Power grid project selection method and system based on artificial intelligence

Abstract

This invention discloses an artificial intelligence-based method and system for selecting power grid projects, relating to the field of artificial intelligence technology. The method includes: collecting and preprocessing raw data to construct a dual dataset; constructing a risk interval model, taking the proposed power grid project dataset as input and outputting a risk interval; verifying the risk interval by setting validation conditions; calculating the revenue stability index, environmental impact score, and transmission reliability; and concatenating these with the risk interval to obtain an input vector; constructing a prediction model, taking the input vector as input and outputting a single-point benefit function value; calculating the comprehensive benefit score and benefit-cost ratio; obtaining decision variables based on binary value conversion; defining an objective function; constructing constraints; and optimizing to generate the optimal power grid project selection combination as the final basis for power grid project selection. This invention significantly improves the scientific rigor, robustness, and real-time performance of the power grid project selection process.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method and system for selecting power grid projects based on artificial intelligence. Background Technology

[0002] In recent years, with the large-scale integration of renewable energy, the implementation of electricity market spot trading mechanisms, and the significant impact of climate fluctuations on electricity load, the uncertainty and complexity of power grid project investment have been continuously increasing. How to scientifically select projects based on multi-dimensional indicators (economic benefits, power supply reliability, environmental impact, etc.) has become a critical issue that power grid companies urgently need to address. At the same time, the application of artificial intelligence methods in power forecasting, market risk modeling, and operation optimization is becoming increasingly mature, providing new tools to support the resolution of the complexity and uncertainty in power grid project selection.

[0003] However, existing AI-based methods for selecting power grid projects fail to provide a risk range for future project returns, leading to significant decision-making biases in extreme cases. Furthermore, these methods largely rely on numerical integration and heuristic search, resulting in high computational complexity and poor real-time performance. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides an artificial intelligence-based method and system for selecting power grid projects, which solves the problems of existing artificial intelligence-based power grid project selection methods failing to provide the risk range of future project returns, resulting in large decision-making biases in extreme cases, and relying heavily on numerical integration and heuristic search, leading to high computational complexity and poor real-time performance.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides an artificial intelligence-based method for selecting power grid projects, comprising,

[0008] Collect raw data for preprocessing and construct dual datasets, including a dataset of the proposed power grid project and a historical dataset;

[0009] The raw data includes proposed power grid projects and historical power grid operation data;

[0010] A risk interval model is constructed, which takes the dataset of the proposed power grid project as input and outputs the risk interval. After verifying the risk interval by setting verification conditions, the revenue stability index, environmental impact score, and power transmission reliability are calculated. The risk interval is then concatenated to obtain the input vector.

[0011] A predictive model is constructed, taking the input vector as input and outputting the single-point benefit function value. The comprehensive benefit score and benefit-cost ratio are calculated. Decision variables are obtained based on binary value conversion. The objective function is defined, and constraints are constructed. The optimal combination of power grid projects is optimized and generated as the final basis for power grid project selection.

[0012] The final basis is stored in a database.

[0013] As a preferred embodiment of the AI-based power grid project selection method of the present invention, wherein: the construction of the risk interval model takes the proposed power grid project dataset as input and the output risk interval refers to the construction of the risk interval model using a boosting regression tree model architecture;

[0014] Based on historical datasets, linear regression is used to obtain predicted market settlement returns. A quantile value is set, and a quantile regression loss function is constructed by combining the predicted market settlement returns.

[0015] The risk interval model is trained using historical datasets. During training, the loss function value is calculated, and the model parameters are iteratively optimized using the gradient boosting algorithm. In each iteration, the prediction residual between the actual value and the loss value is calculated. A regression tree is fitted using the prediction residual, and the minimum quantile regression loss function value is used as the leaf node value in the regression tree. The prediction value is then updated using the learning rate. During the update process, the iteration stops when the loss value of the loss function no longer decreases significantly, and the final risk interval model is output.

[0016] The dataset of the proposed power grid project is input into the final risk interval model, and the lower limit, median and upper limit of the output revenue are combined to form the risk interval.

[0017] As a preferred embodiment of the AI-based power grid project selection method of the present invention, the following steps are taken: after setting verification conditions to verify the risk interval, the revenue stability index, environmental impact score, and transmission reliability are calculated, and the input vector is obtained by splicing the risk interval. The Delphi method is used to set verification conditions, and the values ​​within the risk interval are verified by the verification conditions. When the risk interval meets the verification conditions, the risk interval is recorded and the revenue stability index is calculated. Otherwise, the project data is marked as abnormal, and the proposed power grid project dataset is re-entered to obtain the risk interval.

[0018] The project approval documents for the proposed power grid project are obtained from the power grid enterprise's project approval system through the API interface, and the corresponding environmental impact report is manually searched based on the project approval documents.

[0019] Based on the environmental impact report, obtain carbon emissions and ecological disturbance area, and calculate the environmental impact score;

[0020] Using the power system simulation software PSCAD, combined with standard meteorological conditions and the design parameters of the power grid in the proposed power grid project, a power grid simulation model is generated to perform power grid simulation operations, and the average power outage time of the power grid in the proposed power grid project is obtained. The operation is repeated to obtain the average power outage time of each proposed power grid project, and the maximum value is taken to calculate the power transmission reliability of the power grid in each proposed power grid project.

[0021] The lower and upper limits of the revenue stability index, environmental impact score, power transmission reliability, and risk range are normalized and concatenated to generate an input vector for each power grid project, and the dimensions are recorded.

[0022] As a preferred embodiment of the AI-based power grid project selection method of the present invention, the prediction model is constructed by using an input vector as input, outputting a single-point benefit function value, and calculating a comprehensive benefit score and a benefit-cost ratio using a single hidden layer neural network architecture.

[0023] The input vector is input into the prediction model, and the prediction model predicts the input vector to obtain the single-point benefit function value of each proposed power grid project. The single-point benefit function value is converted into integral form using the analytical integration method to obtain the analytical integration formula. Combined with the recorded input vector dimension, the analytical integration formula is extended to generate a closed-form solution formula, and the comprehensive benefit score of the proposed power grid project is calculated.

[0024] Based on the comprehensive benefit score, the benefit-cost ratio between the comprehensive benefit score and the construction cost of the corresponding proposed power grid project is calculated using a proportional formula.

[0025] As a preferred embodiment of the artificial intelligence-based power grid project selection method of the present invention, wherein: the step of obtaining decision variables based on binary value conversion, defining an objective function, constructing constraints, and optimizing the generation of the optimal power grid project selection combination as the final basis for power grid project selection refers to setting a judgment threshold, combining the benefit-cost ratio, generating the binary value of each proposed power grid project, and defining it as the decision variable of each proposed power grid project;

[0026] Based on all decision variables with a result of 1, a decision vector is generated. Combined with the comprehensive benefit score, an objective function is defined to maximize the value of the objective function.

[0027] Construct constraints based on decision variables;

[0028] The comprehensive benefit scores corresponding to all decision variables in the decision vector of the objective function are concatenated and defined as a gradient vector. The objective function value is iteratively updated using the projection gradient method. In each iteration, the inertia coefficient is set using the accelerated gradient method, and the decision vector is updated according to the inertia coefficient. The updated decision vector is then projected back to the feasible region defined by the constraint conditions using the projection operator. Each decision variable in the projected decision vector is then truncated component by component to generate truncated decision variables.

[0029] Based on the truncated decision variables, it is determined whether the constraints are met. If the constraints are not met, the Lagrange multipliers are calculated and the truncated decision variables are corrected according to the Lagrange multipliers. Otherwise, the iteration continues until the maximum number of iterations is reached, and the optimal power grid project selection combination is output.

[0030] The optimal combination of power grid projects will be used as the basis for selecting power grid projects.

[0031] As a preferred embodiment of the AI-based power grid project selection method of the present invention, the step of storing the final basis in the database refers to compressing the final basis using a compression algorithm, forming a compressed file, adding a project number, and then storing it in a MySQL database.

[0032] As a preferred embodiment of the AI-based power grid project selection method of the present invention, the step of collecting raw data and preprocessing it to construct a dual dataset refers to obtaining the power grid long-term development plan document, feasibility study report, GIS file and feasibility study report from the power grid dispatch center through an API interface;

[0033] Based on the long-term development plan documents for the power grid, the proposed power grid projects were obtained, and text analysis tools were used to extract the design parameters of the proposed power grid projects from the feasibility study reports and GIS documents, including voltage level, conductor cross-sectional area, number of circuits, and line length.

[0034] The construction cost of the proposed power grid project is obtained from the feasibility study report; the future forecast meteorological data of the proposed power grid project is obtained by calling the API of the China Meteorological Administration; and the forecast electricity price is obtained through the power trading center.

[0035] Historical power grid operation data is collected from the power grid dispatch center via API interface, and historical electricity prices and market settlement revenues corresponding to the historical power grid operation data are obtained from the power trading center.

[0036] The historical power grid operation data includes daily maximum load, power factor, transmission loss, line operating voltage, actual power generation, and planned power generation.

[0037] Linear interpolation, outlier detection, and normalization are performed on all data to generate datasets and historical datasets for each proposed power grid project.

[0038] Secondly, the present invention provides an artificial intelligence-based power grid project selection system, comprising,

[0039] The processing and splicing module is used to collect raw data for preprocessing, construct dual datasets, build a risk interval model, form risk intervals for verification, calculate the revenue stability index, environmental impact score, and power transmission reliability, and splice them together with the risk intervals to obtain the input vector.

[0040] The prediction transformation module is used to build prediction models, output single-point benefit function values, calculate comprehensive benefit scores and benefit-cost ratios, and obtain decision variables based on binary value transformation.

[0041] The optimization generation module is used to define the objective function and construct constraints, and optimize and generate the optimal combination of power grid projects as the final basis for power grid project selection;

[0042] The storage module is used to store the final data through a database.

[0043] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the artificial intelligence-based power grid project selection method as described in the first aspect of the present invention.

[0044] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the artificial intelligence-based power grid project selection method as described in the first aspect of the present invention.

[0045] The beneficial effects of this invention are as follows: by combining the quantile regression loss function, prediction model, analytical integral method, and projection gradient method with accelerated gradient update and Lagrange multiplier correction, this invention can not only output the upper and lower limits and median of the proposed power grid project's revenue and construct a robust risk range, but also significantly improve the scientific nature, robustness, and real-time performance of the power grid project selection process. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is a flowchart of the AI-based power grid project selection method in Example 1.

[0048] Figure 2 This is a structural diagram of the power grid project selection system based on artificial intelligence in Example 1.

[0049] Figure 3 This is a flowchart for calculating the comprehensive benefit score of the proposed power grid project in Example 1. Detailed Implementation

[0050] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0051] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0052] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0053] Example 1, referring to Figures 1-3 This is the first embodiment of the present invention, which provides an artificial intelligence-based method for selecting power grid projects, including the following steps:

[0054] S1. Collect raw data and preprocess it to construct a dual dataset, including the proposed power grid project dataset and the historical dataset;

[0055] Specifically, raw data is collected and preprocessed to construct dual datasets, including the proposed power grid project dataset and historical datasets. This involves obtaining long-term power grid development planning documents, feasibility study reports, GIS files, and feasibility study reports from the power grid dispatch center via API interfaces.

[0056] Based on the long-term development plan documents for the power grid, the proposed power grid projects were obtained, and text analysis tools were used to extract the design parameters of the proposed power grid projects from the feasibility study reports and GIS documents, including voltage level, conductor cross-sectional area, number of circuits, and line length.

[0057] The construction cost of the proposed power grid project is obtained from the feasibility study report; the future forecast meteorological data of the proposed power grid project is obtained by calling the API of the China Meteorological Administration; and the forecast electricity price is obtained through the power trading center.

[0058] Historical power grid operation data is collected from the power grid dispatch center via API interface, and historical electricity prices and market settlement revenues corresponding to the historical power grid operation data are obtained from the power trading center.

[0059] The historical power grid operation data includes daily maximum load, power factor, transmission loss, line operating voltage, actual power generation, and planned power generation.

[0060] Linear interpolation, outlier detection, and normalization are performed on all data to generate datasets and historical datasets for each proposed power grid project.

[0061] By constructing a dual-dataset framework (a proposed power grid project dataset and a historical dataset), the proposed project dataset provides parameter input for future project predictions, while the historical dataset provides real samples for model training. Furthermore, by combining text analysis, GIS parsing, and cross-domain API calls, the barriers between engineering parameters, environmental variables, and market signals are broken down. In the data preprocessing stage, multiple strategies, including linear interpolation, anomaly detection, and normalization, are integrated to effectively improve data quality.

[0062] S2. Construct a risk interval model, take the dataset of the proposed power grid project as input, output the risk interval, set verification conditions to verify the risk interval, calculate the revenue stability index, environmental impact score, and power transmission reliability, and combine them with the risk interval to obtain the input vector.

[0063] Specifically, a risk interval model is constructed by taking the dataset of the proposed power grid project as input and outputting the risk interval using a boosting regression tree model architecture.

[0064] Based on historical datasets, linear regression is used to predict market settlement returns. Based on experiments and relevant knowledge, quantile values ​​are set, and a quantile regression loss function is constructed by combining the predicted market settlement returns. The formula is as follows:

[0065]

[0066] In the formula, This represents the quantile regression loss function value. Indicates the first Sample Market settlement revenue, Indicates the first Sample The predicted value (predicted market settlement revenue). Indicates quantile values ​​(which can be set through experiments and relevant knowledge);

[0067] The risk interval model is trained using historical datasets. During training, the loss function value is calculated, and the model parameters are iteratively optimized using the gradient boosting algorithm. In each iteration, the prediction residual between the actual value (market settlement return) and the loss value is calculated. A regression tree is fitted using the prediction residual, and the minimum quantile regression loss function value is used as the leaf node value in the regression tree. The prediction value is then updated using the learning rate. During the update process, the iteration stops when the loss value of the loss function no longer decreases significantly, and the final risk interval model is output.

[0068] The formula for calculating the prediction residual between the actual value and the loss value is as follows:

[0069]

[0070] In the formula, Indicates the first In the first iteration Sample The predicted residuals Indicates the first In the first iteration Sample The predicted value;

[0071] The formula for updating the predicted value using the learning rate is:

[0072]

[0073] In the formula, Indicates the first In the first iteration Sample The predicted value, The learning rate is represented (set based on experiments and domain knowledge). Indicates the first In the first iteration Sample The regression tree;

[0074] The dataset of the proposed power grid project is input into the final risk interval model, and the lower limit, median and upper limit of the output revenue are combined to form the risk interval.

[0075] Traditional boosting regression trees use only mean squared error as the objective function, failing to distinguish between upper and lower deviations in the prediction interval. In contrast, the quantile regression loss function proposed in this invention combines with tree node residual fitting, enabling the risk interval model to dynamically learn the revenue distribution characteristics at different quantiles during training. This design, through a linkage mechanism of "residual—quantile—leaf node value," effectively mitigates the impact of extreme high loads or soaring electricity prices in power grid data on the results, ensuring the stability of predictions under atypical conditions.

[0076] Furthermore, most existing models remain at the level of mean prediction. Even when confidence intervals are introduced, they are mostly passive derivations based on the assumption of normal distribution, which cannot adapt to the reality of highly asymmetric electricity prices and load distribution. In contrast, this invention directly outputs the "lower limit - median - upper limit" results and combines them into a complete risk interval, breaking through the limitations of statistical assumptions. Moreover, the output risk interval can intuitively reflect the distribution of returns under different market conditions, avoiding overly optimistic or conservative investments guided by a single numerical value.

[0077] Secondly, by deeply coupling the quantile regression loss function with the improved regression tree iteration mechanism, this invention ensures that it will not fall into the mean trap when approximating the true return distribution, thereby achieving complete capture of risk uncertainty.

[0078] Furthermore, after setting verification conditions to verify the risk interval, the revenue stability index, environmental impact score, and transmission reliability are calculated. The input vector is obtained by concatenating the risk interval. The Delphi method is used to set verification conditions, and the values ​​within the risk interval (these values ​​refer to the lower limit, median, and upper limit) are verified through the verification conditions. When the risk interval meets the verification conditions, the risk interval is recorded, and the revenue stability index is calculated. Otherwise, the project data is marked as abnormal, and the risk interval is re-entered into the proposed power grid project dataset.

[0079] The Delphi method is used to set the verification conditions, and the formula is as follows:

[0080]

[0081] In the formula, This represents the lower limit of the risk range. This represents the median value within the risk range. This represents the upper limit of the risk range;

[0082] The return stability index is calculated using the following formula:

[0083]

[0084] In the formula, Indicates the first The revenue stability index of a proposed power grid project;

[0085] The project approval documents for the proposed power grid project are obtained from the power grid enterprise's project approval system through the API interface, and the corresponding environmental impact report is manually searched based on the project approval documents.

[0086] Based on the carbon emissions and ecological disturbance area obtained from the environmental impact report, the environmental impact score is calculated using the following formula:

[0087]

[0088] In the formula, Indicates the environmental impact score. The conversion factor can be obtained from the guide to carbon emission accounting in power construction. This indicates the maximum carbon emissions (which can be obtained from the environmental impact report). This indicates the area of ​​ecological disturbance (which can be obtained from the environmental impact report). Indicates the maximum area of ​​ecological disturbance (which can be obtained from the environmental impact report);

[0089] Using the power system simulation software PSCAD, combined with standard meteorological conditions and the design parameters of the power grid in the proposed power grid project, a power grid simulation model was generated for power grid simulation operations. The average outage time of the power grid in the proposed power grid project was obtained. This process was repeated to obtain the average outage time for each proposed power grid project. The maximum value was then used to calculate the transmission reliability of the power grid in each proposed power grid project. The formula is as follows:

[0090]

[0091] In the formula, Indicates the first The transmission reliability of the proposed power grid project Indicates the first Average outage time for each proposed power grid project Indicates the maximum average power outage time;

[0092] The lower and upper limits of the revenue stability index, environmental impact score, power transmission reliability, and risk range are normalized and concatenated to generate an input vector for each power grid project, and the dimensions are recorded.

[0093] By setting verification conditions using the Delphi method, the rationality and interpretability of risk range judgment are improved. Processing the range data yields a revenue stability index, which can intuitively quantify the project's robustness in future revenue. This not only provides a straightforward risk assessment tool for power grid projects but also facilitates horizontal comparisons between different projects in terms of stability. Furthermore, the average outage time obtained through PSCAD simulation serves as a direct indicator, directly linking it to the power grid service level and enhancing the indicator's practicality. Concatenating the revenue stability index, environmental impact score, transmission reliability, and the lower and upper limits of the risk range into an input vector provides a unified input data structure for subsequent steps.

[0094] S3. Construct a prediction model, take the input vector as input, output the single-point benefit function value, calculate the comprehensive benefit score and benefit-cost ratio, obtain the decision variables based on binary value conversion, define the objective function, construct constraints, and optimize to generate the optimal power grid project selection combination as the final basis for power grid project selection.

[0095] Specifically, a prediction model is constructed, which takes the input vector as input, outputs a single-point benefit function value, and calculates the comprehensive benefit score and benefit-cost ratio. The prediction model is constructed using a single hidden layer neural network (SHL-NN) architecture, which includes an input layer, a hidden layer, and an output layer.

[0096] The hidden layer contains neurons;

[0097] The mean squared error is defined as the loss function, and the Adam optimizer is used to iteratively optimize the parameter combination of the prediction model. During the iteration process, when the loss value of the loss function no longer decreases significantly, the iteration stops and the final prediction model is output.

[0098] The input vector is input into the prediction model, and the prediction model predicts the input vector to obtain the single-point benefit function value of each proposed power grid project. The single-point benefit function value is converted into integral form using the analytical integration method to obtain the analytical integration formula. Combined with the recorded input vector dimension, the analytical integration formula is extended to generate a closed-form solution formula, and the comprehensive benefit score of the proposed power grid project is calculated.

[0099] The prediction of the input vector using the prediction model is expressed by the following formula:

[0100]

[0101] In the formula, Represents the input vector The single-point benefit function value, This represents the bias term of the output layer. This represents the total number of neurons (set via cross-validation). Indicates the first hidden layer The weight vector from each neuron to the output layer This represents the Sigmoid activation function. Indicates the first hidden layer Bias terms for each neuron, This indicates the nth time in the input layer to the hidden layer. The weight vector of each neuron. Represents the input vector;

[0102] The analytical integration method is used to convert the integral into integral form, resulting in the analytical integration formula, which is:

[0103]

[0104] In the formula, Represents the input vector The weighting coefficients (can be set through experiments and relevant knowledge content). This represents the bias term of the Sigmoid function. Represents the derivative of the input vector. Represents the natural constant. Represents the natural logarithm function;

[0105] The formula for generating the closed-form solution and calculating the comprehensive benefit score of the proposed power grid project is as follows:

[0106]

[0107] In the formula, Indicates the first A comprehensive benefit score for each proposed power grid project. This represents the total dimension of the input vector. Indicates the first The dimension is passed to the hidden layer. The weight vector of each neuron. Represents the natural logarithm function. Represents the natural constant. Indicates the first The upper bound of the integral in each dimension (can be set through industry standards and relevant knowledge). Indicates the first The lower bound of the integral in each dimension (which can be set through industry standards and relevant knowledge);

[0108] Based on the comprehensive benefit score, the benefit-cost ratio between the comprehensive benefit score and the construction cost of the corresponding proposed power grid project is calculated using a proportional formula.

[0109] By employing a three-dimensional coupling evaluation mechanism of "neural network prediction—analytic integral transformation—closed-loop benefit decoupling," this invention achieves a paradigm shift from "black-box prediction" to "white-box decision support." Existing technologies often employ numerical integration or heuristic weighted summation to calculate project benefits, methods that are computationally intensive and lack analytical expressions. This invention, however, directly transforms the single-point benefit function value predicted by a single-hidden-layer neural network into an analytical integral form, and further extends it into a closed-loop solution formula, achieving overall benefit evaluation of multi-dimensional input vectors. Moreover, utilizing the integral analytical property of the sigmoid function, it transforms it into a closed-loop logarithmic function, ensuring analytical solvability even under multi-dimensional input conditions. This operational step differs from traditional neural networks being used merely as black-box prediction methods; it fundamentally changes the usage mode of neural networks.

[0110] Secondly, while numerical integration or Monte Carlo simulation can perform similar benefit integration calculations, their computational complexity increases exponentially with dimensionality, failing to meet the real-time requirements of power grid project selection. The closed-form solution of this invention, however, maintains computational efficiency unaffected by the number of samples when the input dimension changes, making it suitable for real-time computation. Furthermore, it can capture the nonlinear coupling relationships between multidimensional features, thus making the selection results more consistent with actual power grid project selection.

[0111] Furthermore, based on the binary value conversion to obtain decision variables, an objective function is defined, and constraints are constructed to optimize and generate the optimal combination of power grid projects as the final basis for power grid project selection. This involves setting a judgment threshold based on experiments and current domain knowledge, combining the benefit-cost ratio, generating the binary value of each proposed power grid project, and defining it as the decision variable for each proposed power grid project.

[0112] The binary value for each proposed power grid project is generated by combining the benefit-cost ratio, using the following formula:

[0113]

[0114] In the formula, Indicates the first Decision variables for a proposed power grid project Indicates the first Benefit-cost ratio of a proposed power grid project Indicates the threshold for judgment;

[0115] Based on all decision variables with an outcome of 1, a decision vector is generated. Combining this with the comprehensive benefit score, an objective function is defined. The objective function value is maximized using the following formula:

[0116]

[0117] In the formula, Represents the decision vector The objective function value, This indicates the total number of planned power grid projects;

[0118] Based on the decision variables, the constraints are constructed as follows:

[0119]

[0120] In the formula, Indicates the first Construction costs of a proposed power grid project This indicates the total budget (which can be obtained from the finance department).

[0121] The comprehensive benefit scores corresponding to all decision variables in the decision vector of the objective function are concatenated and defined as a gradient vector. The objective function value is iteratively updated using the projection gradient method. In each iteration, the inertia coefficient is set using the accelerated gradient method, and the decision vector is updated according to the inertia coefficient. The updated decision vector is then projected back to the feasible region defined by the constraint conditions using the projection operator. Each decision variable in the projected decision vector is then truncated component by component to generate truncated decision variables.

[0122] The inertia coefficient is set using the acceleration gradient method, and the formula is as follows:

[0123]

[0124]

[0125] In the formula, Represents the coefficient of inertia. Indicates the first Round iteration;

[0126] The decision vector is then updated based on the inertia coefficient, using the following formula:

[0127]

[0128] In the formula, Indicates the first The decision vector updated during each iteration. This indicates that the updated solution is guaranteed to fall within the feasible region. Projection operator within, Indicates the first The decision vector during each iteration. Indicates the step size (set via backtracking line search). Indicates the first Decision vector during round iteration The gradient vector of the objective function value. Represents the coefficient of inertia. Indicates the first The decision vector during round iteration;

[0129] The next step involves truncating each decision variable within the projected decision vector component-by-component, as shown in the formula:

[0130]

[0131] In the formula, Indicates the first Cutoff decision variables for a proposed power grid project This indicates the operation of finding the minimum value. This indicates the operation of retrieving the maximum value;

[0132] Based on the truncated decision variables, it is determined whether the constraints are met. If the constraints are not met, the Lagrange multipliers are calculated and the truncated decision variables are corrected according to the Lagrange multipliers. Otherwise, the iteration continues until the maximum number of iterations is reached, and the optimal power grid project selection combination is output.

[0133] The formula for calculating the Lagrange multipliers is as follows:

[0134]

[0135] In the formula, Represents the Lagrange multipliers;

[0136] The truncated decision variables are then modified according to the Lagrange multipliers, as shown in the formula:

[0137]

[0138] In the formula, Indicates the first Corrected decision variables for a proposed power grid project;

[0139] The optimal combination of power grid projects will be used as the basis for selecting power grid projects.

[0140] By converting the benefit-cost ratio into a binary variable through a threshold decision, the problem size is significantly reduced, and the solution process is simplified. For example, using continuous variables (such as fractional selection) can lead to spurious solutions of "incomplete construction," which is impractical in the actual selection of power grid projects. Furthermore, the objective function focuses on maximizing the comprehensive benefit score, with the constraint that construction costs do not exceed the budget. This design balances investment benefits with financial constraints, ensuring the economic feasibility and practicality of the selected power grid projects. Secondly, the inertia coefficient is determined using the accelerated gradient method to improve the iteration speed, and then the solution is projected onto the feasible region using a projection operator, ensuring that constraints such as the budget are always satisfied. This operation achieves rapid convergence in complex multidimensional search spaces, avoiding getting stuck in infeasible solutions and repeated "solution-correction-re-optimization." Moreover, component-by-component truncation of the projected solution avoids unreasonable "partial selection" situations in practical applications, ensuring that the output matches the actual project decisions.

[0141] When the solution fails to meet the constraints, Lagrange multipliers are calculated and introduced to correct the variables, achieving a dynamic balance between the objective function and the constraints. This method avoids the loss of optimality caused by hard truncation, ensuring that the result remains close to the global optimum even under stressful constraints. This not only enhances the robustness and applicability of the invention but also allows it to cope with budget adjustments and changes in project conditions. Finally, the optimal power grid project combination is output. This result can be directly used as the basis for power grid project investment and construction decisions, avoiding the uncertainty and subjectivity of manual evaluation.

[0142] S4. Store the final data in a database;

[0143] Specifically, storing the final data in the database means using a compression algorithm to compress the final data, forming a compressed file, adding a project number, and then storing it in a MySQL database.

[0144] By compressing the "final evidence" and storing it in a MySQL database, not only can storage efficiency be significantly improved, but the integrity and traceability of the data can also be guaranteed.

[0145] This embodiment also provides an artificial intelligence-based power grid project selection system, including:

[0146] The processing and splicing module is used to collect raw data for preprocessing, construct dual datasets, build a risk interval model, form risk intervals for verification, calculate the revenue stability index, environmental impact score, and power transmission reliability, and splice them together with the risk intervals to obtain the input vector.

[0147] The prediction transformation module is used to build prediction models, output single-point benefit function values, calculate comprehensive benefit scores and benefit-cost ratios, and obtain decision variables based on binary value transformation.

[0148] The optimization generation module is used to define the objective function and construct constraints, and optimize and generate the optimal combination of power grid projects as the final basis for power grid project selection;

[0149] The storage module is used to store the final data through a database.

[0150] This embodiment also provides a computer device applicable to the power grid project selection method based on artificial intelligence, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the power grid project selection method based on artificial intelligence as proposed in the above embodiment.

[0151] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0152] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the AI-based power grid project selection method proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0153] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An artificial intelligence-based power grid project selection method, characterized in that: The method comprises the following steps: The original data is collected and preprocessed to construct a double data set, including a data set of the proposed power grid project and a historical data set; The original data includes the proposed power grid project and historical power grid operation data; A risk interval model is constructed, the data set of the proposed power grid project is taken as input, and a risk interval is output. After the risk interval is verified by setting a verification condition, the benefit stability index, the environmental impact score, and the power transmission reliability are calculated, and the input vector is obtained by splicing the risk interval; A prediction model is constructed, the input vector is taken as input, and a single-point benefit function value is output. The comprehensive benefit score and the benefit-cost ratio are calculated using a single-hidden-layer neural network architecture to construct the prediction model; The input vector is input into the prediction model, the input vector is predicted by the prediction model, and a single-point benefit function value of each proposed power grid project is obtained. The single-point benefit function value is converted into an integral form using the analytical integration method, an analytical integral formula is obtained, the analytical integral formula is expanded in combination with the recorded input vector dimension, a closed-form solution formula is generated, and the comprehensive benefit score of the proposed power grid project is calculated; Based on the comprehensive benefit score, the benefit-cost ratio between the comprehensive benefit score and the corresponding construction cost of the proposed power grid project is calculated using a proportional formula; Based on the binary value conversion, a decision variable is obtained, a target function is defined, and a constraint condition is constructed to generate an optimal power grid project selection combination as the final basis for selecting the power grid project. A decision threshold is set, and the binary value of each proposed power grid project is generated in combination with the benefit-cost ratio, which is defined as the decision variable of each proposed power grid project; The binary value of each proposed power grid project is generated in combination with the benefit-cost ratio, and the formula is: wherein, denotes the decision variable of the th planned power grid project, denotes the benefit-cost ratio of the th planned power grid project, denotes a decision threshold value; Based on all the decision variables with the result of 1, a decision vector is generated, a target function is defined in combination with the comprehensive benefit score, and the target function value is maximized; Based on the decision variable, a constraint condition is constructed; All the comprehensive benefit scores corresponding to the decision variables in the decision vector in the target function are spliced to define a gradient vector. The target function value is iteratively updated using the projection gradient method. In each iteration, an inertia coefficient is set using the accelerated gradient method, and the decision vector is updated according to the inertia coefficient. The updated decision vector is projected back to the feasible region defined by the constraint condition, and then each decision variable in the projected decision vector is truncated to generate a truncated decision variable; Based on the truncated decision variable, it is judged whether the constraint condition is met. When the constraint condition is not met, the Lagrange multiplier is calculated, and the truncated decision variable is modified according to the Lagrange multiplier. Otherwise, the iteration continues until the maximum number of iterations is reached, and the optimal power grid project selection combination is output. The optimal power grid project selection combination is used as the basis for selecting the power grid project; The final basis is stored in the database.

2. The artificial intelligence-based power grid project selection method of claim 1, wherein: The risk interval model is constructed using a boosting regression tree model architecture, taking the data set of the proposed power grid project as input and outputting a risk interval; Based on the historical data set, the predicted market settlement benefit is obtained using linear regression. A quantile point value is set, and a quantile regression loss function is constructed in combination with the predicted market settlement benefit. The training of the risk interval model is performed using a historical data set, during the training, a loss function value is calculated, and an iterative optimization of the model parameters is performed using a gradient boosting algorithm, in each iteration, a prediction residual between an actual value and a loss value is calculated, a regression tree is fitted using the prediction residual, and a minimum quantile regression loss function value is used as a leaf node value in the regression tree, and then the prediction value is updated by a learning rate, during the updating process, when the loss value of the loss function no longer significantly decreases, the iteration is stopped, and the final risk interval model is output; The data set of the proposed power grid project is input into the final risk interval model, and the lower limit value, the median value and the upper limit value of the income are combined to form the risk interval.

3. The artificial intelligence-based power grid project selection method of claim 2, wherein: The set verification condition is used to verify the risk interval, and the income stability index, the environmental impact score and the power transmission reliability are calculated, and the input vector is obtained by splicing the risk interval; the verification condition is set by using the Delphi method, and the values in the risk interval are verified by the verification condition, when the risk interval meets the verification condition, the risk interval is recorded, and the income stability index is calculated, otherwise, the project data is marked as abnormal, and the risk interval is obtained by inputting the proposed power grid project data set again; The project approval of the proposed power grid project is obtained from the grid enterprise project approval system through the API interface, and the corresponding environmental impact report is manually found based on the project approval; The carbon emission and the ecological disturbance area are obtained from the environmental impact report, and the environmental impact score is calculated; The power system simulation software PSCAD is used, combined with the standard meteorological condition and the design parameter of the power grid in the proposed power grid project, to generate a power grid simulation model for power grid simulation operation, to obtain the average outage time of the power grid in the proposed power grid project, and the maximum value is taken to calculate the power transmission reliability of each proposed power grid project; The income stability index, the environmental impact score, the power transmission reliability and the lower limit value and the upper limit value in the risk interval are normalized and spliced to generate the input vector of each power grid project, and the dimension is recorded.

4. The artificial intelligence-based power grid project selection method of claim 3, wherein: The final basis is compressed by using a compression algorithm, a compressed file is formed, a project number is added, and then stored in the MySQL database.

5. The artificial intelligence-based power grid project selection method of claim 4, wherein: The original data is collected and preprocessed to construct a double data set, which is obtained from the power grid dispatching center through the API interface, including the long-term development planning file, the feasibility report, the GIS file and the feasibility report; The proposed power grid project is obtained from the long-term development planning file, and the design parameters of the proposed power grid project are extracted from the feasibility report and the GIS file by using a text analysis tool, including voltage level, conductor cross-sectional area, loop number and line length; The construction cost of the proposed power grid project is obtained from the feasibility report, the future predicted meteorological data of the proposed power grid project is obtained by calling the API of the China Meteorological Administration, and the predicted electricity price is obtained from the power trading center; The historical power grid operation data is collected from the power grid dispatching center through the API interface, and the historical electricity price and market settlement income corresponding to the historical power grid operation data are obtained from the power trading center; The historical power grid operation data includes daily maximum load value, power factor, power transmission loss, line operation voltage, actual power generation and planned power generation; Linear interpolation, outlier detection and normalization are performed on all data to generate each proposed power grid project data set and historical data set.

6. The power grid project selection system based on artificial intelligence according to any one of claims 1-5, characterized in that: It comprises, The processing splicing module is used for collecting raw data for preprocessing, constructing double data sets, constructing a risk interval model, forming a risk interval for verification, calculating a yield stability index, an environmental impact score and a power transmission reliability, and splicing the input vector in combination with the risk interval; The prediction conversion module is used for constructing a prediction model, outputting a single-point benefit function value, calculating a comprehensive benefit score and a benefit-cost ratio, and converting a binary value to obtain a decision variable; The optimization generation module is used for defining an objective function and constructing a constraint condition to optimize and generate an optimal power grid project selection combination as the final basis for selecting a power grid project; The storage module is used for storing the final basis in a database. 7.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: The processor executes the computer program to implement the steps of the power grid project selection method based on artificial intelligence according to any one of claims 1-5.

8. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to implement the steps of the power grid project selection method based on artificial intelligence according to any one of claims 1-5.

Images