionality of the collected real-time operating data of thermal power units, reconstructing more representative power output prediction features. Specifically, by analyzing the local neighbor samples of each data sample, a local fuzzy simplex set centered on that data sample is obtained. The distance to the nearest sample center is then used as the weight of this local simplex set. During this process, it is ensured that each sample has at least one adjacent edge with a weight of 1. This connects all the local fuzzy simplex sets of data samples. The local fuzzy simplex sets of data samples are merged based on a symmetric directed weighted graph. Finally, the merged set is mapped to a low-dimensional space using cross-entropy calculation to obtain a new set of unit output features.

[0045] Specifically, high-dimensional running datasets The data points in the dataset are used as nodes to obtain multiple nodes. A high-dimensional dataset includes s types of parameters, with k data points for each parameter, forming an s x k matrix. Each element of the matrix can be considered a node, resulting in a total of... There are 10 nodes. Calculate the local distance and length parameters for each node, as shown in Formula 1 below:

[0046] Formula 1:

[0047]

[0048] in, Represents a node Local distances are the foundation for establishing local relationships in the sample space. Represents a node The length parameter is used to ensure that at least one node is related to the node. The edge weight of connected nodes is 1. Represents nodes The node with the smallest Euclidean distance. Represents a node With nodes The Euclidean distance, where n represents the number of nodes.

[0049] Next, a high-dimensional directed weighted graph is constructed using multiple nodes and the local distance and length parameters of each node. It includes all adjacent edges of a node, captures the local adjacency relationships of data in a high-dimensional space, and utilizes a high-dimensional directed weighted graph. Constructing high-dimensional undirected weighted graphs using symmetry properties As shown in Formula 2 below:

[0050] Formula 2:

[0051] in, This represents a high-dimensional undirected weighted graph. This represents a high-dimensional directed weighted graph. This represents the Hadamard product. Weighted graphs can characterize the topological structure of data, and since some nodes are nearest neighbors, some adjacent edges have directionality; therefore, an undirected weighted matrix is ​​constructed.

[0052] The equivalent weighted graph H of the feature dataset is a low-dimensional representation of the weighting matrix. To obtain the optimal representation result, spectral embedding based on the symmetric normalized Laplacian operator is performed to obtain a new point set. Specifically, a high-dimensional undirected weighting matrix B is constructed using the high-dimensional undirected weighted graph G. The high-dimensional undirected weighting matrix B is mapped to a low-dimensional space to obtain the low-dimensional space coordinates of each node. The low-dimensional equivalent weighted graph H is constructed using the low-dimensional space coordinates of multiple nodes, as shown in Formula 3 below:

[0053] Formula 3:

[0054] in, Represents the nodes in the low-dimensional equivalent weighted graph H With nodes Association weights Represents a node Low-dimensional space coordinates, Represents a node The low-dimensional coordinates are given by , where a represents the first hyperparameter and b represents the second hyperparameter.

[0055] Then, the cross-entropy metric is used to construct the cross-entropy topological difference metric function between the high-dimensional undirected weighted graph and the high-dimensional undirected weighted graph G, so that the low-dimensional equivalent weighted graph H can match the source data topology captured by the high-dimensional undirected weighted graph G, thereby obtaining a low-dimensional representation of the overall data, as shown in Formula 4 below:

[0056] Formula 4:

[0057] in, Let represent the cross-entropy topological difference measure function between a high-dimensional undirected weighted graph G and a low-dimensional equivalent weighted graph H. Represents nodes in a high-dimensional undirected weighted graph With nodes Association weights Represents nodes in a low-dimensional equivalent weighted graph With nodes The association weight.

[0058] Finally, the stochastic gradient descent algorithm is used to optimize the cross-entropy topology difference metric function to obtain a low-dimensional feature set, which is then used as the unit output feature set.

[0059] During normal operation of a power unit, numerous operating parameters are involved on both the turbine and boiler sides, and these parameters may still exhibit certain coupling relationships. Calculations based on the physical principles of the coal-fired power unit's boiler-turbine coordination system struggle to comprehensively consider the large volume of variable data. Traditional feature engineering's simple screening not only fails to retain key information but also hinders network learning. Therefore, this application uses UMAP to analyze the original high-dimensional steady-state features, reducing the correlation of high-dimensional time-frequency features. By extracting the manifold structure features within the original sample classes and the local structure features of adjacent classes, the features are embedded from the high-dimensional space into the target space to preserve the global relationships between all sample classes, significantly reducing the feature dimensionality and improving the accuracy of the model's prediction results.

[0060] 203. Input the unit output feature set into the pre-trained primary frequency regulation output prediction model of the thermal power unit to obtain the unit output prediction result.

[0061] In this embodiment, the primary frequency regulation capability of a coal-fired power unit is mainly reflected in its ability to quickly pull the frequency back to the standard frequency when the unit's rotor frequency deviates from the grid's standard frequency of 50Hz. It is difficult to propose a systematic and quantitative evaluation index for assessing the primary frequency regulation capability of a unit using grid-connected operation data. Current practical engineering assessments of the primary frequency regulation capability of coal-fired power units are based on data from off-grid primary frequency regulation tests. In actual operation, the primary frequency regulation capability of a coal-fired power unit is affected by various factors and may vary under different loads and operating environments. Even under the same conditions, the primary frequency regulation capability may differ due to unit operating losses. However, frequent off-grid primary frequency regulation tests would exacerbate unit losses, increase power plant expenses, and in severe cases, lead to insufficient grid frequency regulation resources, impacting grid stability, which is impractical in real-world engineering applications. Therefore, it is necessary to predict the output of the unit after the AGC command under the current operating conditions is disturbed, until it stabilizes to the target value, which is considered to be the completion of one frequency regulation test, and then obtain the frequency regulation capability evaluation factor.

[0062] Therefore, the assessment of the primary frequency regulation capability of coal-fired power units using a mechanistic model is based on the evaluation of the unit's output power during the primary frequency regulation process, calculated according to the physical principles of the coal-fired boiler-generator coordination system. Mechanistic models involve numerous characteristic parameters during the primary frequency regulation process of coal-fired power units and often fail to achieve ideal results. Therefore, neural network technology is used to establish the relationship between the unit's operating input and output, thereby creating a neural network applicable to predicting the primary frequency regulation output of coal-fired power units. The unit's output feature set is input into a pre-trained primary frequency regulation output prediction model for thermal power units to obtain the predicted output results, maximizing the network's learning ability. The primary frequency regulation output prediction model for thermal power units is continuously trained using historical data from the unit over a period of time. After training, current online data is used as input to predict the current output of the thermal power unit. This allows for the use of network output to conduct primary frequency regulation tests with frequency differences under different load conditions, calculating the unit's output power during this process, and thus achieving online calculation of evaluation factors. Figure 3 As shown, the primary frequency regulation output prediction model for thermal power units is a BiLSTM-Attention structure, including a BiLSTM layer, an Attention layer, a fully connected layer, and an input layer. Used to receive the unit output feature set after dimensionality reduction by UMAP, representing the input sequence at different times; the bidirectional LSTM layer (i.e., BiLSTM layer) includes forward LSTM and backward LSTM, the forward LSTM ( Extracting temporal dependencies from left to right, then using a backward LSTM ( Capture reverse associations from right to left to generate forward hidden states. and backward hidden state The Attention layer is used to calculate the weights at each time step. Focusing on the key stages of the frequency modulation process, this avoids information overload caused by equal weighting. Fully connected layers integrate attention-weighted features, outputting the predicted output value of the unit during primary frequency modulation, providing accurate dynamic data support for subsequent evaluation factor calculations. BiLSTM can encode information sequentially from back to front, exhibiting stronger ability to capture bidirectional dependencies and effectively extracting information from time-series data. The attention mechanism focuses limited attention on interpreting key information, thus acquiring information more efficiently. Combining Attention with BiLSTM networks allows the network to capture useful information more efficiently, improving network performance.

[0063] 204. Identify multiple thermal power units to be evaluated.

[0064] In the embodiments of this application, horizontal comparisons between generating units can be performed, such as the power grid dispatching simultaneously evaluating 10 generating units in the region to quickly select the optimal generating unit with fast frequency regulation response, low economic cost, and large safety margin; vertical analysis of single generating units under multiple operating conditions can also be performed, such as power plants analyzing the differences in frequency regulation capabilities of the same generating unit under high / low load and different coal qualities. This can adapt to the complex operating scenarios of multi-unit coordinated frequency regulation and dynamic switching of multiple operating conditions in the new power system, breaking the limitations of traditional single-unit and single-operating-condition evaluation, allowing the power grid side to achieve real-time dispatch-level multi-unit capability ranking, and the power plant side to carry out full-operating-condition domain operation strategy optimization.

[0065] 205. Using the unit output prediction results and high-dimensional operation dataset, the evaluation dimensions of each object to be evaluated are quantitatively calculated to obtain the response time evaluation factor, stability time evaluation factor, frequency regulation contribution rate evaluation factor, load regulation rate evaluation factor, fuel cost change evaluation factor, heat consumption change evaluation factor, and unit power margin evaluation factor for each object to be evaluated.

[0066] In this embodiment, the primary frequency regulation capability of coal-fired power units is evaluated from three aspects: primary frequency regulation performance, primary frequency regulation economy, and primary frequency regulation safety limits. The evaluation of primary frequency regulation performance mainly refers to the national standard "Guidelines for Primary Frequency Regulation Test and Performance Acceptance of Thermal Power Generating Units" (GB / T40595—2021) and the dynamic performance indicators and requirements for primary frequency regulation tests in the "two detailed rules", selecting the response time, settling time, primary frequency regulation contribution rate, and load regulation rate of primary frequency regulation; the primary frequency regulation economy is selected based on changes in fuel cost and heat consumption; and the primary frequency regulation safety limits are selected based on the unit's power margin.

[0067] Specifically, for each object to be evaluated, the unit output prediction curve of the object to be evaluated is extracted from the unit output prediction results. The disturbance response time is identified in the unit output prediction curve. The response time evaluation factor of the object to be evaluated is calculated using the disturbance response, as shown in Formula 5 below:

[0068] Formula 5:

[0069] in, This represents the response time evaluation factor. This indicates the disturbance response time. Response time refers to the time between a disturbance occurring at the system input and the corresponding response being generated at the output. 'a' represents the preset acceptable response time, which is generally 3 seconds.

[0070] Identify the settling time in the unit output prediction curve, and use the settling time to calculate the settling time evaluation factor of the object to be evaluated, as shown in Formula 6 below:

[0071] Formula 6:

[0072] in, Indicates the evaluation factor for stable time. This refers to the settling time, which is the time during which the system response reaches and remains at the system's final value. The minimum time required within the specified range, where b represents the preset qualified stabilization time, which is generally 45 seconds.

[0073] The actual output power of the thermal power unit during the primary frequency regulation process is determined from the unit output prediction curve. The ratio of the actual output power to the theoretical contribution power during the primary frequency regulation process is taken as the primary frequency regulation contribution rate. The frequency regulation contribution rate evaluation factor of the object to be evaluated is determined based on the primary frequency regulation contribution rate, as shown in Formula 7 below:

[0074] Formula 7:

[0075] in, This represents the evaluation factor for frequency modulation contribution rate. This represents the primary frequency modulation contribution rate, typically 1. above.

[0076] Extract the initial power of the object to be evaluated from the high-dimensional operating dataset, extract the maximum and minimum power values ​​from the unit output prediction curve, calculate the load regulation rate using the initial power, maximum power, and minimum power, and determine the load regulation rate evaluation factor of the object to be evaluated based on the load regulation rate, as shown in Formula 8 below:

[0077] Formula 8:

[0078]

[0079] in, This represents the load regulation rate evaluation factor. Indicates the load regulation rate. Indicates the initial power. Indicates the maximum power. This indicates the minimum power value.

[0080] Extract the real-time coal consumption rate, real-time output power, real-time efficiency, real-time coal consumption, and frequency deviation of the unit to be evaluated from the high-dimensional operating dataset. Determine the evaluation factor for the change in fuel cost of the unit to be evaluated based on the real-time coal consumption rate, real-time output power, real-time efficiency, real-time coal consumption, and frequency deviation, as shown in Formula 9 below:

[0081] Formula 9:

[0082]

[0083]

[0084]

[0085]

[0086]

[0087]

[0088] in, This represents the evaluation factor for changes in fuel costs. Indicates the unit's fuel cost, This indicates the price of coal for fuel. This indicates the unit's real-time coal consumption rate. This indicates the real-time output power of the generator set. This indicates the real-time efficiency of the generator set. This indicates the unit's real-time coal consumption. This indicates the mechanical efficiency of the steam turbine. Indicates generator efficiency. This indicates the lower heating value of coal, which is generally taken as 19625 kJ / kg. Indicates the start time of a frequency modulation. Indicates the end time of one frequency modulation. This represents the fuel cost without any frequency regulation operation. This indicates the unit load at the start of frequency regulation. This represents the coal consumption cost under ideal conditions for a single frequency regulation load output. This indicates the unit load at the end of frequency regulation. Represents the normalized function. Indicates frequency difference, This indicates changes in fuel costs, which refer to the difference in fuel expenses between a single frequency regulation operation and a stable operating condition.

[0089] Extract the main steam enthalpy, steam flow rate, boiler feedwater enthalpy, boiler feedwater flow rate, superheated desuperheating water enthalpy, superheated desuperheating water flow rate, overall reheat section enthalpy, overall reheat section flow rate, multiple furnace-side single-stream exhaust steam enthalpy, multiple furnace-side single-stream exhaust steam flow rate, and frequency difference from the high-dimensional operating dataset. Based on these parameters, determine the evaluation factor for the change in heat consumption of the object under evaluation, as shown in Formula 10 below.

[0090] Formula 10:

[0091]

[0092]

[0093] in, The evaluation factor represents the change in heat consumption. This represents the calculated value of heat consumption. This indicates the enthalpy value of the main steam. Indicates steam flow rate. This indicates the enthalpy value of the boiler feedwater. Indicates the boiler feedwater flow rate. This indicates the enthalpy value of the superheated desuperheated water. Indicates the flow rate of the superheated desuperheating water. This indicates the overall enthalpy value of the reheat section. This indicates the overall flow rate in the reheat section. This represents the enthalpy value of the single exhaust steam on the i-th furnace side. This represents the single-stream exhaust steam flow rate on the i-th furnace side. Represents the normalized function. This represents the heat consumption under ideal conditions with a single frequency regulation load output. This represents the heat dissipation when no frequency modulation operation occurs. Indicates frequency difference, This indicates the change in heat consumption, which refers to the difference in heat obtained by the turbine unit from the external heat source under two conditions: a frequency regulation operation and a stable operating condition.

[0094] The power margin evaluation factor of the unit to be evaluated is calculated using the minimum power value and the initial power value, as shown in Formula 11 below:

[0095] Formula 11:

[0096]

[0097] in, This represents the power margin evaluation factor for the generating unit. This indicates the unit's power margin, which refers to the difference between the minimum power output during the first frequency regulation process and the unit's minimum safe output. This represents the minimum power. This represents the minimum safe output calculation factor for the generator unit, typically 1. , Indicates the rated power of the unit. Indicates the initial power.

[0098] 206. The evaluation factors include the response time evaluation factor, stability time evaluation factor, frequency regulation contribution rate evaluation factor, load regulation rate evaluation factor, fuel cost change evaluation factor, heat consumption change evaluation factor, and unit power margin evaluation factor of multiple objects to be evaluated.

[0099] In this embodiment, the evaluation factor set covers performance, economy, and safety, avoiding the problem of traditional evaluations that only consider response speed and overlook issues such as sudden increases in coal consumption or exceeding power limits. This makes the evaluation results more aligned with the multi-objective requirements of safe, economical, and efficient frequency regulation, providing a standardized input matrix for the subsequent entropy-weighted TOPSIS algorithm, and also enabling a systematic comparison between units. Compared with other primary frequency regulation capability evaluation methods, this approach not only considers the performance of primary frequency regulation but also provides a quantitative evaluation of its safety and economy.

[0100] 207. The entropy-weighted TOPSIS method is used to assign weights to the evaluation factor set.

[0101] In this embodiment, to form a comprehensive evaluation index for the primary frequency regulation capability of coal-fired power units, the constructed evaluation factors need to be shaped into a comprehensive quantitative score. Considering the distinction of each indicator on the primary frequency regulation capability, the entropy-weighted TOPSIS method is used to calculate the weights of each evaluation factor. Based on the different uncertainties of each indicator, different weights can be assigned to each indicator, so that the final comprehensive score can reflect the influence of different factors on the evaluation result.

[0102] Specifically, multiple evaluation objects of the thermal power units to be evaluated are identified. Multiple evaluation factors for each evaluation object are extracted from the evaluation factor set. The multiple evaluation factors for each evaluation object are standardized to obtain an evaluation matrix. The entropy value of each evaluation factor type is calculated using the evaluation matrix, as shown in Formula 12 below:

[0103] Formula 12:

[0104]

[0105] in, This represents the entropy value of the j-th evaluation factor type. represents the evaluation factor of the j-th evaluation factor type of the i-th object to be evaluated in the evaluation matrix, n represents the number of objects to be evaluated, and k represents the normalization coefficient.

[0106] Next, the weight of each evaluation factor type is calculated using the entropy value of each evaluation factor type, as shown in Formula 13 below:

[0107] Formula 13:

[0108] in, This represents the weight of the j-th evaluation factor type. Let q represent the entropy value of the j-th evaluation factor type, and let q represent the number of evaluation factor types, which is 7 in this embodiment.

[0109] Then, a weighted standardized matrix is ​​constructed using the evaluation matrix and the weight of each evaluation factor type, as shown in Formula 14 below:

[0110] Formula 14:

[0111]

[0112] in, This represents the normalized value of the j-th evaluation factor type for the i-th object to be evaluated. This represents the evaluation factor of the j-th evaluation factor type for the i-th object to be evaluated in the evaluation matrix, where n represents the number of objects to be evaluated. This represents the value of the j-th evaluation factor type for the i-th object to be evaluated in the weighted standardized matrix. This represents the weight of the j-th evaluation factor type.

[0113] Then, based on the weighted standardization matrix, determine the positive and negative ideal solutions of the evaluation indicators for each object to be evaluated, as shown in Formula 15 below:

[0114] Formula 15:

[0115] in, This represents the positive ideal solution for the evaluation index of the i-th object to be evaluated. This represents the positive ideal of the first evaluation factor type for the i-th object to be evaluated. This represents the positive ideal of the second evaluation factor type for the i-th object to be evaluated. This represents the positive ideal of the q-th evaluation factor type for the i-th object to be evaluated. Let represent the negative ideal solution of the evaluation index for the i-th object to be evaluated. This represents the negative ideal of the first evaluation factor type for the i-th object to be evaluated. This represents the negative ideal of the second evaluation factor type for the i-th object to be evaluated. This represents the negative ideal of the i-th object to be evaluated and the q-th evaluation factor type. The positive ideal solution is the optimal value, and the negative ideal solution is the worst value.

[0116] Then, the target weight for each evaluation factor type is calculated using the weighted standardization matrix, the positive ideal solution of the evaluation index for each object to be evaluated, and the negative ideal solution of the evaluation index, as shown in Formula 16 below:

[0117] Formula 16:

[0118]

[0119]

[0120] in, This represents the target weight for the i-th evaluation factor type. The closer the value is to 1, the higher the weight. This represents the Euclidean distance between the type of the i-th evaluation factor and the positive ideal solution. Let represent the Euclidean distance between the type of the i-th evaluation factor and the negative ideal solution. This represents the value of the j-th evaluation factor type for the i-th object to be evaluated in the weighted standardized matrix. This represents the positive ideal solution for the evaluation index of the i-th object to be evaluated. Let represent the negative ideal of the first evaluation factor type for the i-th object to be evaluated, and n represent the number of objects to be evaluated. Finally, the target weights of multiple evaluation factor types are used as the weight allocation result.

[0121] 208. Based on the weight allocation results, perform a weighted average calculation on the set of evaluation factors to generate a comprehensive evaluation index of the primary frequency regulation capability of the thermal power unit to be evaluated.

[0122] In this embodiment of the application, multiple evaluation objects of the thermal power unit to be evaluated are determined, multiple evaluation factors for each evaluation object are extracted from the evaluation factor set, and a weighted average is calculated for the multiple evaluation factors of each evaluation object based on the weight allocation result to obtain the comprehensive evaluation index of each evaluation object, as shown in the following formula 17:

[0123] Formula 17:

[0124] in, This represents the comprehensive evaluation index for the i-th object to be evaluated. This represents the evaluation factor of the j-th evaluation factor type for the i-th object to be evaluated. Let represent the target weight of the i-th evaluation factor type, and q represent the number of evaluation factor types. Then, the comprehensive evaluation index of multiple objects to be evaluated is used as the comprehensive evaluation index of the primary frequency regulation capability of the thermal power unit to be evaluated.

[0125] In an optional implementation scheme, when multiple objects to be evaluated are multiple generating units, the comprehensive evaluation index of multiple objects to be evaluated is calculated by weighted average to obtain the comprehensive evaluation index of the thermal power unit to be evaluated, and the comprehensive evaluation index of the thermal power unit to be evaluated is used as the comprehensive evaluation index of the primary frequency regulation capability of the thermal power unit to be evaluated.

[0126] Based on the above process, the flowchart of the online evaluation method for the primary frequency regulation capability of thermal power units proposed in this application is as follows:

[0127] like Figure 4 As shown, firstly, historical operating data and online operating data of the units are collected synchronously; secondly, the historical data is reduced and purified using the UMAP algorithm to retain the manifold structure characteristics of the boiler-turbine coupling, and then input into the BiLSTM-Attention network to train a data-driven primary frequency regulation process output calculation model. In the online stage, the online unit operating data is used to output the unit output prediction results of the primary frequency regulation process; then, the primary frequency regulation safety, primary frequency regulation performance, and primary frequency regulation economy are quantified into evaluation factors; finally, the entropy weight TOPSIS method is used to first dynamically assign weights based on the factor dispersion, then expose hidden shortcomings by comparing positive and negative ideal solutions, and finally generate a weighted aggregation to generate a comprehensive evaluation result at the operating condition level, supporting the safe, economical, and efficient frequency regulation of thermal power units.

[0128] This application provides an online evaluation method for the primary frequency regulation capability of thermal power units. Compared with existing technologies, this application obtains real-time operating data of the thermal power unit to be evaluated, preprocesses the real-time operating data to obtain a high-dimensional operating dataset. Then, the UMAP algorithm is used to perform feature dimensionality reduction and reconstruction on the high-dimensional operating dataset to obtain a unit output feature set. Unlike the linear dimensionality reduction limitations of the traditional PCA algorithm, the UMAP algorithm can capture the manifold structure of the data while preserving local correlations and global distribution, significantly reducing the dimensionality of the features and improving the accuracy of the model prediction results. Subsequently, the unit output feature set is input into a pre-trained primary frequency regulation output prediction model for thermal power units to obtain the unit output prediction results, achieving online real-time frequency regulation output prediction. In terms of evaluation dimensions, this method breaks through the traditional framework that only focuses on frequency regulation performance. Based on the unit output prediction results and the high-dimensional operating dataset, the evaluation dimensions of the thermal power unit to be evaluated are quantitatively calculated, constructing a set of evaluation factors from three dimensions: primary frequency regulation performance, safety, and economy. Finally, the entropy-weighted TOPSIS method is used to assign weights to the evaluation factor set. Based on the weight assignment results, a weighted average is calculated on the evaluation factor set to generate a comprehensive evaluation index for the primary frequency regulation capability of the thermal power unit to be evaluated. The entropy-weighted TOPSIS method can achieve objective weighting and quantify the gap between each evaluation factor and the optimal state using positive and negative ideal solutions. The final output of the comprehensive evaluation index can intuitively reflect the quality of the unit's frequency regulation capability and improve the accuracy of frequency regulation capability evaluation.

[0129] Furthermore, as Figure 1 To specifically implement the method, this application provides an online evaluation device for the primary frequency regulation capability of thermal power units, such as... Figure 5 As shown, the device includes: a preprocessing module 301, an output prediction module 302, a factor calculation module 303, and an evaluation module 304.

[0130] Preprocessing module 301 is used to acquire real-time operating data of the thermal power unit to be evaluated, and preprocess the real-time operating data to obtain a high-dimensional operating dataset.

[0131] The output prediction module 302 is used to perform feature dimensionality reduction and reconstruction processing on the high-dimensional operating dataset using the UMAP algorithm to obtain the unit output feature set, and input the unit output feature set into the pre-trained thermal power unit primary frequency regulation output prediction model to obtain the unit output prediction result.

[0132] The factor calculation module 303 is used to perform quantitative calculation of the evaluation dimensions of the thermal power unit to be evaluated based on the unit output prediction results and the high-dimensional operation dataset, and obtain a set of evaluation factors.

[0133] The evaluation module 304 is used to assign weights to the set of evaluation factors using the entropy weight TOPSIS method, and to perform a weighted average calculation on the set of evaluation factors based on the weight assignment results, thereby generating a comprehensive evaluation index of the primary frequency regulation capability of the thermal power unit to be evaluated.

[0134] In specific application scenarios, the output prediction module 302 is used to take the data points in the high-dimensional running dataset as nodes to obtain multiple nodes; and calculate the local distance and length parameters of each node.

[0135]

[0136]

[0137] in, Represents a node Local distance, Represents a node The length parameter, Represents nodes The node with the smallest Euclidean distance. Represents a node With nodes The Euclidean distance is given, where n represents the number of nodes; a high-dimensional directed weighted graph is constructed using the multiple nodes, as well as the local distance and length parameters of each node; and a high-dimensional undirected weighted graph is constructed using the high-dimensional directed weighted graph.

[0138]

[0139] in, This represents a high-dimensional undirected weighted graph. This represents a high-dimensional directed weighted graph. This represents the Hadamard product; a high-dimensional undirected weighted matrix is ​​constructed using the high-dimensional undirected weighted graph; the high-dimensional undirected weighted matrix is ​​mapped to a low-dimensional space to obtain the low-dimensional space coordinates of each node; and a low-dimensional equivalent weighted graph is constructed using the low-dimensional space coordinates of the multiple nodes.

[0140]

[0141] in, Represents the nodes in the low-dimensional equivalent weighted graph H With nodes Association weights Represents a node Low-dimensional space coordinates, Represents a node The low-dimensional spatial coordinates are given, where 'a' represents the first hyperparameter and 'b' represents the second hyperparameter. A cross-entropy topological difference metric function is constructed between the high-dimensional undirected weighted graph and the low-dimensional equivalent weighted graph using the cross-entropy metric method.

[0142]

[0143] in, Let represent the cross-entropy topological difference measure function between a high-dimensional undirected weighted graph G and a low-dimensional equivalent weighted graph H. Represents nodes in a high-dimensional undirected weighted graph With nodes Association weights Represents nodes in a low-dimensional equivalent weighted graph With nodes The association weights are determined; the cross-entropy topological difference metric function is optimized using the stochastic gradient descent algorithm to obtain a low-dimensional feature set, which is then used as the unit output feature set; the unit output feature set is input into a pre-trained thermal power unit primary frequency regulation output prediction model to obtain the unit output prediction result, wherein the thermal power unit primary frequency regulation output prediction model includes a BiLSTM layer, an Attention layer, and a fully connected layer.

[0144] In a specific application scenario, the factor calculation module 303 is used to determine multiple evaluation objects of the thermal power unit to be evaluated; the unit output prediction results and the high-dimensional operation dataset are used to perform quantitative calculation of the evaluation dimensions of each evaluation object to obtain the response time evaluation factor, stability time evaluation factor, frequency regulation contribution rate evaluation factor, load regulation rate evaluation factor, fuel cost change evaluation factor, heat consumption change evaluation factor, and unit power margin evaluation factor for each evaluation object; the response time evaluation factor, stability time evaluation factor, frequency regulation contribution rate evaluation factor, load regulation rate evaluation factor, fuel cost change evaluation factor, heat consumption change evaluation factor, and unit power margin evaluation factor of the multiple evaluation objects are used as the evaluation factor set.

[0145] In specific application scenarios, the factor calculation module 303 is used to extract the unit output prediction curve of each object to be evaluated from the unit output prediction results, identify the disturbance response time in the unit output prediction curve, and calculate the response time evaluation factor of the object to be evaluated using the disturbance response.

[0146]

[0147] in, This represents the response time evaluation factor. The disturbance response time is represented by 'a', where 'a' represents the preset acceptable response time. The settling time is identified in the unit output prediction curve, and this settling time is used to calculate the settling time evaluation factor for the object to be evaluated.

[0148]

[0149] in, Indicates the evaluation factor for stable time. 'b' represents the stabilization time, and 'b' represents the preset qualified stabilization time. The actual output power of the thermal power unit during the primary frequency regulation process is determined from the unit output prediction curve. The ratio of the actual output power to the theoretical contribution power during the primary frequency regulation process is used as the primary frequency regulation contribution rate. Based on the primary frequency regulation contribution rate, the frequency regulation contribution rate evaluation factor of the object to be evaluated is determined.

[0150]

[0151] in, This represents the evaluation factor for frequency modulation contribution rate. This represents the primary frequency regulation contribution rate; the initial power of the object to be evaluated is extracted from the high-dimensional operating dataset, and the maximum and minimum power values ​​are extracted from the unit output prediction curve. The load regulation rate is calculated using the initial power, the maximum power, and the minimum power, and the load regulation rate evaluation factor of the object to be evaluated is determined based on the load regulation rate.

[0152]

[0153]

[0154] in, This represents the load regulation rate evaluation factor. Indicates the load regulation rate. Indicates the initial power. Indicates the maximum power. This represents the minimum power value. The real-time coal consumption rate, real-time output power, real-time efficiency, real-time coal consumption, and frequency difference of the unit to be evaluated are extracted from the high-dimensional operating dataset. Based on these parameters, the fuel cost change evaluation factor for the unit to be evaluated is determined.

[0155]

[0156]

[0157]

[0158]

[0159]

[0160]

[0161] in, This represents the evaluation factor for changes in fuel costs. Indicates the unit's fuel cost, This indicates the price of coal for fuel. This indicates the unit's real-time coal consumption rate. This indicates the real-time output power of the generator set. This indicates the real-time efficiency of the generator set. This indicates the unit's real-time coal consumption. This indicates the mechanical efficiency of the steam turbine. Indicates generator efficiency. This indicates the lower heating value of coal. Indicates the start time of a frequency modulation. Indicates the end time of one frequency modulation. This represents the fuel cost without any frequency regulation operation. This indicates the unit load at the start of frequency regulation. This represents the coal consumption cost under ideal conditions for a single frequency regulation load output. This indicates the unit load at the end of frequency regulation. Represents the normalized function. The frequency difference is used to represent the main steam enthalpy, steam flow rate, boiler feedwater enthalpy, boiler feedwater flow rate, superheated desuperheating water enthalpy, superheated desuperheating water flow rate, overall reheat section enthalpy, overall reheat section flow rate, multiple furnace-side single-stream exhaust steam enthalpy, multiple furnace-side single-stream exhaust steam flow rate, and frequency difference of the object to be evaluated. Based on the main steam enthalpy, steam flow rate, boiler feedwater enthalpy, boiler feedwater flow rate, superheated desuperheating water enthalpy, superheated desuperheating water flow rate, overall reheat section enthalpy, overall reheat section flow rate, multiple furnace-side single-stream exhaust steam enthalpy, multiple furnace-side single-stream exhaust steam flow rate, and frequency difference, the evaluation factor for the change in heat consumption of the object to be evaluated is determined.

[0162]

[0163]

[0164] in, The evaluation factor represents the change in heat consumption. This represents the calculated value of heat consumption. This indicates the enthalpy value of the main steam. Indicates steam flow rate. This indicates the enthalpy value of the boiler feedwater. Indicates the boiler feedwater flow rate. This indicates the enthalpy value of the superheated desuperheated water. Indicates the flow rate of the superheated desuperheating water. This indicates the overall enthalpy value of the reheat section. This indicates the overall flow rate in the reheat section. This represents the enthalpy value of the single exhaust steam on the i-th furnace side. This represents the single-stream exhaust steam flow rate on the i-th furnace side. Represents the normalized function. This represents the heat consumption under ideal conditions with a single frequency regulation load output. This represents the heat dissipation when no frequency modulation operation occurs. The frequency difference is represented; the minimum power value and the initial power are used to calculate the unit power margin evaluation factor of the object to be evaluated.

[0165]

[0166]

[0167] in, This represents the power margin evaluation factor for the generating unit. This indicates the power margin of the generating unit. This represents the minimum power. This represents the calculation coefficient for the minimum safe output of the generator unit. Indicates the rated power of the unit. Indicates the initial power.

[0168] In a specific application scenario, the evaluation module 304 is used to identify multiple evaluation objects of the thermal power unit to be evaluated, extract multiple evaluation factors for each evaluation object from the set of evaluation factors, standardize the multiple evaluation factors of each evaluation object to obtain an evaluation matrix, and use the evaluation matrix to calculate the entropy value of each evaluation factor type.

[0169]

[0170]

[0171] in, This represents the entropy value of the j-th evaluation factor type. Let represent the evaluation factor of the j-th evaluation factor type for the i-th object to be evaluated in the evaluation matrix, where n represents the number of objects to be evaluated and k represents the normalization coefficient; the weight of each evaluation factor type is calculated using the entropy value of each evaluation factor type.

[0172]

[0173] in, This represents the weight of the j-th evaluation factor type. Let represent the entropy value of the j-th evaluation factor type, and q represent the number of evaluation factor types; a weighted standardized matrix is ​​constructed using the evaluation matrix and the weight of each evaluation factor type.

[0174]

[0175]

[0176] in, This represents the normalized value of the j-th evaluation factor type for the i-th object to be evaluated. This represents the evaluation factor of the j-th evaluation factor type for the i-th object to be evaluated in the evaluation matrix, where n represents the number of objects to be evaluated. This represents the value of the j-th evaluation factor type for the i-th object to be evaluated in the weighted standardized matrix. This represents the weight of the j-th evaluation factor type; based on the weighted standardization matrix, the positive and negative ideal solutions of the evaluation indicators for each object to be evaluated are determined.

[0177]

[0178] in, This represents the positive ideal solution for the evaluation index of the i-th object to be evaluated. This represents the positive ideal of the first evaluation factor type for the i-th object to be evaluated. This represents the positive ideal of the second evaluation factor type for the i-th object to be evaluated. This represents the positive ideal of the q-th evaluation factor type for the i-th object to be evaluated. Let represent the negative ideal solution of the evaluation index for the i-th object to be evaluated. This represents the negative ideal of the first evaluation factor type for the i-th object to be evaluated. This represents the negative ideal of the second evaluation factor type for the i-th object to be evaluated. Let represent the negative ideal of the q-th evaluation factor type for the i-th object to be evaluated; calculate the target weight for each evaluation factor type using the weighted standardization matrix, the positive ideal solution of the evaluation index for each object to be evaluated, and the negative ideal solution of the evaluation index.

[0179]

[0180]

[0181]

[0182] in, This represents the target weight for the i-th evaluation factor type. This represents the Euclidean distance between the type of the i-th evaluation factor and the positive ideal solution. Let represent the Euclidean distance between the type of the i-th evaluation factor and the negative ideal solution. This represents the value of the j-th evaluation factor type for the i-th object to be evaluated in the weighted standardized matrix. This represents the positive ideal solution for the evaluation index of the i-th object to be evaluated. The negative ideal of the first evaluation factor type of the i-th object to be evaluated is represented by n, and the target weights of the multiple evaluation factor types are used as the weight allocation result.

[0183] In a specific application scenario, the evaluation module 304 is used to identify multiple evaluation objects of the thermal power unit to be evaluated, extract multiple evaluation factors for each evaluation object from the set of evaluation factors, and calculate a weighted average of the multiple evaluation factors for each evaluation object based on the weight allocation result to obtain a comprehensive evaluation index for each evaluation object.

[0184]

[0185] in, This represents the comprehensive evaluation index for the i-th object to be evaluated. This represents the evaluation factor of the j-th evaluation factor type for the i-th object to be evaluated. q represents the target weight of the i-th evaluation factor type, and q represents the number of evaluation factor types; the comprehensive evaluation index of the multiple objects to be evaluated is used as the comprehensive evaluation index of the primary frequency regulation capability of the thermal power unit to be evaluated.

[0186] In specific application scenarios, the evaluation module 304 is used to calculate the weighted average of the comprehensive evaluation indicators of the multiple evaluation objects when the multiple evaluation objects are multiple units, to obtain the comprehensive evaluation indicator of the thermal power unit to be evaluated, and to use the comprehensive evaluation indicator of the thermal power unit to be evaluated as the comprehensive evaluation indicator of the primary frequency regulation capability of the thermal power unit to be evaluated.

[0187] This application provides an apparatus that, compared to existing technologies, acquires real-time operating data of the thermal power unit to be evaluated, preprocesses the real-time operating data to obtain a high-dimensional operating dataset, and then uses the UMAP algorithm to perform feature dimensionality reduction and reconstruction on the high-dimensional operating dataset to obtain a unit output feature set. Unlike the linear dimensionality reduction limitations of the traditional PCA algorithm, the UMAP algorithm can capture the manifold structure of the data while preserving local correlations and global distribution, significantly reducing the dimensionality of the features and improving the accuracy of the model prediction results. Subsequently, the unit output feature set is input into a pre-trained thermal power unit primary frequency regulation output prediction model to obtain the unit output prediction results, realizing online real-time frequency regulation output prediction. In terms of evaluation dimensions, it breaks through the traditional framework of only focusing on frequency regulation performance. Based on the unit output prediction results and the high-dimensional operating dataset, it performs quantitative calculations of the evaluation dimensions of the thermal power unit to be evaluated, constructing a set of evaluation factors from three dimensions: primary frequency regulation performance, safety, and economy. Finally, the entropy-weighted TOPSIS method is used to assign weights to the evaluation factor set. Based on the weight assignment results, a weighted average is calculated on the evaluation factor set to generate a comprehensive evaluation index for the primary frequency regulation capability of the thermal power unit to be evaluated. The entropy-weighted TOPSIS method can achieve objective weighting and quantify the gap between each evaluation factor and the optimal state using positive and negative ideal solutions. The final output of the comprehensive evaluation index can intuitively reflect the quality of the unit's frequency regulation capability and improve the accuracy of frequency regulation capability evaluation.

[0188] It should be noted that other corresponding descriptions of the functional units involved in the online evaluation device for primary frequency regulation capability of thermal power units provided in this application embodiment can be found in the following references. Figure 1 and Figures 2 to 4 The corresponding descriptions in [the document] will not be repeated here.

[0189] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0190] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0191] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

[0192] In an exemplary embodiment, see Figure 6 Furthermore, a device is provided, comprising a bus, a processor, a memory, and a communication interface. It may also include an input / output interface and a display device, wherein the various functional units can communicate with each other via the bus. The memory stores a computer program, and the processor executes the program stored in the memory to perform the online evaluation method for the primary frequency regulation capability of thermal power units described in the above embodiments.

[0193] A medium storing a computer program thereon, which, when executed by a processor, implements the steps of the online evaluation method for the primary frequency regulation capability of thermal power units.

[0194] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented in hardware or by using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) and includes several instructions to cause a computer device (such as a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0195] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of a preferred embodiment, and the modules or processes shown in the drawings are not necessarily essential for implementing this application.

[0196] Those skilled in the art will understand that the modules in the apparatus of the implementation scenario can be distributed within the apparatus of the implementation scenario as described, or they can be located in one or more apparatuses different from this implementation scenario, with corresponding changes. The modules of the above-described implementation scenario can be combined into one module, or they can be further divided into multiple sub-modules.

[0197] The serial numbers in this application are for descriptive purposes only and do not represent the superiority or inferiority of the implementation scenario.

[0198] The above disclosures are only a few specific implementation scenarios of this application. However, this application is not limited to these. Any variations that can be conceived by those skilled in the art should fall within the protection scope of this application.

Claims

1. A method for online evaluation of the primary frequency regulation capability of a thermal power unit, characterized in that, include: The real-time operating data of the thermal power unit to be evaluated is obtained, and the real-time operating data is preprocessed to obtain a high-dimensional operating dataset; The UMAP algorithm is used to perform feature dimensionality reduction and reconstruction on the high-dimensional operating dataset to obtain the unit output feature set. This feature set is then input into a pre-trained primary frequency regulation output prediction model for thermal power units to obtain the unit output prediction result. This includes: treating data points in the high-dimensional operating dataset as nodes to obtain multiple nodes; and calculating the local distance and length parameters of each node. in, Represents a node Local distance, Represents a node The length parameter, Represents nodes The node with the smallest Euclidean distance. Represents a node With nodes The Euclidean distance is given, where n represents the number of nodes; a high-dimensional directed weighted graph is constructed using the multiple nodes, as well as the local distance and length parameters of each node; and a high-dimensional undirected weighted graph is constructed using the high-dimensional directed weighted graph. Where G represents a high-dimensional undirected weighted graph, This represents a high-dimensional directed weighted graph. This represents the Hadamard product; a high-dimensional undirected weighted matrix is ​​constructed using the high-dimensional undirected weighted graph; the high-dimensional undirected weighted matrix is ​​mapped to a low-dimensional space to obtain the low-dimensional space coordinates of each node; and a low-dimensional equivalent weighted graph is constructed using the low-dimensional space coordinates of the multiple nodes. in, Represents the nodes in the low-dimensional equivalent weighted graph H With nodes Association weights Represents a node Low-dimensional space coordinates, Represents a node The low-dimensional spatial coordinates are given, where 'a' represents the first hyperparameter and 'b' represents the second hyperparameter. A cross-entropy topological difference metric function is constructed between the high-dimensional undirected weighted graph and the low-dimensional equivalent weighted graph using the cross-entropy metric method. in, Let represent the cross-entropy topological difference measure function between a high-dimensional undirected weighted graph G and a low-dimensional equivalent weighted graph H. Represents nodes in a high-dimensional undirected weighted graph With nodes Association weights Represents nodes in a low-dimensional equivalent weighted graph With nodes The association weights are determined; the cross-entropy topological difference metric function is optimized using the stochastic gradient descent algorithm to obtain a low-dimensional feature set, which is then used as the unit output feature set; the unit output feature set is input into a pre-trained thermal power unit primary frequency regulation output prediction model to obtain the unit output prediction result, wherein the thermal power unit primary frequency regulation output prediction model includes a BiLSTM layer, an Attention layer, and a fully connected layer; Based on the unit output prediction results and the high-dimensional operation dataset, the evaluation dimensions of the thermal power unit to be evaluated are quantitatively calculated to obtain a set of evaluation factors. The set of evaluation factors includes a subset of primary frequency regulation performance evaluation dimensions, a subset of primary frequency regulation safety evaluation dimensions, and a subset of primary frequency regulation economic evaluation dimensions. The subset of primary frequency regulation performance evaluation dimensions includes response time evaluation factors, stability time evaluation factors, frequency regulation contribution rate evaluation factors, and load regulation rate evaluation factors. The subset of primary frequency regulation safety evaluation dimensions includes unit power margin evaluation factors. The subset of primary frequency regulation economic evaluation dimensions includes fuel cost change evaluation factors and heat consumption change evaluation factors. The TOPSIS method of entropy weight is used to assign weights to the set of evaluation factors. Based on the weight assignment results, a weighted average is calculated on the set of evaluation factors to generate a comprehensive evaluation index of the primary frequency regulation capability of the thermal power unit to be evaluated.

2. The method according to claim 1, characterized in that, The evaluation dimension quantification calculation is performed on the thermal power unit to be evaluated based on the unit output prediction results and the high-dimensional operation dataset to obtain a set of evaluation factors, including: Identify multiple objects to be evaluated for the thermal power unit to be evaluated; The unit output prediction results and the high-dimensional operation dataset are used to perform quantitative calculations on the evaluation dimensions of each of the objects to be evaluated, and the response time evaluation factor, stability time evaluation factor, frequency regulation contribution rate evaluation factor, load regulation rate evaluation factor, fuel cost change evaluation factor, heat consumption change evaluation factor, and unit power margin evaluation factor are obtained for each of the objects to be evaluated. The evaluation factors are defined as the response time evaluation factor, stability time evaluation factor, frequency regulation contribution rate evaluation factor, load regulation rate evaluation factor, fuel cost change evaluation factor, heat consumption change evaluation factor, and unit power margin evaluation factor of the multiple objects to be evaluated.

3. The method according to claim 2, characterized in that, The evaluation dimensions of each object to be evaluated are quantified using the unit output prediction results and the high-dimensional operation dataset, resulting in response time evaluation factors, stabilization time evaluation factors, frequency regulation contribution rate evaluation factors, load regulation rate evaluation factors, fuel cost change evaluation factors, heat consumption change evaluation factors, and unit power margin evaluation factors for each object to be evaluated, including: For each of the objects to be evaluated, the unit output prediction curve of the object to be evaluated is extracted from the unit output prediction results. The disturbance response time is identified in the unit output prediction curve, and the response time evaluation factor of the object to be evaluated is calculated using the disturbance response. in, This represents the response time evaluation factor. This indicates the disturbance response time, where 'a' represents the preset acceptable response time. The settling-off time is identified in the unit output prediction curve, and the settling-off time is used to calculate the settling-off time evaluation factor of the object to be evaluated. in, Indicates the evaluation factor for stable time. b represents the settling time, and b represents the preset qualified settling time; The actual output power of the thermal power unit during the primary frequency regulation process is determined from the unit output prediction curve. The ratio of the actual output power to the theoretical contribution power during the primary frequency regulation process is taken as the primary frequency regulation contribution rate. The frequency regulation contribution rate evaluation factor of the object to be evaluated is determined based on the primary frequency regulation contribution rate. in, This represents the evaluation factor for frequency modulation contribution rate. Indicates the contribution rate of primary frequency modulation; The initial power of the object to be evaluated is extracted from the high-dimensional operating dataset. The maximum and minimum power values ​​are extracted from the unit output prediction curve. The load regulation rate is calculated using the initial power, the maximum power, and the minimum power. The load regulation rate evaluation factor of the object to be evaluated is determined based on the load regulation rate. in, This represents the load regulation rate evaluation factor. Indicates the load regulation rate. Indicates the initial power. Indicates the maximum power. Indicates the minimum power value; Extract the real-time coal consumption rate, real-time output power, real-time efficiency, real-time coal consumption, and frequency deviation of the unit to be evaluated from the high-dimensional operating dataset. Determine the fuel cost change evaluation factor for the unit to be evaluated based on these parameters. in, This represents the evaluation factor for changes in fuel costs. Indicates the unit's fuel cost, This indicates the price of coal for fuel. This indicates the unit's real-time coal consumption rate. This indicates the real-time output power of the generator set. This indicates the real-time efficiency of the generator unit. This indicates the unit's real-time coal consumption. This indicates the mechanical efficiency of the steam turbine. Indicates generator efficiency. This indicates the lower heating value of coal. Indicates the start time of a frequency modulation. Indicates the end time of a frequency modulation. This represents the fuel cost without any frequency regulation operation. This indicates the unit load at the start of frequency regulation. This represents the coal consumption cost under ideal conditions for a single frequency regulation load output. This indicates the unit load at the end of frequency regulation. Represents the normalized function. Indicates frequency difference; The following parameters are extracted from the high-dimensional operational dataset: main steam enthalpy, steam flow rate, boiler feedwater enthalpy, boiler feedwater flow rate, superheated desuperheating water enthalpy, superheated desuperheating water flow rate, overall reheat section enthalpy, overall reheat section flow rate, multiple furnace-side single-stream exhaust steam enthalpy values, multiple furnace-side single-stream exhaust steam flow rates, and frequency difference. Based on these parameters, the evaluation factor for the change in heat consumption of the object under evaluation is determined. in, The evaluation factor represents the change in heat consumption. This represents the calculated value of heat consumption. Indicates the enthalpy of the main steam. Indicates steam flow rate. This indicates the enthalpy value of the boiler feedwater. Indicates the boiler feedwater flow rate. This indicates the enthalpy value of the superheated desuperheated water. Indicates the flow rate of the superheated desuperheating water. This indicates the overall enthalpy value of the reheat section. This indicates the overall flow rate in the reheat section. This represents the enthalpy value of the single exhaust steam on the i-th furnace side. This represents the single-stream exhaust steam flow rate on the i-th furnace side. Represents the normalized function. This represents the heat consumption under ideal conditions with a single frequency regulation load output. This represents the heat dissipation when no frequency modulation operation occurs. Indicates frequency difference; The power margin evaluation factor of the unit to be evaluated is calculated using the minimum power value and the initial power value. in, This represents the power margin evaluation factor for the generating unit. This indicates the power margin of the generating unit. This represents the minimum power. This represents the calculation coefficient for the minimum safe output of the generator unit. Indicates the rated power of the unit. Indicates the initial power.

4. The method according to claim 1, characterized in that, The step of assigning weights to the evaluation factor set using the entropy-weighted TOPSIS method includes: Identify multiple objects to be evaluated for the thermal power unit to be evaluated, and extract multiple evaluation factors for each object to be evaluated from the set of evaluation factors. The evaluation factors for each object to be evaluated are standardized to obtain an evaluation matrix. The entropy value for each evaluation factor type is then calculated using the evaluation matrix. in, This represents the entropy value of the j-th evaluation factor type. The evaluation factor represents the evaluation factor of the j-th evaluation factor type of the i-th object to be evaluated in the evaluation matrix, n represents the number of objects to be evaluated, and k represents the normalization coefficient. The weight of each evaluation factor type is calculated using the entropy value of each evaluation factor type. in, This represents the weight of the j-th evaluation factor type. q represents the entropy value of the j-th evaluation factor type, and q represents the number of evaluation factor types. A weighted standardized matrix is ​​constructed using the evaluation matrix and the weights of each evaluation factor type. in, This represents the normalized value of the j-th evaluation factor type for the i-th object to be evaluated. This represents the evaluation factor of the j-th evaluation factor type for the i-th object to be evaluated in the evaluation matrix, where n represents the number of objects to be evaluated. This represents the value of the j-th evaluation factor type for the i-th object to be evaluated in the weighted standardized matrix. This represents the weight of the j-th evaluation factor type; Based on the weighted standardization matrix, determine the positive ideal solution and the negative ideal solution of the evaluation index for each of the objects to be evaluated. in, This represents the positive ideal solution for the evaluation index of the i-th object to be evaluated. This represents the positive ideal of the first evaluation factor type for the i-th object to be evaluated. This represents the positive ideal of the second evaluation factor type for the i-th object to be evaluated. This represents the positive ideal of the q-th evaluation factor type for the i-th object to be evaluated. Let represent the negative ideal solution of the evaluation index for the i-th object to be evaluated. This represents the negative ideal of the first evaluation factor type for the i-th object to be evaluated. This represents the negative ideal of the second evaluation factor type for the i-th object to be evaluated. This represents the negative ideal of the q-th evaluation factor type for the i-th object to be evaluated; The target weight for each evaluation factor type is calculated using the weighted standardization matrix, the positive ideal solution of the evaluation index for each of the objects to be evaluated, and the negative ideal solution of the evaluation index. in, This represents the target weight for the i-th evaluation factor type. This represents the Euclidean distance between the type of the i-th evaluation factor and the positive ideal solution. Let represent the Euclidean distance between the type of the i-th evaluation factor and the negative ideal solution. This represents the value of the j-th evaluation factor type for the i-th object to be evaluated in the weighted standardized matrix. This represents the positive ideal solution for the evaluation index of the i-th object to be evaluated. represents the negative ideal of the first evaluation factor type of the i-th object to be evaluated, and n represents the number of objects to be evaluated; The target weights of the multiple evaluation factor types are used as the weight allocation results.

5. The method according to claim 1, characterized in that, The step of calculating a weighted average of the evaluation factor set based on the weight allocation results to generate a comprehensive evaluation index for the primary frequency regulation capability of the thermal power unit to be evaluated includes: Identify multiple objects to be evaluated for the thermal power unit to be evaluated, and extract multiple evaluation factors for each object to be evaluated from the set of evaluation factors. Based on the weight allocation results, a weighted average of multiple evaluation factors for each of the objects to be evaluated is calculated to obtain a comprehensive evaluation index for each object. in, This represents the comprehensive evaluation index for the i-th object to be evaluated. This represents the evaluation factor of the j-th evaluation factor type for the i-th object to be evaluated. represents the target weight of the i-th evaluation factor type, and q represents the number of evaluation factor types; The comprehensive evaluation index of the multiple objects to be evaluated shall be used as the comprehensive evaluation index of the primary frequency regulation capability of the thermal power unit to be evaluated.

6. The method according to claim 5, characterized in that, The method further includes: When the multiple objects to be evaluated are multiple units, the comprehensive evaluation index of the multiple objects to be evaluated is calculated by weighted average to obtain the comprehensive evaluation index of the thermal power unit to be evaluated, and the comprehensive evaluation index of the thermal power unit to be evaluated is used as the comprehensive evaluation index of the primary frequency regulation capability of the thermal power unit to be evaluated.

7. An online evaluation device for the primary frequency regulation capability of a thermal power unit, characterized in that, include: The preprocessing module is used to acquire real-time operating data of the thermal power unit to be evaluated, and to preprocess the real-time operating data to obtain a high-dimensional operating dataset. The power output prediction module is used to perform feature dimensionality reduction and reconstruction processing on the high-dimensional operating dataset using the UMAP algorithm to obtain a unit power output feature set. This unit power output feature set is then input into a pre-trained thermal power unit primary frequency regulation power output prediction model to obtain the unit power output prediction result. This includes: treating data points in the high-dimensional operating dataset as nodes to obtain multiple nodes; and calculating the local distance and length parameters of each node. in, Represents a node Local distance, Represents a node The length parameter, Represents nodes The node with the smallest Euclidean distance. Represents a node With nodes The Euclidean distance is given, where n represents the number of nodes; a high-dimensional directed weighted graph is constructed using the multiple nodes, as well as the local distance and length parameters of each node; and a high-dimensional undirected weighted graph is constructed using the high-dimensional directed weighted graph. Where G represents a high-dimensional undirected weighted graph, This represents a high-dimensional directed weighted graph. This represents the Hadamard product; a high-dimensional undirected weighted matrix is ​​constructed using the high-dimensional undirected weighted graph; the high-dimensional undirected weighted matrix is ​​mapped to a low-dimensional space to obtain the low-dimensional space coordinates of each node; and a low-dimensional equivalent weighted graph is constructed using the low-dimensional space coordinates of the multiple nodes. in, Represents the nodes in the low-dimensional equivalent weighted graph H With nodes Association weights Represents a node Low-dimensional space coordinates, Represents a node The low-dimensional spatial coordinates are given, where 'a' represents the first hyperparameter and 'b' represents the second hyperparameter. A cross-entropy topological difference metric function is constructed between the high-dimensional undirected weighted graph and the low-dimensional equivalent weighted graph using the cross-entropy metric method. in, Let represent the cross-entropy topological difference measure function between a high-dimensional undirected weighted graph G and a low-dimensional equivalent weighted graph H. Represents nodes in a high-dimensional undirected weighted graph With nodes Association weights Represents nodes in a low-dimensional equivalent weighted graph With nodes The association weights are determined; the cross-entropy topological difference metric function is optimized using the stochastic gradient descent algorithm to obtain a low-dimensional feature set, which is then used as the unit output feature set; the unit output feature set is input into a pre-trained thermal power unit primary frequency regulation output prediction model to obtain the unit output prediction result, wherein the thermal power unit primary frequency regulation output prediction model includes a BiLSTM layer, an Attention layer, and a fully connected layer; The factor calculation module is used to perform quantitative calculations of the evaluation dimensions of the thermal power unit to be evaluated based on the unit output prediction results and the high-dimensional operation dataset, to obtain a set of evaluation factors. The set of evaluation factors includes a subset of primary frequency regulation performance evaluation dimensions, a subset of primary frequency regulation safety evaluation dimensions, and a subset of primary frequency regulation economic evaluation dimensions. The subset of primary frequency regulation performance evaluation dimensions includes response time evaluation factors, stability time evaluation factors, frequency regulation contribution rate evaluation factors, and load regulation rate evaluation factors. The subset of primary frequency regulation safety evaluation dimensions includes unit power margin evaluation factors. The subset of primary frequency regulation economic evaluation dimensions includes fuel cost change evaluation factors and heat consumption change evaluation factors. The evaluation module is used to assign weights to the set of evaluation factors using the entropy-weighted TOPSIS method, and to calculate a weighted average of the set of evaluation factors based on the weight assignment results, thereby generating a comprehensive evaluation index of the primary frequency regulation capability of the thermal power unit to be evaluated.

8. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images