New energy project ESG score calculation method and device, storage medium and computer

Abstract

The invention discloses a new energy project ESG score calculation method and device, a storage medium and a computer. The method comprises the steps of determining a project stage, an index parameter and an index value of a new energy project; obtaining a first weight and a second weight of the index parameters calculated based on different algorithms, and obtaining a plurality of historical project stages and a parameter group corresponding to each historical project stage, the parameter group comprising a plurality of historical index parameters and a historical weight value of each historical index parameter; inputting each historical project stage, each parameter group, the project stage and each index parameter into a long short-term memory network model to obtain a third weight of the index parameter; obtaining a fusion index weight of the index parameters based on the first weight, the second weight and the third weight of the index parameters; and obtaining an ESG score of the new energy project based on the fusion index weight and the index value corresponding to each index parameter. According to the scheme, the calculation accuracy of the ESG score of the new energy project can be improved.

Description

Technical Field

[0001] This invention relates to the field of new energy project management technology, and in particular to a method, apparatus, storage medium and computer for calculating the ESG score of new energy projects. Background Technology

[0002] With the rapid expansion of new energy projects such as photovoltaics and wind power, investors, financial institutions, and regulators have become increasingly reliant on ESG (Environmental, Social, and Governance) scores in project financing and risk pricing. The accuracy of ESG score calculation directly affects the effectiveness of related decisions.

[0003] Currently, ESG scoring for new energy projects has significant limitations: it relies heavily on third-party ratings or self-reporting by companies, and the evaluation standards are inconsistent and difficult to quantify. Specifically, traditional scoring methods mainly use fixed weights to assign scores to indicators, with these weights often set manually. This subjective approach, which depends on expert experience, is highly subjective and lacks industry applicability. As a result, ESG scores deviate significantly from the actual ESG performance of projects, and the calculation accuracy is low, making it difficult to meet the demand for accurate ESG data in risk management and investment decisions in financial scenarios. Summary of the Invention

[0004] In view of this, this application provides a method, apparatus, storage medium and computer for calculating the ESG score of new energy projects, with the main purpose of solving the technical problem of low accuracy in calculating the ESG score of new energy projects.

[0005] According to a first aspect of the present invention, a method for calculating the ESG score of a new energy project is provided, the method comprising: Determine the current project stage of the new energy project, multiple indicator parameters of the new energy project at the project stage, and the indicator value corresponding to each indicator parameter; Obtain the first indicator weight corresponding to each of the indicator parameters calculated in advance based on the entropy method, and obtain the second indicator weight corresponding to each of the indicator parameters determined in advance based on the machine learning model; Obtain multiple consecutive historical project phases and a parameter set corresponding to each historical project phase, wherein the parameter set includes multiple historical indicator parameters and a historical weight value corresponding to each historical indicator parameter; The historical project stages, the parameter groups corresponding to each historical project stage, the project stage, and each of the indicator parameters are input into the pre-trained first long short-term memory network model to obtain the third indicator weight for each of the indicator parameters. For each of the aforementioned indicator parameters, the first indicator weight, the second indicator weight, and the third indicator weight corresponding to the indicator parameter are fused to obtain the fused indicator weight for each of the aforementioned indicator parameters. Based on the fusion index weight corresponding to each of the aforementioned index parameters, the index values ​​of all the aforementioned index parameters are weighted and summed to obtain the ESG score of the new energy project.

[0006] In an optional embodiment, each of the indicator parameters constitutes a preset parameter group with other preset number of indicator parameters; determining the multiple indicator parameters of the new energy project at the project stage and the indicator value corresponding to each indicator parameter includes: determining whether there is a null value among the indicator values ​​of each indicator parameter; if there is a null value, then the indicator parameter corresponding to the indicator value is determined as a null value indicator parameter; determining the indicator parameters belonging to the same preset parameter group as the null value indicator parameter as associated indicator parameters; inputting the associated indicator parameters, the indicator values ​​of the associated indicator parameters, and the null value indicator parameters into an indicator prediction model pre-trained for the preset parameter group to obtain the predicted indicator value output by the indicator prediction model; and using the predicted indicator value as the indicator value corresponding to the null value indicator parameter.

[0007] In an optional embodiment, after determining the current project stage of the new energy project, the multiple indicator parameters of the new energy project at the project stage, and the indicator value corresponding to each indicator parameter, the method further includes: performing feature standardization processing on each indicator value to obtain the indicator value after feature standardization processing; and performing the steps of obtaining the first indicator weight corresponding to each indicator parameter pre-calculated based on the entropy method, and obtaining the second indicator weight corresponding to each indicator parameter pre-determined based on the machine learning model.

[0008] In an optional embodiment, the step of fusing the first indicator weight, the second indicator weight, and the third indicator weight corresponding to the indicator parameter to obtain the fused indicator weight for each indicator parameter includes: calculating the average weight of the first indicator weight, the second indicator weight, and the third indicator weight; and calculating the first weight difference between the first indicator weight and the average weight, the second weight difference between the second indicator weight and the average weight, and the third weight difference between the third indicator weight and the average weight; determining whether the first weight difference, the second weight difference, and the third weight difference are all less than a preset threshold; if the first weight difference, the second weight difference, and the third weight difference are not all less than the preset threshold, determining the average weight as the fused indicator weight.

[0009] In an optional embodiment, the calculation method of the first indicator weight includes: obtaining a historical indicator dataset corresponding to multiple historical project stages, wherein the historical indicator dataset includes multiple indicator parameters and an indicator value corresponding to each indicator parameter; associating the same indicator parameters among all the indicator parameters to obtain a set of similar indicators, and calculating the coefficient of variation among the indicator values ​​corresponding to all the indicator parameters in the set of similar indicators; and determining the first indicator weight corresponding to the indicator parameter based on the magnitude of the coefficient of variation corresponding to each indicator parameter.

[0010] In an optional embodiment, the calculation of the second indicator weight includes: obtaining a historical indicator dataset corresponding to multiple historical project stages, and an ESG score corresponding to each historical project stage, wherein the historical indicator dataset includes multiple indicator parameters and an indicator value corresponding to each indicator parameter; inputting the historical indicator dataset and ESG score corresponding to each historical project stage into a pre-trained ensemble learning model to determine the contribution of the indicator value corresponding to each indicator parameter to the ESG score; and determining the second indicator weight corresponding to the indicator parameter based on the contribution of the indicator value corresponding to the indicator parameter to the ESG score.

[0011] In an optional embodiment, after obtaining the ESG score of the new energy project by weighted summation of the index values ​​of all the index parameters based on the fusion index weight corresponding to each index parameter, the method further includes: obtaining the ESG scores corresponding to multiple consecutive project stages, and smoothing the multiple ESG scores based on the Kalman filter algorithm to obtain multiple smoothed ESG scores; inputting each ESG score and the project stage corresponding to the ESG score into a pre-trained second long short-term memory network model to obtain the ESG score of the next project stage of the new energy project output by the second long short-term memory network model.

[0012] According to a second aspect of the present invention, a device for calculating the ESG score of a new energy project is provided, the device comprising: The data acquisition module is used to determine the current project stage of the new energy project, multiple indicator parameters of the new energy project at the project stage, and the indicator value corresponding to each indicator parameter. The weight acquisition module is used to acquire the first indicator weight corresponding to each of the indicator parameters calculated in advance based on the entropy method, and to acquire the second indicator weight corresponding to each of the indicator parameters determined in advance based on the machine learning model. The parameter acquisition module is used to acquire multiple consecutive historical project stages and the parameter group corresponding to each historical project stage, wherein the parameter group includes multiple historical indicator parameters and the historical weight value corresponding to each historical indicator parameter. The model calculation module is used to input the historical project stage, the parameter group corresponding to each historical project stage, the project stage, and each of the indicator parameters into the pre-trained first long short-term memory network model to obtain the third indicator weight of each indicator parameter. The weight fusion module is used to fuse the first indicator weight, the second indicator weight, and the third indicator weight corresponding to each indicator parameter to obtain the fused indicator weight for each indicator parameter. The scoring calculation module is used to perform a weighted summation of the index values ​​of all the index parameters based on the fusion index weight corresponding to each index parameter, so as to obtain the ESG score of the new energy project.

[0013] According to a third aspect of the present invention, a storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the above-described method for calculating the ESG score of a new energy project.

[0014] According to a fourth aspect of the present invention, a computer is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-mentioned method for calculating the ESG score of new energy projects.

[0015] This invention provides a method, apparatus, storage medium, and computer for calculating ESG scores of new energy projects. It acquires the indicator parameters corresponding to the current stage of the new energy project, as well as the corresponding indicator values ​​for each parameter. It introduces three types of weights: a first indicator weight calculated using the entropy method (objectively reflecting the degree of indicator variation), a second indicator weight determined by a machine learning model (capturing nonlinear correlations between indicators), and a third indicator weight based on a Long Short-Term Memory (LSTM) network model combined with historical data (reflecting the dynamic evolution of indicators). By fusing these three types of weights, a fused indicator weight that combines objectivity, correlation, and dynamism is obtained. This solves the scoring bias caused by traditional single or fixed weights, ensures the consistency between ESG scores and actual performance, and improves the accuracy of ESG score calculation for new energy projects.

[0016] 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

[0017] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 A flowchart illustrating a method for calculating the ESG score of a new energy project according to an embodiment of the present invention is shown. Figure 2 This diagram illustrates the structure of a device for calculating the ESG score of a new energy project according to an embodiment of the present invention. Figure 3 This invention provides a schematic diagram of the structure of another device for calculating the ESG score of a new energy project according to an embodiment of the present invention. Figure 4 The diagram shows a schematic of the structure of a computer that performs a method for calculating the ESG score of a new energy project, according to an embodiment of the present invention. Detailed Implementation

[0018] The present invention will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the present application can be combined with each other.

[0019] Currently, ESG scoring for new energy projects has significant limitations. It relies heavily on third-party ratings or self-reporting by companies, with inconsistent evaluation standards and difficulty in quantification. Specifically, traditional evaluation methods primarily use fixed weighted scores for each indicator parameter. These weights are often set manually, relying on expert experience, which is highly subjective and lacks industry adaptability. This results in a large discrepancy between the ESG score and the actual ESG performance of the project, and low calculation accuracy, making it difficult to meet the demand for accurate ESG data in risk management and investment decisions in financial scenarios.

[0020] To address the above problems, in one embodiment, such as Figure 1 As shown, a method for calculating the ESG score of a new energy project is provided. The method is applied to a computer to calculate the ESG score of a new energy project as an example, and includes the following steps: 101. Determine the current project stage of the new energy project, multiple indicator parameters of the new energy project at the project stage, and the indicator value corresponding to each indicator parameter.

[0021] In this context, the project phases of a new energy project can refer to the continuous stages throughout the entire lifecycle of the project. These phases are typically logically divided into "preliminary preparation," "construction and implementation," "operation and maintenance," and "final exit." Alternatively, they can be divided according to the time frame of the project, such as monthly or annually, with each month or year considered a project phase. No specific limitation is made here. The indicator parameters can be the names of the indicators within the new energy project, and the indicator values ​​can be the corresponding numerical values.

[0022] Furthermore, the indicators and parameters for new energy projects can be pre-defined, employing a hierarchical structure to ensure comprehensive coverage and clear logic. Specifically, primary indicators include three dimensions: Environment, Social, and Governance. Based on this, secondary sub-dimensions are established. For example, indicators under the Environment dimension include carbon emissions, energy efficiency, and waste disposal; indicators under the Social dimension include safe production, employment contribution, and supply chain compliance; and indicators under the Governance dimension include information disclosure transparency, board structure, and compliance management. At the tertiary level, each secondary-level indicator is further refined into directly quantifiable parameters, such as carbon emission intensity (in tons). (e.g., per megawatt-hour), power generation per unit installed capacity, accident rate, governance transparency score, etc. Furthermore, each indicator parameter corresponds to a numerical value obtained through data collection; for example, the numerical value corresponding to the indicator parameter "carbon emission intensity" could be "20 tons". / megawatt-hours.

[0023] Furthermore, for indicator parameters that cannot be directly quantified, such as "board structure," the indicator value can be quantified through expert evaluation, text mining, or natural language processing methods. Here, the content of each unquantifiable indicator parameter can be pre-scored, and the score can be used as the indicator value. For example, if the content of the indicator parameter "board structure" is "the chairman and general manager are combined into one," and the preset score for this content is "60," then the indicator value corresponding to the indicator parameter "board structure" can be determined as 60.

[0024] In practical work, indicator parameters corresponding to the current stage of a new energy project can be collected from operational data, carbon emission monitoring data, corporate financial and governance data, and social responsibility data. The values ​​corresponding to these indicator parameters from the aforementioned information sources are then determined as the indicator values. Alternatively, each required indicator parameter can be pre-determined, and based on a large language model, these parameters can be identified from the operational data, carbon emission monitoring data, corporate financial and governance data, and social responsibility data. The values ​​corresponding to these indicator parameters are then determined as the indicator values. For the same new energy project, the indicator parameters corresponding to different project stages are the same; for example, they may all include indicator parameters such as "power generation per unit installed capacity," "accident incidence rate," and "governance transparency score," only the indicator values ​​differ.

[0025] Furthermore, if the indicator parameter is an unquantifiable indicator parameter, the content of the indicator parameter can be obtained from the aforementioned information source, and a pre-set score can be determined for the content, and this score can be determined as the indicator value of the indicator parameter.

[0026] For example, information about the current project stage of a new energy project can be obtained, such as the "preliminary preparation" stage or "XXXX year X month", and the corresponding indicator parameters and values ​​of the project stage can be linked for later use.

[0027] 102. Obtain the first indicator weight corresponding to each of the indicator parameters calculated in advance based on the entropy method, and obtain the second indicator weight corresponding to each of the indicator parameters determined in advance based on the machine learning model; Here, the entropy method, a statistical approach, can be used to calculate the first indicator weight for each parameter. The entropy method measures the degree of variability of an indicator by calculating its information entropy. If, in historical data, the numerical differences of the same indicator parameter across different project phases are significant, it indicates that the parameter provides richer information, and its weight should be higher. For example, if the numerical fluctuation of the same type of indicator parameter, "carbon emission intensity," is large across different project phases, its first indicator weight can be increased accordingly; conversely, for indicator parameters with small numerical fluctuations, their first indicator weight can be appropriately decreased.

[0028] Specifically, the calculation method for the weight of the first indicator can be as follows: First, obtain historical indicator datasets corresponding to multiple historical project stages. Each historical indicator dataset includes multiple indicator parameters and the corresponding indicator value for each parameter. Here, the indicator parameters corresponding to each type of new energy project in different project stages in history, as well as the corresponding indicator value for each parameter, can be obtained. The indicator parameters and indicator values ​​belonging to the same historical project stage are then associated to obtain the historical indicator dataset.

[0029] Then, among all the indicator parameters, the same indicator parameters are associated to obtain a set of similar indicators, and the coefficient of variation (COP) between the indicator values ​​corresponding to all the indicator parameters in the set of similar indicators is calculated. Here, the same indicator parameters can be identified among all the indicator parameters corresponding to multiple historical project stages. For example, all the "carbon emission intensity" indicator parameters can be associated to obtain a set of similar indicators corresponding to "carbon emission intensity". Further, the COP between the indicator values ​​of each indicator parameter in the set of similar indicators is calculated.

[0030] Finally, based on the magnitude of the discrete coefficient corresponding to each indicator parameter, the first indicator weight corresponding to the indicator parameter is determined. Specifically, multiple different discrete coefficient intervals can be preset. The larger the value of the discrete coefficient interval, the larger the preset first indicator weight for that discrete coefficient interval. If the discrete coefficient corresponding to a set of similar indicators falls within a certain discrete coefficient interval, then the first indicator weight of that discrete coefficient interval is determined as the first indicator weight of the indicator parameter of the set of similar indicators.

[0031] Furthermore, principal component analysis (PCA) can be used in advance to determine the first indicator weight for each indicator parameter. Here, PCA is used to reduce the dimensionality and linearly combine multidimensional ESG indicators, extract the key principal components that can explain most of the variance, and calculate the indicator weights based on the principal component loadings, thereby reducing redundancy and multicollinearity issues among indicators.

[0032] Furthermore, at the machine learning method level, models such as Random Forest and XGBoost can be introduced to determine the secondary indicator weight for each indicator parameter. Here, the Random Forest model can construct multiple decision trees and integrate the results. Based on the contribution of each indicator parameter during tree splitting (such as Gini importance or information gain), the importance score of the indicator parameter can be determined. If the indicator value of a certain parameter contributes significantly to the ESG score, it can be assigned a higher secondary indicator weight. Furthermore, the XGBoost model, as a gradient boosting algorithm, can capture the non-linear relationship between indicator parameters and ESG scores. Its feature importance is measured by the cumulative gain of the feature during model splitting, reflecting the marginal contribution of different indicator parameters to the overall ESG performance, thereby determining the secondary indicator weight for each indicator parameter.

[0033] Here, the calculation method for the weight of the second indicator can be as follows: First, obtain the historical indicator dataset corresponding to multiple consecutive historical project stages, and the ESG score corresponding to each historical project stage. The historical indicator dataset includes multiple indicator parameters and the indicator value corresponding to each indicator parameter. Here, the ESG score, indicator parameters, and indicator value corresponding to each indicator parameter for each type of new energy project in different project stages in history can be obtained. The indicator parameters and indicator values ​​belonging to the same historical project stage are associated to obtain the historical indicator dataset.

[0034] Then, the historical indicator dataset and ESG scores corresponding to each historical project stage are input into a pre-trained ensemble learning model to determine the contribution of each indicator parameter's value to the ESG score. Here, the ensemble learning model can be a pre-trained random forest model or an XGBoost model. The ESG scores, indicator parameters, and indicator values ​​corresponding to each historical project stage can be input into the ensemble learning model, which then outputs the contribution of each indicator parameter.

[0035] Finally, based on the contribution of the indicator values ​​corresponding to the indicator parameters to the ESG score, the weight of the second indicator corresponding to the indicator parameters is determined.

[0036] 103. Obtain multiple consecutive historical project phases and parameter sets corresponding to each historical project phase, wherein the parameter sets include multiple historical indicator parameters and historical weight values ​​corresponding to each historical indicator parameter; Here, based on the pre-defined sequential relationship of each project stage, multiple historical project stages preceding the current project stage of a new energy project can be identified. For example, if all project stages of a new energy project are sequentially divided into "preliminary preparation", "construction and implementation", "operation and maintenance" and "closing and exit", and the current project stage of the new energy project is "operation and maintenance", then the project stages preceding the current project stage, "preliminary preparation" and "construction and implementation", can be identified as historical project stages.

[0037] Furthermore, for each historical project stage, historical indicator parameters and historical weight values ​​corresponding to each historical indicator parameter can be obtained from past data, and the historical indicator parameters and historical weight values ​​of the historical project stage can be associated as a parameter group.

[0038] 104. Input the historical project stage, the parameter group corresponding to each historical project stage, the project stage, and each of the indicator parameters into the pre-trained first long short-term memory network model to obtain the third indicator weight of each of the indicator parameters; The first Long Short-Term Memory (LSTM) network model is suitable for capturing the dynamic changes in historical weight values ​​of new energy projects at different historical project stages over time. It can identify long-term dependencies in the input sequence, thereby enabling the time-series update of the third indicator weights and overall scores of the indicator parameters. Here, historical indicator parameters and historical weight values ​​from different historical project stages can be input into the LSTM model as different batches of time series data to obtain the third indicator weights of each indicator parameter for the current project stage output by the LSTM model.

[0039] Here, historical indicator parameters and historical weight values ​​of different historical project stages can be obtained as training data. The weight values ​​of each indicator parameter corresponding to the next project stage can be used as training labels to train the LSTM model and obtain the first long short-term memory network model.

[0040] Furthermore, the first long short-term memory network model can also be a Transformer model. The Transformer model can focus on the changes in the historical weight values ​​of historical indicator parameters in multiple project stages based on the self-attention mechanism. It is suitable for dynamic modeling of high-dimensional, multi-source heterogeneous ESG data and can improve the accuracy and robustness of weight prediction.

[0041] 105. For each of the indicator parameters, the first indicator weight, the second indicator weight, and the third indicator weight corresponding to the indicator parameter are fused to obtain the fused indicator weight for each of the indicator parameters. Specifically, the average weights of the first indicator, the second indicator, and the third indicator can be calculated to determine the average of the three values. Furthermore, the first weight difference between the first indicator weight and the average weight, the second weight difference between the second indicator weight and the average weight, and the third weight difference between the third indicator weight and the average weight can be calculated respectively. Furthermore, it is determined whether the first, second, and third weight differences are all less than a preset threshold. If none of these three differences are less than the preset threshold, the average weight is determined as the weight of the fusion index. The preset threshold can be used to determine whether the three weight differences deviate too much from the average weight; its value can be determined based on the actual situation. Further, if all three weight differences are less than the preset threshold, any one of these three values ​​can be determined as the weight of the fusion index.

[0042] 106. Based on the fusion index weight corresponding to each of the aforementioned index parameters, the index values ​​of all the aforementioned index parameters are weighted and summed to obtain the ESG score of the new energy project.

[0043] Here, the ESG score of a new energy project can be calculated based on Formula 1: (1) in, For ESG score, This refers to the dynamic weights determined by a combination of statistical methods (entropy method, PCA) and machine learning methods (random forest, XGBoost, LSTM / Transformer dynamic update), i.e., fused index weights. Let be the value of the i-th indicator parameter in project phase t.

[0044] Furthermore, the calculated ESG scores can be used as training data for different types of financial models, and embedded as key explanatory variables into multiple financial models to achieve a quantitative assessment of the financing risks and investment value of new energy projects. Specifically, in credit risk prediction, default risk models such as Logit, Probit, or Cox Hazard Model can be used to incorporate ESG scores into input variables, thereby revealing the impact of ESG scores on the probability of project default. Furthermore, in green bond and financing rate modeling, multiple regression models, XGBoost models, random forests, and other machine learning methods can be used to predict the marginal effect of ESG scores on financing rates and credit spreads, quantifying the effect of ESG scores on adjusting financing costs. Further, in portfolio optimization, ESG constraints can be introduced into the Markowitz mean-variance framework to construct an ESG-adjusted Efficient Frontier architecture, which can improve portfolio sustainability while balancing returns and risks. Finally, in financial market risk contagion modeling, ESG scores can be introduced as a moderating factor in DCC-GARCH dynamic correlation models and Hawkes Process jump risk models, characterizing the amplification or buffering effect of ESG performance on market volatility and risk spillover effects. Through the above methods, this solution achieves deep integration of ESG evaluation results with financial risk modeling, interest rate spread pricing, and investment decisions, promoting the practical application of ESG scores in the field of green finance.

[0045] The method for calculating ESG scores for new energy projects provided in this embodiment obtains the indicator parameters corresponding to the current stage of the new energy project, as well as the indicator values ​​corresponding to each indicator parameter. It introduces three types of weights: the first indicator weight calculated by the entropy method (objectively reflecting the degree of indicator variation), the second indicator weight determined by the machine learning model (capturing the nonlinear correlation of indicators), and the third indicator weight based on the LSTM model combined with historical data (reflecting the dynamic evolution law of indicators). By fusing the three types of weights, a fused indicator weight that combines objectivity, correlation, and dynamism is obtained. This solves the score calculation deviation caused by traditional single or fixed weights, ensures the consistency between ESG scores and actual performance, and improves the accuracy of ESG score calculation for new energy projects.

[0046] In an optional embodiment, each indicator parameter can be pre-grouped with a predetermined number of other indicator parameters from all the indicator parameters at each project stage. Here, certain indicator parameters in a new energy project may be correlated, and the values ​​of each indicator parameter may influence each other. Therefore, correlated indicator parameters can be pre-organized to obtain different preset parameter groups.

[0047] Furthermore, determining the multiple indicator parameters of the new energy project at the project stage and the corresponding indicator value of each indicator parameter includes: first, determining whether there are null values ​​in the indicator value of each indicator parameter; if there are null indicator values, then the indicator parameter corresponding to the indicator value is determined as a null indicator parameter; specifically, all indicator values ​​of the current project stage of the new energy project can be checked to determine whether there are null values ​​in the indicator value of each indicator parameter, and the indicator parameters with null indicator values ​​are determined as null indicator parameters.

[0048] Then, other indicator parameters belonging to the same preset parameter group as the null value indicator parameter are identified as associated indicator parameters. Here, the preset parameter group to which the null value indicator parameter belongs can be determined, and other indicator parameters in the preset parameter group other than the null value indicator parameter are identified as associated indicator parameters.

[0049] Then, the associated indicator parameters, the indicator values ​​of the associated indicator parameters, and the null value indicator parameters are input into the indicator prediction model pre-trained for the preset parameter group to obtain the predicted indicator values ​​output by the indicator prediction model. Here, the training method for the indicator prediction model can be as follows: obtain multiple historical parameter groups, each historical parameter group including multiple indicator parameters belonging to the same preset parameter group and the indicator value of each indicator parameter; further, set the indicator value of one indicator parameter to a null value, use the indicator value before setting to a null value as the label of the predicted indicator value, train the neural network model, and obtain an indicator prediction model that can identify the predicted indicator value corresponding to the null value.

[0050] Finally, the predicted index value is used as the index value corresponding to the null value index parameter. The embodiments provided in this application can handle the null value problem by grouping related index parameters into preset parameter groups and utilizing the characteristics of mutual influence between indicators. By accurately identifying null value index parameters and related index parameters, and then using a neural network model to predict null values ​​based on the related index values, errors caused by simply removing or filling null values ​​are avoided. This provides high-quality data support for subsequent ESG score calculations and reduces the interference of null values ​​on the scoring results.

[0051] In an optional embodiment, after determining the current project stage of the new energy project, the multiple indicator parameters of the new energy project at the project stage, and the indicator value corresponding to each indicator parameter, the method further includes: First, each indicator value is subjected to feature standardization to obtain the indicator value after feature standardization. Specifically, the indicator values ​​can be standardized and normalized based on algorithms such as Min-Max and Z-score. Here, the Min-Max algorithm can map the indicator values ​​to the [0,1] interval, and the Z-score algorithm can make the mean of the indicator values ​​0 and the standard deviation 1. Both can eliminate the differences caused by the unit of measurement (such as tons or times) and the range of values, and avoid the problem of unbalanced weights in ESG scoring due to different indicator values ​​(such as carbon emission intensity and accident rate), thus ensuring the fairness of subsequent weight calculation and score evaluation.

[0052] Then, the steps described in step 102—obtaining the first indicator weight corresponding to each indicator parameter pre-calculated based on the entropy method and obtaining the second indicator weight corresponding to each indicator parameter pre-determined based on the machine learning model—are executed. The embodiments provided in this application standardize indicator values ​​using Min-Max or Z-score, eliminating differences in measurement units and numerical ranges between different indicators. This avoids weight imbalances in ESG scoring due to differences in indicator values, making the processed indicator values ​​more comparable. It provides a fair data foundation for subsequent weight calculations using the entropy method and machine learning model, ensuring the accuracy and rationality of ESG score calculations.

[0053] In an optional embodiment, after obtaining the ESG score of the new energy project by weighted summation of the index values ​​of all the index parameters based on the fusion index weight corresponding to each index parameter, the method further includes: obtaining the ESG scores corresponding to multiple consecutive project stages, and smoothing the multiple ESG scores based on the Kalman filter algorithm to obtain multiple smoothed ESG scores. Here, the Kalman filter algorithm can update and smooth ESG scores in real time under noisy environments, ensuring the robustness and timeliness of subsequent evaluation results. Specifically, the ESG scores calculated for the new energy project in multiple consecutive project stages in step 106 can be collected, and Kalman filters can be applied to them to obtain multiple smoothed ESG scores.

[0054] Then, each ESG score and the corresponding project stage are input into a pre-trained second long short-term memory (LSTM) network model to obtain the ESG score corresponding to the next project stage of the new energy project, as output by the second LSTM network model. For example, the ESG scores calculated in the "preliminary preparation," "construction and implementation," and "operation and maintenance" stages of a new energy project can be input into the second LSTM network model to predict the ESG score corresponding to the subsequent "closing and exit" stage of the new energy project, as output by the second LSTM network model.

[0055] Here, the second Long Short-Term Memory (LSTM) network model can also be an LSTM model. This LSTM model captures the ESG scores of new energy projects at different project stages, identifying long-term dependencies between ESG scores in the input sequence composed of multiple ESG scores, thereby achieving time-series updates of ESG scores. Specifically, the ESG scores of different project stages can be input as time series data into the LSTM model, yielding the LSTM model outputting the ESG scores for the next project stage.

[0056] Furthermore, the calculated ESG scores for each project stage can be dynamically coupled with relevant financial indicators (such as financing rates, green bond spreads, and credit risk premiums). For example, a view containing the relationship between ESG scores and financial indicators for each project stage can be generated, including curves showing the change of ESG scores over time and graphs showing the correlation between ESG scores and financial indicators. This allows the relationship between ESG scores and various financial indicators to be presented graphically in a visualization platform, thereby providing intuitive and dynamic decision support for investors, financial institutions, and regulators.

[0057] Here, in calculating the weights of indicator parameters, in addition to using the entropy method, principal component analysis, random forest, and XGBoost, Bayesian hierarchical models or factor analysis can also be introduced for weight estimation. Bayesian hierarchical models, by introducing prior and posterior update mechanisms into a hierarchical structure, can still derive robust indicator weights even in scenarios with scarce data or uncertainty. Factor analysis, by modeling the potential correlations among multiple indicators, groups highly correlated variables into common factors, thereby achieving a reasonable allocation of indicator weights. These methods can effectively reflect the relative importance of ESG indicators, ensuring the scientific rigor and interpretability of the overall ESG score.

[0058] Furthermore, regarding the dynamic evolution and update mechanism of ESG scores, in addition to schemes based on LSTM (Long Short-Term Memory) networks and Kalman filters, time-varying parameter vector autoregressive (TVP-VAR) models or Bayesian dynamic models (BDM) can also be used to achieve real-time correction. TVP-VAR can capture the interaction relationship between ESG indicators over time, clearly reflecting the dynamic dependency structure of the indicators; BDM, by introducing a Bayesian estimation framework, gradually updates the fluctuations of ESG scores over time, exhibiting good adaptability and robustness. Through the above alternative methods, the overall goal of this invention can also be achieved, namely, ensuring the objectivity, flexibility, and robustness of the ESG weight allocation and dynamic update process.

[0059] The ESG score calculation method for new energy projects provided in this embodiment integrates the entropy method, random forest / XGBoost, and LSTM / Transformer algorithm dynamic models through a weight fusion system. This overcomes the shortcomings of traditional fixed weight assignment, which is too subjective, and makes the final output ESG score highly synchronized with the actual performance of the project. It can be directly embedded into credit risk, bond spread, portfolio optimization, and risk contagion models, providing financial institutions with real-time, auditable, and implementable green investment and financing decision-making basis, and significantly improving the efficiency and transparency of capital allocation throughout the entire life cycle of new energy projects.

[0060] 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, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. In addition, the labels corresponding to each step in the above embodiments are only for identification purposes and are not intended to limit the execution order of the steps. The execution order of the steps in each embodiment can be set according to the actual situation.

[0061] Furthermore, as Figure 1 The specific implementation of the method shown in this embodiment provides a device for calculating the ESG score of new energy projects, such as... Figure 2 As shown, the device includes: a data acquisition module 21, a weight acquisition module 22, a parameter acquisition module 23, a model calculation module 24, a weight fusion module 25, and a score calculation module 26.

[0062] Data acquisition module 21 can be used to determine the current project stage of the new energy project, multiple indicator parameters of the new energy project at the project stage, and the indicator value corresponding to each indicator parameter; The weight acquisition module 22 can be used to acquire the first index weight corresponding to each of the index parameters calculated in advance based on the entropy method, and to acquire the second index weight corresponding to each of the index parameters determined in advance based on the machine learning model. The parameter acquisition module 23 can be used to acquire multiple consecutive historical project stages and the parameter group corresponding to each historical project stage, wherein the parameter group includes multiple historical indicator parameters and the historical weight value corresponding to each historical indicator parameter. The model calculation module 24 can be used to input the historical project stage, the parameter group corresponding to each historical project stage, the project stage, and each of the indicator parameters into the pre-trained first long short-term memory network model to obtain the third indicator weight of each indicator parameter. The weight fusion module 25 can be used to fuse the first indicator weight, the second indicator weight and the third indicator weight corresponding to each indicator parameter to obtain the fused indicator weight of each indicator parameter. The scoring calculation module 26 can be used to perform a weighted summation of the index values ​​of all the index parameters based on the fusion index weight corresponding to each index parameter, so as to obtain the ESG score of the new energy project.

[0063] In specific application scenarios, each of the indicator parameters forms a preset parameter group with other preset number of indicator parameters. The data acquisition module 21 is specifically used to determine whether there are null values ​​in the indicator values ​​of each indicator parameter. If there are null indicator values, the indicator parameter corresponding to the indicator value is determined as a null indicator parameter. The indicator parameters belonging to the same preset parameter group as the null indicator parameter are determined as associated indicator parameters. The associated indicator parameters, the indicator values ​​of the associated indicator parameters, and the null indicator parameters are input into an indicator prediction model pre-trained for the preset parameter group to obtain the predicted indicator value output by the indicator prediction model. The predicted indicator value is used as the indicator value corresponding to the null indicator parameter.

[0064] In specific application scenarios, the data acquisition module 21 can also be used to perform feature standardization processing on each of the indicator values ​​to obtain the indicator values ​​after feature standardization processing.

[0065] In a specific application scenario, the weight fusion module 25 can be used to calculate the average weight of the first indicator weight, the second indicator weight, and the third indicator weight, and to calculate the first weight difference between the first indicator weight and the average weight, the second weight difference between the second indicator weight and the average weight, and the third weight difference between the third indicator weight and the average weight; determine whether the first weight difference, the second weight difference, and the third weight difference are all less than a preset threshold; if the first weight difference, the second weight difference, and the third weight difference are not all less than the preset threshold, the average weight is determined as the fused indicator weight.

[0066] In specific application scenarios, such as Figure 3As shown, the device also includes a score prediction module 27, which is specifically used to obtain ESG scores corresponding to multiple consecutive project stages, and to smooth the multiple ESG scores based on the Kalman filter algorithm to obtain multiple smoothed ESG scores; each ESG score and the project stage corresponding to the ESG score are input into a pre-trained second long short-term memory network model to obtain the ESG score corresponding to the next project stage of the new energy project output by the second long short-term memory network model.

[0067] It should be noted that other corresponding descriptions of the functional units involved in the ESG score calculation device for new energy projects provided in this embodiment can be found in [reference needed]. Figure 1 The corresponding description in [the document] will not be repeated here.

[0068] Based on the above, Figure 1 Accordingly, this embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the above-described method. Figure 1 The method for calculating the ESG score of new energy projects is shown.

[0069] Based on this understanding, the technical solution of this application can be embodied in the form of a software product. The software product to be identified can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause a computer (such as a personal computer, server, or network device, etc.) to execute the methods described in the various implementation scenarios of this application.

[0070] Based on the above, Figure 1 The method shown, and Figure 2 and Figure 3 The embodiment of the calculation device for the ESG score of new energy projects shown is intended to achieve the above objective, such as... Figure 4 As shown, this embodiment also provides a computer for calculating the ESG score of a new energy project. Specifically, it can be a personal computer, server, smartphone, tablet, smartwatch, or other network device. The computer includes a storage medium and a processor; the storage medium stores the computer program and operating system; the processor executes the computer program to achieve the above-mentioned... Figure 1 The method shown.

[0071] Optionally, the computer may also include internal memory, a communication interface, a network interface, a camera, radio frequency (RF) circuitry, sensors, audio circuitry, a Wi-Fi module, a display screen, and input devices such as a keyboard. The communication interface may also include a USB interface, a card reader interface, etc. The network interface may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0072] Those skilled in the art will understand that the computer structure for recognizing operational actions provided in this embodiment does not constitute a limitation on the computer, and may include more or fewer components, or combine certain components, or have different component arrangements.

[0073] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the aforementioned computer hardware and software resources to be identified, supporting the operation of information processing programs and other software and / or programs to be identified. The network communication module is used to enable communication between the various components within the storage medium, as well as communication with other hardware and software in the information processing computer.

[0074] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platform, or it can be implemented by hardware. By applying the technical solution of this application, firstly, the current project stage of the new energy project, multiple indicator parameters of the new energy project in the project stage, and the indicator value corresponding to each indicator parameter are determined; then, a first indicator weight corresponding to each indicator parameter, pre-calculated based on the entropy method, is obtained, and a second indicator weight corresponding to each indicator parameter, pre-determined based on a machine learning model, is obtained; next, multiple consecutive historical project stages and parameter groups corresponding to each historical project stage are obtained, wherein the parameter groups include multiple historical indicator parameters and historical weight values ​​corresponding to each historical indicator parameter; next, the historical project stages, the parameter groups corresponding to each historical project stage, the project stage, and each indicator parameter are input into a pre-trained first long short-term memory network model to obtain a third indicator weight for each indicator parameter; next, for each indicator parameter, the first indicator weight, the second indicator weight, and the third indicator weight corresponding to the indicator parameter are fused to obtain a fused indicator weight for each indicator parameter; finally, based on the fused indicator weight corresponding to each indicator parameter, the indicator values ​​of all indicator parameters are weighted and summed to obtain the ESG score of the new energy project. Compared with existing technologies, this method can improve the accuracy of calculating ESG scores for new energy projects.

[0075] 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. Those skilled in the art will understand that the modules in the apparatus of the embodiment can be distributed within the apparatus of the embodiment as described, or they can be modified to be located in one or more apparatuses different from this embodiment. The modules of the above-described embodiment can be combined into one module, or further divided into multiple sub-modules.

[0076] The serial numbers in this application are for descriptive purposes only and do not represent the superiority or inferiority of any particular implementation scenario. The above disclosures are merely a few specific implementation scenarios of this application; however, this application is not limited thereto, and any variations conceived by those skilled in the art should fall within the protection scope of this application.

Claims

1. A method for calculating the ESG score of a new energy project, characterized in that, The method includes: Determine the current project stage of the new energy project, multiple indicator parameters of the new energy project at the project stage, and the indicator value corresponding to each indicator parameter; Obtain the first indicator weight corresponding to each of the indicator parameters calculated in advance based on the entropy method, and obtain the second indicator weight corresponding to each of the indicator parameters determined in advance based on the machine learning model; Obtain multiple consecutive historical project phases and a parameter set corresponding to each historical project phase, wherein the parameter set includes multiple historical indicator parameters and a historical weight value corresponding to each historical indicator parameter; The historical project stages, the parameter groups corresponding to each historical project stage, the project stage, and each of the indicator parameters are input into the pre-trained first long short-term memory network model to obtain the third indicator weight for each of the indicator parameters. For each of the aforementioned indicator parameters, the first indicator weight, the second indicator weight, and the third indicator weight corresponding to the indicator parameter are fused to obtain the fused indicator weight for each of the aforementioned indicator parameters. Based on the fusion index weight corresponding to each of the aforementioned index parameters, the index values ​​of all the aforementioned index parameters are weighted and summed to obtain the ESG score of the new energy project.

2. The method according to claim 1, characterized in that, Each of the aforementioned indicator parameters, together with a preset number of other indicator parameters, constitutes a preset parameter group; The determination of multiple indicator parameters of the new energy project at the project stage and the corresponding indicator value for each indicator parameter includes: Determine whether there are null values ​​in the indicator values ​​of each indicator parameter. If there are null values, then the indicator parameter corresponding to the indicator value is determined as a null value indicator parameter. The index parameters that belong to the same preset parameter group as the null value index parameters are identified as associated index parameters; The associated indicator parameters, the indicator values ​​of the associated indicator parameters, and the null value indicator parameters are input into the indicator prediction model pre-trained for the preset parameter group to obtain the predicted indicator values ​​output by the indicator prediction model. The predicted index value is used as the index value corresponding to the null value index parameter.

3. The method according to claim 1, characterized in that, After determining the current project stage of the new energy project, the multiple indicator parameters of the new energy project at the project stage, and the indicator value corresponding to each indicator parameter, the method further includes: Each of the aforementioned index values ​​is subjected to feature standardization processing to obtain the index values ​​after feature standardization. The steps are as follows: obtaining the first indicator weight corresponding to each indicator parameter calculated in advance based on the entropy method, and obtaining the second indicator weight corresponding to each indicator parameter determined in advance based on the machine learning model.

4. The method according to any one of claims 1 to 3, characterized in that, The step of fusing the first indicator weight, the second indicator weight, and the third indicator weight corresponding to the indicator parameter to obtain the fused indicator weight for each indicator parameter includes: Calculate the average weight of the first indicator weight, the second indicator weight, and the third indicator weight, and calculate the first weight difference between the first indicator weight and the average weight, the second weight difference between the second indicator weight and the average weight, and the third weight difference between the third indicator weight and the average weight. Determine whether the first weight difference, the second weight difference, and the third weight difference are all less than a preset threshold. If the first weight difference, the second weight difference, and the third weight difference are not all less than the preset threshold, determine the average weight as the fusion index weight.

5. The method according to claim 4, characterized in that, The calculation method for the weight of the first indicator includes: Obtain historical indicator datasets corresponding to multiple historical project phases, wherein the historical indicator datasets include multiple indicator parameters and indicator values ​​corresponding to each indicator parameter; Among all the indicator parameters, the same indicator parameters are associated to obtain a set of similar indicators, and the coefficient of variation among the indicator values ​​corresponding to all the indicator parameters in the set of similar indicators is calculated. The first indicator weight corresponding to each indicator parameter is determined based on the magnitude of the discrete coefficient.

6. The method according to claim 4, characterized in that, The calculation method for the weight of the second indicator includes: Obtain historical indicator datasets corresponding to multiple historical project phases, and ESG scores corresponding to each historical project phase, wherein the historical indicator datasets include multiple indicator parameters and indicator values ​​corresponding to each indicator parameter; The historical indicator dataset and ESG scores corresponding to each historical project stage are input into a pre-trained ensemble learning model to determine the contribution of the indicator value corresponding to each indicator parameter to the ESG score. Based on the contribution of the indicator value corresponding to the indicator parameter to the ESG score, the weight of the second indicator corresponding to the indicator parameter is determined.

7. The method according to claim 1, characterized in that, After obtaining the ESG score of the new energy project by weighted summation of the index values ​​of all the index parameters based on the fusion index weight corresponding to each index parameter, the method further includes: Obtain the ESG scores corresponding to multiple consecutive project phases, and smooth the multiple ESG scores based on the Kalman filter algorithm to obtain multiple smoothed ESG scores. Each ESG score and the corresponding project stage are input into a pre-trained second long short-term memory network model to obtain the ESG score corresponding to the next project stage of the new energy project output by the second long short-term memory network model.

8. A device for calculating the ESG score of a new energy project, characterized in that, The device includes: The data acquisition module is used to determine the current project stage of the new energy project, multiple indicator parameters of the new energy project at the project stage, and the indicator value corresponding to each indicator parameter. The weight acquisition module is used to acquire the first indicator weight corresponding to each of the indicator parameters calculated in advance based on the entropy method, and to acquire the second indicator weight corresponding to each of the indicator parameters determined in advance based on the machine learning model. The parameter acquisition module is used to acquire multiple consecutive historical project stages and the parameter group corresponding to each historical project stage, wherein the parameter group includes multiple historical indicator parameters and the historical weight value corresponding to each historical indicator parameter. The model calculation module is used to input the historical project stage, the parameter group corresponding to each historical project stage, the project stage, and each of the indicator parameters into the pre-trained first long short-term memory network model to obtain the third indicator weight of each indicator parameter. The weight fusion module is used to fuse the first indicator weight, the second indicator weight, and the third indicator weight corresponding to each indicator parameter to obtain the fused indicator weight for each indicator parameter. The scoring calculation module is used to perform a weighted summation of the index values ​​of all the index parameters based on the fusion index weight corresponding to each index parameter, so as to obtain the ESG score of the new energy project.

9. A 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 7.

10. A computer, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, 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 7.

Images