Power equipment purchase analysis method and system based on adaptation degree

Abstract

The invention belongs to the technical field of power equipment purchase, and provides a power equipment purchase analysis method and system based on the degree of adaptation, and the method comprises the steps: obtaining equipment parameters participating in bidding; obtaining an equipment adaptation degree according to the equipment parameters and a preset equipment adaptation degree evaluation model; all the adaptation degrees are arranged according to the sequence from high to low, the highest adaptation degree is a first-level tendency, and in the same way, purchasing scheme tendency levels of all the devices are obtained; a polynomial model is adopted as an equipment adaptation degree evaluation model, and a correction coefficient, a power index and an expression weight corresponding to each parameter in the polynomial model are determined through a recursive segmentation algorithm and an iterative optimization method, so that adaptation degree evaluation can be quickly carried out on various types of power equipment, and on the premise of ensuring accuracy and reliability, the equipment adaptation degree evaluation efficiency is improved. The limitation of dependence on artificial experience and complex mathematical modeling in a traditional method is avoided.

Description

Technical Field

[0001] The present invention belongs to the technical field of power equipment procurement, and particularly relates to a power equipment procurement analysis method and system based on fitness. Background Technique

[0002] With the development of data science and mining technology, it is a highly feasible and urgently explored technical route to use data-driven intelligent algorithms to evaluate the fitness of power equipment procurement, so as to determine the tendency of power equipment procurement plans. Currently, the methods for evaluating the fitness of new projects based on historical data of power equipment can be mainly divided into two categories. One is the traditional mathematical modeling method, which constructs a scoring calculation model of fitness by analyzing factors such as equipment parameters, performance indicators, operating environment, and manufacturing processes; the other is the deep learning method, which trains a deep neural network using historical equipment data to capture the non-linear mapping relationship between equipment parameters and fitness.

[0003] However, both the mathematical modeling method and the deep learning method have certain limitations. For example, the mathematical modeling method relies too much on the subjective experience of experts and is difficult to comprehensively cover the multi-dimensional parameter relationships of complex equipment. Therefore, for the fitness evaluation of large equipment, the technical route of traditional mathematical modeling is difficult to effectively implement; although the deep learning method can better mine the implicit relationship between equipment parameters and fitness, the neural network model itself lacks interpretability and cannot clearly show the basis for fitness evaluation. Therefore, it is difficult to gain the trust of the purchaser in practical engineering applications and cannot provide strong reference for power equipment procurement. Summary of the Invention

[0004] In order to solve the above problems, the present invention proposes a power equipment procurement analysis method and system based on fitness. The present invention uses a polynomial model as the equipment fitness evaluation model, and determines the correction coefficient, power exponent, and representation weight corresponding to each parameter in the polynomial model through a recursive partitioning algorithm and an iterative optimization method, which can quickly evaluate the fitness of various power equipment and avoid the limitations of relying on manual experience and complex mathematical modeling in the traditional method on the premise of ensuring accuracy and reliability.

[0005] In order to achieve the above object, the present invention is implemented through the following technical solutions:

[0006] In the first aspect, the present invention provides a power equipment procurement analysis method based on fitness, including:

[0007] Obtain the equipment parameters participating in the bid;

[0008] According to the equipment parameters and a preset equipment fitness evaluation model, obtain the equipment fitness;

[0009] Arrange all adaptation degrees in descending order. The highest adaptation degree is the first-level tendency, and so on to obtain the procurement plan tendency levels of all devices;

[0010] Among them, the device adaptation degree evaluation model is a polynomial model. Each parameter in the polynomial model corresponds to a correction coefficient, a power exponent, and a parameter importance index representing the weight; the correction coefficient and the power exponent are optimized and determined by an iterative optimization method; the parameter importance index is determined by a recursive partitioning algorithm. Specifically, the device parameters are hierarchically grouped, and according to the data of the device parameters and the adaptation degree scores, the partitioning threshold of each device parameter is calculated, and the device parameters that can reduce the adaptation degree scoring error are selected in each layer of grouping until the preset termination condition is met.

[0011] Further, the device parameters include the pressure, temperature, and material grade of the device; for different stages of power equipment procurement, the device parameters are classified and identified. The stages include the bidding stage, the tendering stage, and the procurement stage; in each stage, the data is labeled as different categories.

[0012] Further, generate a single-parameter description template, including the influence weight of each parameter on the adaptation degree, as well as the correction coefficient and the power exponent; generate a comprehensive adaptation degree description template, summarize the contributions of all parameters to the adaptation degree, and generate a complete description in combination with the predicted total adaptation degree.

[0013] Further, the adaptation degree suitability_evaluate is:

[0014] suitability_evaluate = ∑(feature_importance_n × a_n × value_n b _ n )

[0015] Among them, a_n is the device parameter - adaptation degree linear scale correction coefficient; b_n is the device parameter non-linear power exponent; value_n is the nth device parameter; feature_importance_n is the nth parameter importance index.

[0016] Further, based on a data-driven iterative optimization method, optimize the parameters a_n and b_n:

[0017] Randomly generate the initial parameter combinations a_n and b_n, and set the objective function to quantify the error between the predicted adaptation degree and the actual adaptation degree score;

[0018] According to the current parameter combination, calculate the error value through the objective function;

[0019] Update the current parameter combination, calculate a new parameter combination to reduce the objective function value;

[0020] Introduce random noise to the parameters or re-explore different initial solutions to expand the search space;

[0021] In each iteration, evaluate the fitness value of the current parameter combination, and retain the parameter combination with higher performance for the next iteration according to the set strategy;

[0022] Repeat the fitness evaluation and parameter adjustment process until the preset termination condition is met.

[0023] Furthermore, the objective function Loss is:

[0024] Loss = (1 / N)∑(suitability_true - suitability_evaluate) 2 + λ∑(a_n 2 + b_n 2 );

[0025] Where: suitability_true is the actual suitability; suitability_evaluate is the predicted suitability; N is the total number of samples; λ is the regularization coefficient.

[0026] Furthermore, sort the parameter values of different samples from smallest to largest; for continuous device parameters, take the midpoint of adjacent values as the candidate splitting point; for discrete parameters, enumerate all possible classification combinations; for each candidate splitting point, divide the dataset into two subsets; calculate the fitness score error of each subset respectively; use the splitting gain to measure the reduction of the error after splitting, and calculate the splitting gain; select the candidate point with the largest splitting gain as the splitting threshold of the current parameter; when determining the parameter that reduces the fitness score error, traverse all device parameters, for each parameter, calculate all splitting thresholds and the corresponding splitting gains; select the parameter with the largest splitting gain as the current grouping basis.

[0027] In a second aspect, the present invention also provides a power equipment procurement analysis system based on suitability, including:

[0028] A data acquisition module, configured to: obtain the device parameters participating in the tender;

[0029] A suitability determination module, configured to: obtain the device suitability according to the device parameters and a preset device suitability evaluation model;

[0030] A procurement plan preference determination module, configured to: arrange all suitability levels from high to low, the highest suitability is the first-level preference, and so on to obtain the procurement plan preference levels of all devices;

[0031] Among them, the device fitness evaluation model is a polynomial model. For each parameter in the polynomial model, there are a correction coefficient, a power exponent, and a parameter importance index representing the weight. The correction coefficient and the power exponent are determined by iterative optimization methods. The parameter importance index is determined by a recursive partitioning algorithm. Specifically, the device parameters are hierarchically grouped. According to the device parameters and the fitness score data, the splitting threshold of each device parameter is calculated. In each layer of grouping, the device parameters that can reduce the fitness score error are selected until the preset termination condition is met.

[0032] In a third aspect, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and capable of running on the processor. When the processor executes the program, the steps of the power device procurement analysis method based on fitness described in the first aspect are implemented.

[0033] In a fourth aspect, the present invention also provides a computer program product. The computer program product includes a computer program. When the computer program is executed by a processor, the steps of the power device procurement analysis method based on fitness described in the first aspect are implemented.

[0034] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0035] 1. The present invention uses a polynomial model as the device fitness evaluation model, and determines the correction coefficient, power exponent, and weight corresponding to each parameter in the polynomial model through a recursive partitioning algorithm and iterative optimization methods. It can quickly evaluate the fitness of various power devices, and on the premise of ensuring accuracy and reliability, it avoids the limitations of relying on manual experience and complex mathematical modeling in traditional methods. Then, according to the sorting order of fitness, the procurement plan preference levels of all devices are obtained, providing a strong reference for device procurement.

[0036] 2. In the process of fitness evaluation, the present invention combines the generation of the importance index of parameters and the details of parameter contributions. The non-linear relationship between the change of each parameter and the fitness is automatically analyzed and explained by the model, and users can intuitively understand the source and logic of the fitness evaluation, thereby enhancing the credibility and practical application value of the model.

[0037] 3. Traditional fitness evaluation description documents need to be manually written by a large amount of manpower. The present invention generates a description document including predicted fitness, details of parameter contributions, and explanations of key rules by automatically analyzing the model results and the rules of recursive partitioning of the feature space, supports multiple format outputs, reduces manual participation, and improves the efficiency of generating descriptive text content. Description of the Drawings

[0038] The accompanying drawings forming a part of this embodiment are used to provide a further understanding of this embodiment. The schematic embodiments and descriptions thereof of this embodiment are used to explain this embodiment and do not constitute an improper limitation of this embodiment.

[0039] Figure 1 It is the flowchart for evaluating the adaptability of Embodiment 1 of the present invention. Specific implementation manners

[0040] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0041] It should be noted that the following detailed descriptions are all exemplary and are intended to provide further explanations for this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which this application belongs.

[0042] Term explanations:

[0043] Procurement Suitability refers to the degree of matching between power equipment or services and requirements in the procurement scenario. Procurement Suitability comprehensively considers factors such as the technical parameters, performance indicators, price range, supply cycle, and historical usage records of the equipment, and is a score given to evaluate whether the equipment meets the specific needs of the purchaser. In the present invention, Procurement Suitability is used as the evaluation target, and the multi-dimensional characteristics of the equipment are quantified through data analysis and intelligent algorithms, so as to provide a scientific and reliable reference basis for procurement decisions.

[0044] Intelligent Algorithm is a set of algorithms that utilize data-driven technologies and optimization methods to simulate human intelligence. In the present invention, Intelligent Algorithm is used to analyze and process the parameter data of power equipment, and optimize the structure and parameters of the adaptability evaluation model, thereby improving the scientificity, accuracy, and interpretability of the evaluation results.

[0045] Suitability Explanation is an explanatory document generated based on the model prediction results, including the details of the impact of each parameter of the power equipment on the adaptability and the calculation process.

[0046] Interpretability means that the model output results can clearly show the functional relationship between the power equipment parameters and the adaptability, facilitating users to understand the logic and basis of the adaptability evaluation.

[0047] Embodiment 1:

[0048] How to accurately evaluate the suitability of power equipment procurement through equipment parameters, so as to determine the tendency of procurement plans, is a major challenge in the current procurement business of the power industry. Its difficulties are mainly reflected in the following aspects: Complexity of power equipment: Power equipment, especially large-scale power equipment, has complex structures and many parameters. It is difficult to comprehensively evaluate the suitability of equipment based on the experience of procurement engineers alone. Information asymmetry: Due to the existence of technical barriers, it is difficult to objectively evaluate the suitability and rationality of equipment from the perspective of equipment manufacturing process or performance matching in the bidding and procurement process. Subjective dependence: The process of suitability evaluation often relies on the subjective experience of procurement engineers, and when writing the suitability evaluation instructions, a lot of manpower is also required to write documents and list analysis methods. Insufficient data utilization: The historical equipment parameters and bidding and procurement data accumulated in the procurement business have not been fully utilized, resulting in the inability to provide effective data support for the current suitability evaluation. For example, various scoring items for equipment in the bid evaluation process, as very important prior knowledge, have not been fully utilized. By building a set of explainable intelligent algorithms, the above problems can be solved, and efficient and scientific evaluation of the adaptability of power equipment procurement can be achieved, providing a more reliable basis for procurement decisions.

[0049] As described in the background technology, the current methods for evaluating the suitability of new projects based on historical data of power equipment mainly include traditional mathematical modeling methods and deep learning methods; however, both mathematical modeling methods and deep learning methods have certain limitations. For example, mathematical modeling methods rely too much on the subjective experience of experts and it is difficult to fully cover the multi-dimensional parameter relationships of complex equipment. Therefore, for the fitness evaluation of large equipment, the technical route of traditional mathematical modeling is difficult to effectively implement; although the deep learning method can better explore the implicit relationship between equipment parameters and fitness, the neural network model itself lacks interpretability and cannot clearly display the basis for fitness evaluation. Therefore, it is difficult to gain the trust of the purchaser in actual engineering applications.

[0050] In order to solve at least one of the above problems, Figure 1 As shown, this embodiment provides a power equipment procurement analysis method based on adaptability, which belongs to a power equipment procurement plan tendency analysis method based on adaptability. For the bidding and procurement process of the power industry, it can scientifically evaluate the equipment procurement adaptability and generate an interpretable evaluation description, realize the power equipment procurement plan tendency analysis, and minimize the interference of subjective factors and engineer experience on the results, thereby improving the scientificity and efficiency of decision-making. The method includes:

[0051] S1. Equipment data collection:

[0052] Optionally, diverse data collection methods are designed according to different working stages and equipment types to ensure coverage of key parameters, operating environments, and historical procurement records of the equipment. The collected data is formatted and stored through a standardized process and stored in a data table or database in a structured manner to achieve unified management and convenient invocation.

[0053] S2. Data processing: Data cleaning. Optionally, through a systematic data cleaning process, ensure the accuracy and representativeness of the input data. The specific steps are as follows:

[0054] S2.1. Missing value handling: Optionally, for missing numerical parameters, use mean imputation, median filling, or regression prediction to complete; for categorical parameters, use mode filling or inference methods based on similar samples to complete.

[0055] S2.2. Outlier correction: Optionally, identify and correct data that exceeds a reasonable range or has an abnormal format, such as unreasonable parameter values or incorrect equipment type markings.

[0056] S2.3. Redundant data deduplication: Optionally, delete duplicate records by comparing them one by one to optimize the data structure.

[0057] S2.4. Outlier handling: Optionally, use box plot analysis, standard deviation method, or principal component analysis (PCA) to detect outlier samples and perform deletion or interpolation smoothing on the outliers.

[0058] S2.5. Data screening and integration: Optionally, screen representative data according to equipment type, working scenario, and adaptation requirements, and eliminate samples without reference value to ensure the balance between diversity and rationality of the data.

[0059] The cleaned data will be stored in the equipment parameter - adaptability sample database in a structured format to provide high-quality data support for subsequent model training and evaluation. The specific sample field templates include key fields such as equipment parameters, operating conditions, historical procurement information, and adaptability scores as follows:

[0060] {Equipment name: name,

[0061] Equipment parameters: {Parameter name 1: value_1, Parameter name 2: value_2,...},

[0062] Adaptability: suitability}.

[0063] S3. Data identification:

[0064] Optionally, for different stages of power equipment procurement, classify and label the data, including the bidding stage, tendering stage, procurement stage, etc. At each stage, the data is labeled with different categories to distinguish its application scenarios and ensure the pertinence and accuracy of subsequent analysis and evaluation.

[0065] S4. Calculation of equipment parameter importance index:

[0066] Optionally, for a certain type of equipment, in this embodiment, the cleaned equipment parameter - fitness sample data is used as the input training data of the model. Among them, the equipment parameter information is used as the input feature, and the fitness score of the equipment is used as the target label. The recursive partitioning of the feature space algorithm is used to train the sample data to generate an equipment fitness evaluation model. Specifically, the recursive partitioning of the feature space algorithm is used to determine the variable importance index in the fitness evaluation model. The equipment parameter information includes pressure, temperature, material grade, etc. Optionally, the equipment parameter information is the average ambient pressure of the equipment within a preset time in the operating environment or the working pressure inside the equipment, the average working temperature within the preset time, and the main material grade of the equipment, which can be obtained by collecting historical working data, etc.

[0067] In this embodiment, by means of recursive partitioning of the feature space, the equipment parameters are grouped, and a hierarchical fitness evaluation model is gradually constructed. Specifically, first, according to the data of equipment parameters (such as pressure, temperature, material grade, etc.) and fitness scores, calculate the segmentation threshold for each parameter to maximize the difference of the grouped data. Then, in each layer of grouping, select the parameter that can minimize the fitness score error to the greatest extent as the basis for the current grouping, and divide a new data subset accordingly. This process is recursively performed layer by layer until a preset termination condition is met, such as until the sample size of the data subset is too small, less than the preset amount, or the error after segmentation no longer decreases significantly.

[0068] Among them, the segmentation threshold is determined by evaluating the possible values of each parameter to determine the demarcation point that can minimize the fitness score error. Specifically:

[0069] Calculate candidate segmentation points: For each parameter (such as pressure, temperature, material grade), sort the parameter values of different samples from small to large. For continuous parameters, take the midpoint of adjacent values as the candidate segmentation point; for discrete parameters, enumerate all possible classification combinations.

[0070] Calculate the fitness score error after segmentation: For each candidate segmentation point, divide the data set into two subsets (left subset and right subset). Calculate the fitness score error of each subset respectively, such as the mean square error (MSE):

[0071] MSE = (1 / n) * Σ (predicted score - actual score)^2;

[0072] Among them, n is a constant.

[0073] Calculate the splitting gain Gain: Use the splitting gain to measure the reduction in error after splitting. The calculation formula is:

[0074] Gain = MSE_parent - (n_L / n) * MSE_L - (n_R / n) * MSE_R;

[0075] Among them, Gain is the splitting gain; MSE_parent is the mean squared error of the parent node, representing the error of the entire dataset before splitting. It is the mean squared error of the samples in the parent node (i.e., the node before splitting); n_L and n_R are the number of samples in the left child node (L) and the right child node (R) respectively. These two values represent the number of samples contained in each child node after splitting. n is the total number of all samples in the parent node, that is, the total number of samples before splitting; MSE_L and MSE_R are the mean squared errors of the left child node and the right child node respectively, which represent the errors of the data in the left child node and the right child node after splitting.

[0076] Select the best splitting point: Use an optimization algorithm (such as the greedy algorithm) to select the candidate point with the largest splitting gain as the splitting threshold for the current parameter.

[0077] When determining the parameters that reduce the fitness score error, in each layer of grouping, select the parameters that can minimize the fitness score error to the greatest extent. Specifically, traverse all parameters: for each parameter, calculate all possible splitting thresholds and the corresponding splitting gains; compare the splitting gains: for each parameter, record its best splitting point and the corresponding splitting gain. Among all parameters, select the parameter with the largest splitting gain as the basis for the current grouping; perform grouping: according to the selected parameter and splitting threshold, divide the dataset into left and right subsets, and recursively perform the next layer of grouping.

[0078] During the construction of the fitness evaluation model, by comparing the impact of parameter splitting on the fitness score, the variable importance index (parameter importance index) of each parameter is automatically calculated. These indexes are used to quantify the contribution degree of device parameters to the fitness evaluation result, so as to provide more interpretable results for subsequent analysis.

[0079] When calculating the variable importance index, the variable importance index quantifies the contribution degree of each parameter to the fitness evaluation result. Specifically, record the split gain: when selecting a parameter for splitting each time, calculate the reduction amount of the fitness score error caused by this split (i.e., the split gain), and accumulate the split gain of this parameter into the variable importance index. Normalization processing: perform normalization processing on the split gains of all parameters, and convert the importance index into a relative weight: importance index = sum of parameter split gains / sum of split gains of all parameters. Final index: the higher the variable importance index, the greater the contribution of the parameter to the reduction of the fitness score error, and thus the more important it is to the evaluation result.

[0080] It should be noted that the variable importance index comes from the accumulation of split gains during the splitting process, and the split gain directly comes from the process of dividing the data into new subsets. The split gain of the parameter is recorded each time of splitting. The split gains are accumulated into the variable importance index of the corresponding parameter. The division of subsets gradually refines the data distribution and dynamically calculates the importance of the parameters at the same time. The final variable importance index reflects the global contribution of the parameter in the entire splitting process by summarizing all split gains.

[0081] After training is completed, the generated fitness evaluation model is stored in an unstructured format for subsequent fitness evaluation tasks. At the same time, the calculated parameter importance index is used as a new field to update the device parameter - fitness sample database formed in the data cleaning step, and finally construct a device parameter - parameter importance index - fitness sample database to provide high-quality data support for further analysis and decision-making. The specific sample field templates include core field templates such as device type, key parameters, working scenarios, importance index, and fitness score:

[0082] {Device name: name,

[0083] Device parameters: {Parameter name 1: value_1, Parameter name 2: value_2,...},

[0084] Fitness: suitability,

[0085] Parameter importance index: {Parameter name 1: feature_importance_2, Parameter name 1: feature_importance_2,...}.

[0086] S5. Construction of interpretability model:

[0087] In order to facilitate the generation of explanatory content subsequently, a polynomial is used to model the device fitness model. Introduce the polynomial:

[0088] f 参数名n(value_n, a_n, b_n) = a_n × value_n b _ n ;

[0089] Among them, a_n is the linear scale correction coefficient of the device parameter - adaptability; b_n is the non - linear power exponent of the device parameter, and value_n is the nth device parameter. Thus, the polynomial model for device adaptability evaluation is the adaptability:

[0090] suitability_evaluate = ∑f 参数名n (value_n, a_n, b_n) = ∑a_n × value_n b _ n ;

[0091] In some other embodiments, the polynomial f 参数名n (value_n, a_n, b_n) = a_n × value_n b _ n can also be replaced by a spline function or a Fourier series.

[0092] Loss function construction: Based on the polynomial model suitability_evaluate = ∑f(value_n, a_n, b_n) of device adaptability evaluation, in this embodiment, a loss function is designed to measure the deviation between the predicted adaptability of the model and the actual adaptability, so as to optimize the undetermined parameters a_n and b_n of the model. The loss function is defined as:

[0093] Loss = (1 / N)∑(suitability_true - suitability_evaluate) 2 + λ∑(a_n 2 + b_n 2 );

[0094] Among them: suitability_true is the actual adaptability; suitability_evaluate is the adaptability predicted by the model; N is the total number of samples; λ is the regularization coefficient used to limit the model complexity; a_n and b_n are the parameters to be optimized in the polynomial model. This loss function consists of two parts: the first part is the mean square error, which is used to measure the deviation between the predicted value and the actual value of the model; the second part is the regularization term, which limits the magnitudes of the parameters a_n and b_n to prevent the model from overfitting. Optionally, the actual adaptability can be determined according to historical data; or set artificially; or obtained by calculation. For example, the actual adaptability = (1 - |budget price - bid price| / budget price) × 100%.

[0095] In this embodiment, an iterative optimization method based on data-driven is adopted to optimize and solve the undetermined parameters a_n and b_n in the device fitness evaluation model to achieve the optimal solution of the model. The specific steps are as follows:

[0096] S5.1. Parameter initialization: Randomly generate the initial parameter combinations a_n and b_n, and set the objective function (loss function) to quantify the error between the model prediction result and the actual fitness score. The form of the objective function is designed according to application requirements. For example, methods such as mean square error (MSE) or weighted error are adopted.

[0097] S5.2. Fitness evaluation: According to the current parameter combination, calculate the error value of the model through the objective function. The lower the error value, the higher the fitness of the current parameter combination. Fitness evaluation provides a basis for subsequent parameter update.

[0098] S5.3. Parameter adjustment: Use optimization strategies (such as gradient descent method, random search or other numerical optimization methods) to update the current parameter combination, calculate the new parameter combination to reduce the objective function value, thereby improving the model performance.

[0099] S5.4. Global search mechanism: To avoid the model falling into a local optimal solution, a perturbation mechanism or random sampling strategy is introduced during the optimization process. For example, introduce random noise to the parameters or re-explore different initial solutions to expand the search space and enhance the global search ability.

[0100] S5.5. Parameter combination evaluation and retention: In each iteration, evaluate the fitness value of the current parameter combination, and retain the parameter combination with higher performance for the next iteration according to the set strategy (such as selecting the optimal or sub-optimal solution).

[0101] S5.6. Termination condition detection: Repeat the fitness evaluation and parameter adjustment process until the preset termination condition is met. For example, the objective function value tends to be stable or reaches the set maximum number of iterations. The setting of the termination condition needs to balance the optimization accuracy and computational efficiency.

[0102] S5.7. Optimal parameter output: At the end of the optimization, select the parameter combinations a_n and b_n with the lowest final fitness value, use them as the optimal solution of the device fitness evaluation model, and apply them to the actual fitness evaluation task to provide high-precision support for subsequent procurement decisions.

[0103] Through the above steps, not only can the model parameters be efficiently optimized, but also the scientificity and reliability of the fitness evaluation can be improved to meet the actual engineering requirements. According to the parameter values finally determined by the above device fitness polynomial and optimization algorithm, the calculation formula for the device fitness is:

[0104] suitability_evaluate = ∑(feature_importance_n × f(value_n, a_n, b_n));

[0105] The linear and power effects of a single parameter on the device suitability are evaluated using a_n and b_n, while feature_importance_n linearly weights the contribution of each parameter to the device suitability, synthesizing the importance and actual values of each parameter to ensure the accuracy and rationality of the suitability evaluation.

[0106] S6. Analysis of procurement plan preference: Calculate the suitability for all devices participating in the bid, then arrange all the suitability levels in descending order. The highest suitability corresponds to the first-level preference, and so on, to obtain the procurement plan preference levels for all devices, providing a basis for procurement.

[0107] Through this embodiment, the efficiency and accuracy of the suitability evaluation are improved. Specifically, by constructing a recursive partitioning feature space model based on historical data and an optimized polynomial suitability calculation formula optimized by an algorithm, the suitability of various power devices can be quickly evaluated, avoiding the limitations of relying on manual experience and complex mathematical modeling in traditional methods. Through multivariate statistical methods and data cleaning processes, the accuracy and reliability of the input data are ensured, significantly reducing the error of the suitability evaluation.

[0108] The method in this embodiment is interpretable, enhancing the credibility of the model. Specifically, during the suitability evaluation process, the importance index feature_importance of the parameter and the generation of the parameter contribution details are combined. The non-linear relationship (such as the first power, second power) between the change of each parameter and the suitability is automatically analyzed and explained by the model, and users can intuitively understand the source and logic of the suitability evaluation, thus enhancing the credibility and practical application value of the model.

[0109] Automatically generating a suitability description document improves work efficiency. Specifically, traditional suitability evaluation description documents require a large amount of manpower for manual writing. In this embodiment, by automatically analyzing the model results and the recursive partitioning feature space rules, a description document including predicted suitability, parameter contribution details, and key rule explanations is generated, and supports output in multiple formats such as JSON, HTML, and PDF, reducing manual participation and improving the efficiency of generating descriptive text content.

[0110] This embodiment fully exploits the potential of historical data. By using historical project data and optimizing through recursive partitioning of the feature space model training and optimization algorithms, the implicit parameter - fitness relationship in historical data is fully utilized, forming a unified equipment parameter - fitness sample database and parameter importance indicators, further improving the scientificity and rationality of the evaluation.

[0111] This embodiment has strong applicability and is convenient for popularization and use. It is applicable to the fitness evaluation requirements in different stages (tendering, bidding, purchasing) of power equipment procurement business. It can flexibly adjust input data and model parameters, covering various equipment types and working stages. The automatically generated description content is standardized and intuitive, facilitating popularization and use in actual procurement.

[0112] Through this embodiment, the defects of the traditional fitness evaluation method, such as low work efficiency, reliance on subjective experience, and lack of interpretability, are solved. The power equipment procurement fitness evaluation and description generation process becomes more scientific, transparent, and efficient, providing important technical support for procurement decisions in the power industry. According to 10 parameters and 50 training samples for a single device, the recursive partitioning of the feature space model training and polynomial model parameter calculation process can be completed within 5 minutes on an ordinary personal computer; the generation process of a fitness calculation and a fitness calculation description can be completed within 10 seconds.

[0113] Embodiment 2:

[0114] Based on Embodiment 1, this embodiment provides a power equipment procurement analysis method based on fitness, including the following steps:

[0115] Fitness evaluation model parsing: According to the equipment fitness evaluation polynomial model suitability_evaluate = ∑f(value_n, a_n, b_n) in Embodiment 1, extract the importance index feature_importance_n, parameter value value_n, and the undetermined parameters a_n and b_n of each equipment parameter in the model.

[0116] Generate a single - parameter description template: Based on the model parsing results, generate a description of the impact of a single parameter on fitness according to the following template: "{The influence weight of parameter_name_n on fitness is feature_importance_n, its current value is value_n, the parameters fitted in the model are coefficient a_n and exponent b_n, so the contribution value to the equipment fitness is feature_importance_n×a_n×value_n b _ n}."

[0117] Comprehensive Adaptability Explanation Generation Template: Summarize the contributions of all parameters to the adaptability, and generate a complete explanation in combination with the predicted total adaptability. The template is as follows:

[0118] "Based on the input device parameters, the device adaptability predicted by the adaptability evaluation model is {suitability_evaluate}. Among them, the contribution details of each parameter are as follows:

[0119] The change of parameter {parameter_name_1} has a relationship of {b_n power relationship} with the adaptability. Calculated according to the current parameter value {value_1}, multiplied by the weight {feature_importance_1} and the model parameter {a_1}, the final contribution to the adaptability is {feature_importance_1×f(value_1, a_1, b_1)};

[0120] The change of parameter {parameter_name_2} has a relationship of {b_n power relationship} with the adaptability. Calculated according to the current parameter value {value_2}, multiplied by the weight {feature_importance_2} and the model parameter {a_2}, the final contribution to the adaptability is {feature_importance_2×f(value_2, a_2, b_2)};

[0121] And so on."

[0122] It clearly shows the influence relationship of each parameter on the adaptability, including its change form (such as first power or second power), the conversion method of parameter values, and the contribution weight, which helps users understand the model calculation logic and verify its rationality.

[0123] Automatic Explanation Generation and Output Formatting: Combine the analysis of the polynomial model and the results of recursively partitioning the feature space rules to generate a complete adaptability evaluation explanation document, including the prediction results of the model, the parameter contribution details, and the explanatory content of the recursively partitioning the feature space rules. Output the generated explanation content in a specified format (such as JSON, HTML, or PDF) for report display or system integration.

[0124] Arrange all adaptabilities in descending order. The highest adaptability corresponds to the first-level tendency, and so on, to obtain the procurement plan tendency levels of all devices, providing a basis for procurement.

[0125] Example 3:

[0126] This example provides a power equipment procurement analysis system based on adaptability, including:

[0127] A data acquisition module, configured to: obtain the device parameters participating in the bid;

[0128] The adaptation degree determination module is configured to: obtain the device adaptation degree according to the device parameters and a preset device adaptation degree evaluation model;

[0129] The procurement plan preference determination module is configured to: arrange all the adaptation degrees in descending order, with the highest adaptation degree being the first-level preference, and so on to obtain the procurement plan preference levels of all devices;

[0130] Wherein, the device adaptation degree evaluation model is a polynomial model, and each parameter in the polynomial model corresponds to a correction coefficient, a power exponent, and a parameter importance index representing the weight; the correction coefficient and the power exponent are optimized and determined by an iterative optimization method; the parameter importance index is determined by a recursive partitioning algorithm. Specifically, the device parameters are hierarchically grouped, and according to the device parameters and the data of the adaptation degree score, the partitioning threshold of each device parameter is calculated, and the device parameters that can reduce the adaptation degree score error are selected in each layer of grouping until the preset termination condition is met.

[0131] The working method of the system is the same as that of the power equipment procurement analysis method based on the adaptation degree in Embodiment 1, and will not be elaborated here.

[0132] Embodiment 4:

[0133] This embodiment provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the steps of the power equipment procurement analysis method based on the adaptation degree described in Embodiment 1 are implemented.

[0134] Embodiment 5:

[0135] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored on the memory and capable of running on the processor. When the processor executes the program, the steps of the power equipment procurement analysis method based on the adaptation degree described in Embodiment 1 are implemented.

[0136] Embodiment 6:

[0137] This embodiment provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the steps of the power equipment procurement analysis method based on the adaptation degree described in Embodiment 1 are implemented.

[0138] The above are only the preferred embodiments of this embodiment and are not used to limit this embodiment. For those skilled in the art, this embodiment can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this embodiment shall be included within the protection scope of this embodiment.

Claims

1. A power equipment procurement analysis method based on adaptability, characterized in that: include: Obtain the parameters of the equipment participating in the bidding; According to the device parameters and the preset device suitability evaluation model, the device suitability is obtained; All the adaptability is arranged in order from high to low, the highest adaptability is the first level tendency, and so on to obtain the procurement plan tendency level of all equipment; The equipment adaptability evaluation model is a polynomial model, and each parameter in the polynomial model corresponds to a correction coefficient, a power exponent, and a parameter importance index representing a weight; The correction coefficient and power index are optimized and determined by the iterative optimization method; the parameter importance index is determined by the recursive segmentation algorithm. Specifically, the device parameters are grouped in layers, and the segmentation threshold of each device parameter is calculated according to the data of the device parameters and the fitness score. The device parameters that can reduce the fitness score error are selected in each layer of grouping until the preset termination condition is met.

2. The power equipment procurement analysis method based on adaptability according to claim 1, characterized in that: The equipment parameters include the pressure, temperature and material grade of the equipment; the equipment parameters are classified and identified according to the different stages of power equipment procurement, including the bidding stage, the tendering stage and the procurement stage; at each stage, the data is marked as different categories.

3. The power equipment procurement analysis method based on adaptability according to claim 1, characterized in that: Generate a single parameter description template, including the weight of each parameter's influence on the fitness, as well as the correction coefficient and power exponent; Generate a comprehensive fitness description template that summarizes the contribution of all parameters to the fitness and generates a complete description combined with the predicted total fitness.

4. The power equipment procurement analysis method based on adaptability according to claim 1, characterized in that: The suitability_evaluate is: suitability_evaluate=∑(feature_importance_n×a_n×value_n b _ n ); Among them, a_n is the linear scale correction coefficient of equipment parameter-fitness; b_n is the nonlinear power exponent of equipment parameter; value_n is the nth equipment parameter; feature_importance_n is the nth parameter importance index.

5. The power equipment procurement analysis method based on adaptability as claimed in claim 4, characterized in that: Based on the data-driven iterative optimization method, the parameters a_n and b_n are optimized: Randomly generate initial parameter combinations a_n and b_n, and set the objective function to quantify the error between the predicted fitness and the actual fitness score; According to the current parameter combination, the error value is calculated through the objective function; Update the current parameter combination and calculate a new parameter combination to reduce the objective function value; Introduce random noise to the parameters or re-explore different initial solutions to expand the search space; In each round of iteration, the fitness value of the current parameter combination is evaluated, and the parameter combination with higher performance is retained for the next round of iteration according to the set strategy; The fitness evaluation and parameter adjustment process is repeated until the preset termination condition is met.

6. The power equipment procurement analysis method based on adaptability as claimed in claim 5, characterized in that: The objective function Loss is: Loss=(1 / N)∑(suitability_true-suitability_evaluate) 2 +λ∑(a_n 2 +b_n 2 ); Where: suitability_true is the actual fitness; suitability_evaluate is the predicted fitness; N is the total number of samples; λ is the regularization coefficient.

7. The power equipment procurement analysis method based on adaptability according to claim 1, characterized in that: Sort the parameter values ​​of different samples from small to large; for continuous device parameters, take the midpoint of adjacent values ​​as the candidate segmentation point; for discrete parameters, enumerate all possible classification combinations; for each candidate segmentation point, divide the data set into two subsets; calculate the fitness score error of each subset respectively; use the segmentation gain to measure the degree of error reduction after segmentation and calculate the segmentation gain; select the candidate point with the largest segmentation gain as the segmentation threshold of the current parameter; when determining the parameter that reduces the fitness score error, traverse all device parameters, and for each parameter, calculate all segmentation thresholds and corresponding segmentation gains; select the parameter with the largest segmentation gain as the current grouping basis.

8. A power equipment procurement analysis system based on adaptability, characterized in that: include: The data acquisition module is configured to: obtain parameters of equipment participating in the bidding; The adaptability determination module is configured to: obtain the device adaptability according to the device parameters and a preset device adaptability evaluation model; The purchase plan preference determination module is configured to: arrange all the adaptability in order from high to low, with the highest adaptability being the first-level preference, and so on to obtain the purchase plan preference level of all equipment; The equipment adaptability evaluation model is a polynomial model, and each parameter in the polynomial model corresponds to a correction coefficient, a power exponent, and a parameter importance index representing a weight; The correction coefficient and power index are optimized and determined by the iterative optimization method; the parameter importance index is determined by the recursive segmentation algorithm. Specifically, the device parameters are grouped in layers, and the segmentation threshold of each device parameter is calculated according to the data of the device parameters and the fitness score. The device parameters that can reduce the fitness score error are selected in each layer of grouping until the preset termination condition is met.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that: When the processor executes the program, the steps of the power equipment procurement analysis method based on adaptability as described in any one of claims 1 to 7 are implemented.

10. A computer program product, characterized in that The computer program product includes a computer program, and when the computer program is executed by a processor, the steps of the power equipment procurement analysis method based on adaptability are implemented as described in any one of claims 1 to 7.

Images