Power grid asset cost analysis method, system and storage medium based on neural network

Abstract

The present invention provides a method, system and storage medium for analyzing the cost of power grid assets based on a neural network, and relates to the technical field of power grid asset management. The method comprises: obtaining basic data and external factor data of each asset in the power grid within the current preset time period, and establishing a relationship to obtain the correlation coefficient of the asset; then performing feature extraction to obtain multiple internal features and multiple external features of the asset; and then obtaining the current cost data of the asset; performing weight distribution according to the correlation coefficient of the asset to obtain the weight values ​​corresponding to each of them; and inputting a neural network model to obtain the initial predicted cost data of the asset in the preset future time period; and obtaining the final predicted cost data of the asset according to the initial predicted cost data combined with the current cost data. The present invention effectively solves the problem that the traditional model has insufficient generalization ability in power grid asset cost analysis and cannot adapt to dynamic changes in the environment and differences in regional characteristics, and improves the accuracy and reliability of the prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid asset management, and in particular to a power grid asset cost analysis method, system and storage medium based on a neural network. Background Art

[0002] At present, the cost analysis methods of power grid assets mostly use traditional statistical and econometric models. Under the specific background of stable data, clear model assumptions, and consistent environmental conditions, cost forecasting analysis of stable assets can provide relatively reliable forecast results, which is convenient for the subsequent generation of control strategies based on the cost forecast results.

[0003] In the existing technology, traditional statistical and econometric models often show insufficient generalization ability when faced with power grid systems in different regions and of different sizes and relatively variable power grid assets. They are unable to fully capture the diversity and complexity of power grid asset data, find it difficult to adapt to the dynamic changes in the power grid operating environment, and fail to take into account the significant differences in the characteristics of power grid assets in different regions. Summary of the invention

[0004] The problem solved by the present invention is how to improve the flexibility and adaptability of power grid asset cost analysis.

[0005] In order to solve the above problems, the present invention provides a method, system and storage medium for analyzing power grid asset cost based on neural network.

[0006] The present invention provides a method for analyzing power grid asset costs based on a neural network, comprising:

[0007] Obtaining basic data and external factor data of each asset in the power grid within a current preset time period, wherein the basic data includes asset acquisition cost, operation and maintenance records, technical parameters and failure rate, and the external factor data includes geographic data of the asset within a preset range, current environmental data and asset usage data;

[0008] Establishing a relationship between the asset purchase cost, the operation and maintenance record, the technical parameters and the failure rate of each asset and the geographic data, the current environment data and the asset usage data of the asset to obtain a correlation coefficient between the basic data of the asset and the external factor data;

[0009] Extracting features from the basic data and the external factor data to obtain a plurality of internal features and a plurality of external features of the asset;

[0010] Obtaining current cost data of the asset based on the internal characteristics and the external characteristics of the asset;

[0011] According to the correlation coefficient of the asset, weights are assigned to the internal features and the external features to obtain weight values ​​corresponding to each of the internal features and the external features;

[0012] Inputting the internal features and the weight values ​​corresponding to the internal features and the external features and the weight values ​​corresponding to the external features into a neural network model to obtain initial predicted cost data of the asset in a preset future time period;

[0013] The final predicted cost data of the asset is obtained based on the initial predicted cost data and the current cost data.

[0014] Optionally, the establishing of a relationship between the asset purchase cost, the operation and maintenance record, the technical parameter and the failure rate of each asset and the geographic data, the current environment data and the asset usage data of the asset to obtain a correlation coefficient between the basic data of the asset and the external factor data includes:

[0015] Performing data analysis on the asset purchase cost, the operation and maintenance record, the technical parameter and the failure rate of the asset to obtain the change trends of the asset purchase cost, the operation and maintenance record, the technical parameter and the failure rate within the current preset time period;

[0016] quantifying the linear relationship of the change trend among the geographic data of the asset, the current environment data, and the asset usage data by a Pearson correlation coefficient;

[0017] The correlation coefficient between the basic data of the asset and the external factor data is calculated based on the linear relationship.

[0018] Optionally, the extracting features of the basic data and the external factor data to obtain a plurality of internal features and a plurality of external features of the asset includes:

[0019] Extracting features from the basic data to obtain the number of maintenance times, maintenance costs, component replacement frequency, equipment capacity, voltage level, and line length of the asset within the current preset time period, and using them as the internal features of the asset;

[0020] Feature extraction is performed on the external factor data to obtain the altitude, terrain type, geographical location, average temperature and humidity, pollution level and usage requirements of the asset within the current preset time period, and use them as the external features of the asset.

[0021] Optionally, obtaining the current cost data of the asset according to the internal characteristics and the external characteristics of the asset includes:

[0022] The number of maintenance, the repair cost, the frequency of component replacement, the equipment capacity, the voltage level, the line length, the altitude, the terrain type, the geographical location, the average temperature and humidity, the pollution level, and the usage demand of the asset are standardized to obtain standardized data of the same magnitude;

[0023] The standardized data is input into a cost prediction model to obtain the current cost data of the asset within the current preset time period.

[0024] Optionally, the weight allocation of the internal features and the external features according to the correlation coefficient of the asset to obtain the weight values ​​corresponding to each of the internal features and the external features respectively includes:

[0025] The internal features and the external features are sorted according to the correlation coefficients corresponding to the asset acquisition cost, the operation and maintenance record, the technical parameters, and the failure rate of the asset in the asset relative to the geographic data, the current environment data, and the asset usage data to obtain a correlation sequence;

[0026] According to the association sequence, the initial weights of the asset acquisition cost, the operation and maintenance record, the technical parameters, the failure rate, the geographic data, the current environmental data and the asset usage data of the asset are adjusted to obtain the weight values ​​corresponding to each of the internal characteristics and the external characteristics.

[0027] Optionally, the step of inputting the internal features and the weight values ​​corresponding to the internal features and the external features and the weight values ​​corresponding to the external features into a neural network model to obtain initial predicted cost data of the asset in a preset future time period includes:

[0028] Inputting the internal features and the weight values ​​corresponding to the internal features and the external features and the weight values ​​corresponding to the external features into a neural network model;

[0029] According to the internal feature, obtaining a feature value corresponding to the internal feature;

[0030] Multiplying the weight value corresponding to the internal feature by the feature value corresponding to the internal feature to obtain the weighted internal feature;

[0031] According to the external feature, obtaining a feature value corresponding to the external feature;

[0032] Multiplying the weight value corresponding to the external feature by the feature value corresponding to the external feature to obtain the weighted external feature;

[0033] The weighted internal features are predicted by combining the neural network model with the weighted external features to obtain the initial predicted cost data of the asset in the preset future time period.

[0034] Optionally, obtaining the final predicted cost data of the asset according to the initial predicted cost data in combination with the current cost data includes:

[0035] Comparing the initial predicted cost data with the current cost data, obtaining the cost change difference of the asset within the preset future time period;

[0036] determining a forecast error based on the cost change difference;

[0037] Assigning weights to the initial predicted cost data and the current cost data according to the prediction error to obtain weight values ​​corresponding to the initial predicted cost data and the current cost data;

[0038] The final predicted cost data of the asset is obtained by taking a weighted average based on the weight values ​​corresponding to the initial predicted cost data and the current cost data.

[0039] Optionally, the neural network-based power grid asset cost analysis method further includes:

[0040] Cost prediction is performed for each of the assets in the power grid through a distributed system, wherein the distributed system includes a plurality of nodes, and each of the nodes performs cost prediction for one of the assets.

[0041] The present invention also provides a power grid asset cost analysis system based on a neural network, comprising:

[0042] A data acquisition unit, used to acquire basic data and external factor data of each asset in the power grid within a current preset time period, wherein the basic data includes asset acquisition cost, operation and maintenance records, technical parameters and failure rate, and the external factor data includes geographic data of the asset within a preset range, current environmental data and asset usage data;

[0043] A correlation coefficient calculation unit, used to establish a relationship between the asset purchase cost, the operation and maintenance record, the technical parameters and the failure rate of each asset and the geographic data, the current environment data and the asset use data of the asset, to obtain a correlation coefficient between the basic data of the asset and the external factor data;

[0044] A feature extraction unit, used to extract features from the basic data and the external factor data to obtain a plurality of internal features and a plurality of external features of the asset;

[0045] A first data processing unit, configured to obtain current cost data of the asset according to the internal characteristics and the external characteristics of the asset;

[0046] A second data processing unit is used to assign weights to the internal features and the external features according to the correlation coefficient of the asset, and obtain weight values ​​corresponding to each of the internal features and the external features;

[0047] A prediction unit, used for inputting the internal features and the weight values ​​corresponding to the internal features and the external features and the weight values ​​corresponding to the external features into a neural network model to obtain initial predicted cost data of the asset in a preset future time period;

[0048] A correction unit is used to obtain final predicted cost data of the asset based on the initial predicted cost data in combination with the current cost data.

[0049] The present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the above-mentioned neural network-based power grid asset cost analysis method is implemented.

[0050] The neural network-based power grid asset cost analysis method, system and storage medium of the present invention, first, systematically collect basic data and external factor data of each asset, such as purchase cost, maintenance records, technical parameters, failure rate, geography, environment and usage, etc., to ensure that the data covers the whole picture of the asset. Secondly, through feature extraction and correlation coefficient calculation, the key factors affecting the cost are excavated, which not only enhances the model's ability to capture data diversity and complexity, but also provides a basis for weight allocation; the weight allocation mechanism further optimizes the influence of the features and ensures the pertinence of model training. Subsequently, the weighted features are input into the neural network model, and the initial predicted cost data of the asset in the future time period is generated by using its powerful nonlinear fitting ability. Finally, by integrating with the current cost data, adjusting and optimizing the prediction results, more accurate and reliable final predicted cost data is generated. The present invention obtains the basic data and external factor data of the power grid assets, uses feature extraction technology to obtain the internal and external features of the assets, and calculates the correlation coefficients between these features and the asset cost. Further, through weight allocation, the importance of features to cost prediction is strengthened, and the weighted features are input into the neural network model for training to obtain the initial predicted cost data of the asset. Finally, the initial forecast results are adjusted in combination with the current cost data to obtain more accurate final forecast cost data. This effectively solves the problem that the traditional model is insufficiently generalized in the cost analysis of power grid assets and cannot adapt to the dynamic changes in the environment and the differences in regional characteristics, improves the accuracy and reliability of the forecast, and provides a more advanced and flexible solution for power grid asset management; it also improves the accuracy and adaptability of power grid asset cost forecasts, providing a powerful decision-making support tool for power grid operators, helping to optimize asset management, reduce operating costs, and improve the economic benefits and operating efficiency of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a flow chart of a method for analyzing power grid asset costs based on a neural network according to an embodiment of the present invention;

[0052] Figure 2 It is a schematic diagram of a power grid asset cost analysis system based on neural network in another embodiment of the present invention. DETAILED DESCRIPTION

[0053] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0054] Combination Figure 1 As shown, the present invention provides a method for analyzing power grid asset costs based on a neural network, comprising:

[0055] Obtain basic data and external factor data for each asset in the power grid within the current preset time period, wherein the basic data includes asset acquisition cost, operation and maintenance records, technical parameters and failure rate, and the external factor data includes geographic data of the asset within a preset range, current environmental data and asset usage data.

[0056] Specifically, basic data is the core of power grid asset management, which covers multiple key parameters of equipment. Among them, asset acquisition cost: refers to the initial investment cost when purchasing assets. Operation and maintenance records: including daily operation data, maintenance logs, repair records, etc. of equipment, which help to understand the use and maintenance of equipment. Technical parameters: refers to the technical specifications of equipment, such as voltage level, current capacity, etc., which are crucial for evaluating equipment performance. Failure rate: records the frequency of equipment failures, which is an important indicator for measuring equipment reliability. External factor data is closely related to the environment and usage of the equipment. Geographical data: including the geographical location, topography, etc. of the equipment, which affect the safe operation and maintenance difficulty of the equipment. Current environmental data: environmental factors such as temperature, humidity, wind force, etc., have an impact on the operation of the power grid. Asset usage data: records the frequency of use of equipment, operating time, etc., which helps to evaluate the efficiency and service life of the equipment. By collecting and analyzing these data, the present invention can more accurately predict the operating status and potential problems of the equipment, so as to take preventive measures, reduce the occurrence of failures, and ensure the stable operation of the power grid. At the same time, these data can also be used for asset evaluation, life prediction, cost-benefit analysis, etc., to provide support for the planning and optimization of the power grid.

[0057] The asset purchase cost, the operation and maintenance record, the technical parameters and the failure rate of each asset are related to the geographic data, the current environment data and the asset usage data of the asset to obtain the correlation coefficient between the basic data of the asset and the external factor data.

[0058] Specifically, first, collect basic data such as the acquisition cost, operation and maintenance records, technical parameters and failure rate of each asset, as well as external factor data such as geographic data, current environmental data and asset usage data. Clean and format the collected data to ensure the accuracy and consistency of the data, including steps such as removing outliers, filling missing values, and standardizing or normalizing the data. Extract or construct features that are helpful for analysis from the original data; for example, maintenance frequency can be extracted from operation and maintenance records, and key performance indicators can be extracted from technical parameters. Use statistical methods and visualization tools to explore the relationships and patterns between data; this helps identify which variables may have a significant impact on asset performance. Select appropriate statistical models or machine learning algorithms to analyze the data, such as linear regression, decision trees, random forests, neural networks, etc. Use models to evaluate the strength of association between basic data and external factor data; this can be achieved by calculating correlation coefficients (such as Pearson correlation coefficients), coefficients of determination (R²) or other statistical indicators. Evaluate the performance of the model, such as testing the generalization ability of the model through cross-validation; adjust model parameters or select different models based on the evaluation results. Interpret the model results, determine which external factors are significantly correlated with the underlying data of the asset, and quantify the strength of this correlation. Finally, organize the analysis results into reports and provide them to decision makers, thereby helping grid operators better understand the factors affecting asset performance and formulate maintenance and upgrade strategies accordingly.

[0059] Feature extraction is performed on the basic data and the external factor data to obtain a plurality of internal features and a plurality of external features of the asset.

[0060] Specifically, feature extraction of basic data and external factor data of power grid assets involves converting raw data into key information that facilitates analysis and prediction; this process first requires clarifying the analysis objectives, and then integrating and exploring the data to identify patterns and trends. Next, internal features are extracted from basic data such as asset acquisition costs, operation and maintenance records, technical parameters, and failure rates, while external features are extracted from geographic data, environmental data, and asset usage data. Feature construction involves creating new indicators that reflect asset performance and status, while feature selection aims to select the most useful features for the target task. Finally, features are processed through feature transformation and dimensionality reduction techniques to optimize model performance and avoid overfitting, ensuring the effectiveness of features and providing data support for power grid asset management.

[0061] According to the internal characteristics and the external characteristics of the asset, current cost data of the asset is obtained.

[0062] Specifically, by combining the internal characteristics of power grid assets, including acquisition cost, technical parameters, operation and maintenance records, and failure rates, and external characteristics, including geographic data, environmental data, and asset usage data, the current cost of assets can be comprehensively evaluated. This involves analyzing maintenance records to predict maintenance costs, considering the impact of technical parameters on energy consumption and component replacement, evaluating the impact of geographic and environmental factors on maintenance difficulty and cost, and analyzing asset usage patterns and potential risks. Combining these factors, the total cost, including direct and indirect costs, can be calculated, and a cost-benefit analysis can be performed, thereby providing data support and decision-making basis for the management and budget formulation of power grid assets.

[0063] According to the correlation coefficient of the asset, weights are assigned to the internal features and the external features to obtain weight values ​​corresponding to each of the internal features and the external features.

[0064] Specifically, by analyzing the correlation coefficients of internal characteristics of power grid assets such as acquisition cost, technical parameters, operation and maintenance records and failure rates, as well as external characteristics such as geographic data, environmental data and asset usage data, the impact of each characteristic on asset performance is quantified; this process includes data standardization, calculation of correlation coefficients, weight allocation, weight optimization, verification and adjustment, and final determination of weight values. Ultimately, these weight values ​​are applied to asset evaluation and management decisions to ensure that the importance of characteristics to asset performance is reasonably reflected, thereby improving the efficiency and effectiveness of asset management.

[0065] The internal features and the weight values ​​corresponding to the internal features and the external features and the weight values ​​corresponding to the external features are input into a neural network model to obtain initial predicted cost data of the asset in a preset future time period.

[0066] Specifically, the internal characteristics and weight values ​​of power grid assets and the external characteristics and weight values ​​thereof are used as inputs, and the initial cost data of the assets in a preset future time period is predicted by training a neural network model; including constructing a suitable neural network architecture, such as a multi-layer perceptron or a convolutional neural network, inputting feature data for learning and adjusting network weights, optimizing model parameters through iterative training, and finally outputting a prediction of the future cost of the asset; the data-driven method of the present invention can capture complex nonlinear relationships and provide more accurate and forward-looking decision support for power grid asset management.

[0067] The final predicted cost data of the asset is obtained based on the initial predicted cost data and the current cost data.

[0068] Specifically, the initial forecast cost data generated by the neural network model is combined with the current actual cost data for further analysis and adjustment to obtain the final forecast cost data of the asset. This process includes sensitivity analysis of the forecast data, evaluating the impact of different factors on cost forecasts, and using statistical methods or machine learning techniques to adjust the forecast model to ensure the accuracy and reliability of the forecast results. Ultimately, the forecast cost data obtained in this way can provide more accurate financial guidance for the long-term planning and management of power grid assets.

[0069] The neural network-based power grid asset cost analysis method of the present invention, first, systematically collects basic data and external factor data of each asset, such as purchase cost, maintenance records, technical parameters, failure rate, geography, environment and usage, etc., to ensure that the data covers the whole picture of the asset. Secondly, through feature extraction and correlation coefficient calculation, the key factors affecting the cost are excavated, which not only enhances the model's ability to capture data diversity and complexity, but also provides a basis for weight allocation; the weight allocation mechanism further optimizes the influence of the features and ensures the pertinence of model training. Subsequently, the weighted features are input into the neural network model, and the initial predicted cost data of the asset in the future time period is generated by using its powerful nonlinear fitting ability. Finally, by integrating with the current cost data, adjusting and optimizing the prediction results, more accurate and reliable final predicted cost data is generated. The present invention obtains the basic data and external factor data of the power grid assets, uses feature extraction technology to obtain the internal and external features of the assets, and calculates the correlation coefficients between these features and the asset cost. Further, through weight allocation, the importance of features to cost prediction is strengthened, and the weighted features are input into the neural network model for training to obtain the initial predicted cost data of the asset. Finally, the initial forecast results are adjusted in combination with the current cost data to obtain more accurate final forecast cost data. This effectively solves the problem that the traditional model is insufficiently generalized in the cost analysis of power grid assets and cannot adapt to the dynamic changes in the environment and the differences in regional characteristics, improves the accuracy and reliability of the forecast, and provides a more advanced and flexible solution for power grid asset management; it also improves the accuracy and adaptability of power grid asset cost forecasts, providing a powerful decision-making support tool for power grid operators, helping to optimize asset management, reduce operating costs, and improve the economic benefits and operating efficiency of the power grid.

[0070] Optionally, the establishing of a relationship between the asset purchase cost, the operation and maintenance record, the technical parameter and the failure rate of each asset and the geographic data, the current environment data and the asset usage data of the asset to obtain a correlation coefficient between the basic data of the asset and the external factor data includes:

[0071] Performing data analysis on the asset purchase cost, the operation and maintenance record, the technical parameter and the failure rate of the asset to obtain the change trends of the asset purchase cost, the operation and maintenance record, the technical parameter and the failure rate within the current preset time period;

[0072] quantifying the linear relationship of the change trend among the geographic data of the asset, the current environment data, and the asset usage data by a Pearson correlation coefficient;

[0073] The correlation coefficient between the basic data of the asset and the external factor data is calculated based on the linear relationship.

[0074] Specifically, we first conduct an in-depth analysis of basic data such as asset acquisition costs, operation and maintenance records, technical parameters, and failure rates to determine their changing trends within a preset time period. This analysis includes methods such as time series analysis and trend line fitting to reveal the evolution of asset costs and performance over time. Next, we use the Pearson correlation coefficient to quantify the linear relationship between these changing trends and the geographic, environmental, and usage data of the assets; the Pearson correlation coefficient is a value between -1 and 1 that measures the strength and direction of the linear correlation between two variables; a positive value indicates a positive correlation between the variables, a negative value indicates a negative correlation, and a value close to 0 indicates no or very weak linear relationship. The calculated correlation coefficient can be used to evaluate the mutual influence between basic data and external factor data, and then a comprehensive evaluation model can be constructed; this model can help identify which external factors have a significant impact on asset costs and performance, thereby providing data support for power grid asset management.

[0075] In this optional embodiment, by accurately quantifying the correlation between internal basic data and external factor data, the accuracy of predicting the dynamic changes in power grid asset costs is significantly improved; this data-based prediction can not only help power grid operators better understand the evolution of asset performance, but also predict potential risks and cost increases, allowing operators to adjust maintenance plans and operating strategies in a timely manner, optimize resource allocation, reduce unnecessary costs, and improve the overall operating efficiency and reliability of the power grid. In addition, this method can also provide a scientific basis for long-term planning and investment decisions of the power grid, and enhance the foresight and adaptability of power grid asset management.

[0076] Optionally, the extracting features of the basic data and the external factor data to obtain a plurality of internal features and a plurality of external features of the asset includes:

[0077] Extracting features from the basic data to obtain the number of maintenance times, maintenance costs, component replacement frequency, equipment capacity, voltage level, and line length of the asset within the current preset time period, and using them as the internal features of the asset;

[0078] Feature extraction is performed on the external factor data to obtain the altitude, terrain type, geographical location, average temperature and humidity, pollution level and usage requirements of the asset within the current preset time period, and use them as the external features of the asset.

[0079] Specifically, through feature extraction technology, the basic data and external factor data of power grid assets are deeply analyzed to identify and quantify the key factors affecting asset performance and cost. First, the internal features extracted from the basic data include maintenance times, repair costs, component replacement frequency, equipment capacity, voltage level, and line length. These features directly reflect the operating status and maintenance needs of assets, providing important information for assessing the health status of assets and predicting future maintenance activities. Secondly, the feature extraction of external factor data focuses on the geographical and environmental conditions of assets, including altitude, terrain type, geographical location, average temperature and humidity, pollution level, and usage requirements; these external features have a significant impact on the durability, performance stability, and maintenance strategy of assets.

[0080] In this optional embodiment, by combining these internal and external characteristics, a comprehensive asset performance evaluation model is constructed; the model can predict the performance changes and potential maintenance needs of assets under different conditions, thereby providing data-driven decision support for power grid operators. For example, by analyzing the frequency of component replacement and maintenance costs, the aging rate of assets and maintenance cost trends can be predicted; and considering geographical location and environmental factors can help evaluate the reliability of assets under extreme weather or pollution conditions. In addition, the application of the present invention can significantly improve the efficiency and effectiveness of power grid asset management; by accurately predicting the maintenance needs and costs of assets, operators can optimize maintenance plans, reduce unexpected failures, extend the service life of assets, and reduce long-term operating costs. At the same time, a deep understanding of external environmental factors helps to take potential environmental risks into account in the asset planning and design stages, thereby improving the overall resilience and reliability of the power grid.

[0081] Optionally, obtaining the current cost data of the asset according to the internal characteristics and the external characteristics of the asset includes:

[0082] The number of maintenance, the repair cost, the frequency of component replacement, the equipment capacity, the voltage level, the line length, the altitude, the terrain type, the geographical location, the average temperature and humidity, the pollution level, and the usage demand of the asset are standardized to obtain standardized data of the same magnitude;

[0083] The standardized data is input into a cost prediction model to obtain the current cost data of the asset within the current preset time period.

[0084] Specifically, through standardization, the comparability of multi-dimensional data, such as the number of maintenance times, repair costs, component replacement frequency, equipment capacity, voltage level, line length, altitude, terrain type, geographical location, average temperature and humidity, pollution level, and usage demand of power grid assets, is ensured; the original data is converted into a dimensionless format, usually using methods such as Z-score standardization or range standardization to eliminate the influence of different dimensions and magnitudes, making it possible to compare and analyze different features at the same level. After the standardization process is completed, the data is used as input for training and building a cost prediction model. The model may use machine learning algorithms such as regression analysis, random forests, or neural networks to identify complex relationships between data and predict the current cost data of assets in the current preset time period.

[0085] In this optional embodiment, the efficiency of model training and the accuracy of prediction are significantly improved through standardization; by ensuring that all input features are at the same magnitude, the model can more accurately capture the impact of different features on costs, avoiding the problem of certain features being over-dominated in the model due to their large numerical range. In addition, the standardized data helps to improve the generalization ability of the model, so that the model can maintain stable prediction performance when faced with new or unseen data. Inputting standardized data into the cost prediction model can provide grid operators with accurate cost estimates, helping them better plan financial budgets, optimize maintenance plans, and improve asset utilization efficiency; this data-driven approach makes grid asset management more scientific and systematic, helps reduce operational risks and costs, and improves the economic benefits and competitiveness of the grid.

[0086] Optionally, the weight allocation of the internal features and the external features according to the correlation coefficient of the asset to obtain the weight values ​​corresponding to each of the internal features and the external features respectively includes:

[0087] The internal features and the external features are sorted according to the correlation coefficients corresponding to the asset acquisition cost, the operation and maintenance record, the technical parameters, and the failure rate of the asset in the asset relative to the geographic data, the current environment data, and the asset usage data to obtain a correlation sequence;

[0088] According to the association sequence, the initial weights of the asset acquisition cost, the operation and maintenance record, the technical parameters, the failure rate, the geographic data, the current environmental data and the asset usage data of the asset are adjusted to obtain the weight values ​​corresponding to each of the internal characteristics and the external characteristics.

[0089] Specifically, by analyzing the internal characteristics of the power grid assets, such as the purchase cost, operation and maintenance records, technical parameters and failure rates, as well as the external characteristics such as geographic data, current environmental data and asset usage data, these characteristics are sorted using their correlation coefficients relative to these characteristics to form a correlation sequence; by calculating the strength of the correlation between each feature and the power grid asset cost, and sorting the features according to these strengths, the relative importance of their impact on the cost is determined. Next, the initial weight of each feature is adjusted according to the correlation sequence; the purpose of the weight adjustment is to ensure that those features that are highly correlated with the cost have a greater influence in the cost prediction model; this can be achieved in a variety of ways, for example, the correlation coefficient can be used as a direct mapping of the weight, or algorithms such as principal component analysis (PCA) or factor analysis can be applied to determine the weight.

[0090] In this optional embodiment, the weight value obtained by this method can more accurately reflect the contribution of each feature to the asset cost; it not only improves the accuracy of the cost prediction model, but also enhances the interpretability of the model, so that power grid operators can identify the key factors affecting the cost and formulate more targeted maintenance and operation strategies accordingly. For example, if technical parameters are highly correlated with costs, operators may pay more attention to technology upgrades and equipment optimization; if environmental data is the main influencing factor, special maintenance measures may need to be taken in specific environments. In addition, by continuously optimizing the weight distribution, the model's adaptability to new data can be improved, ensuring the timeliness and reliability of the prediction results. This method provides a dynamic, data-driven decision support tool for power grid asset management, which helps to maximize cost-effectiveness, reduce operational risks, and improve the overall performance and stability of the power grid.

[0091] Optionally, the step of inputting the internal features and the weight values ​​corresponding to the internal features and the external features and the weight values ​​corresponding to the external features into a neural network model to obtain initial predicted cost data of the asset in a preset future time period includes:

[0092] Inputting the internal features and the weight values ​​corresponding to the internal features and the external features and the weight values ​​corresponding to the external features into a neural network model;

[0093] According to the internal feature, obtaining a feature value corresponding to the internal feature;

[0094] Multiplying the weight value corresponding to the internal feature by the feature value corresponding to the internal feature to obtain the weighted internal feature;

[0095] According to the external feature, obtaining a feature value corresponding to the external feature;

[0096] Multiplying the weight value corresponding to the external feature by the feature value corresponding to the external feature to obtain the weighted external feature;

[0097] The weighted internal features are predicted by combining the neural network model with the weighted external features to obtain the initial predicted cost data of the asset in the preset future time period.

[0098] Specifically, the internal and external features of power grid assets and their corresponding weight values ​​are input into the neural network model to achieve accurate prediction of the future cost of assets; internal features include asset acquisition cost, operation and maintenance records, technical parameters and failure rate, while external features include geographic data, environmental data and asset usage data. First, weights are assigned to these features based on the importance of the associated sequence and features, and these weights reflect the influence of the features on cost prediction. In terms of obtaining feature values, the internal features are quantified to obtain the specific value of each feature. Subsequently, these feature values ​​are multiplied by the corresponding weight values ​​to obtain the weighted internal features. This step ensures that the model can be adjusted according to the importance of the features and strengthens the role of features that have a significant impact on costs. The weighted internal features are input into the neural network model for training and prediction. By learning the relationship between features and costs, the neural network model can capture complex nonlinear patterns, thereby providing an initial prediction of the cost of assets in a preset future time period; this process involves multiple layers of the model, including input layers, hidden layers and output layers, where the hidden layers process the input data through activation functions to identify patterns and trends in the data.

[0099] In this optional embodiment, the nonlinear characteristics of the neural network enable it to handle the complexity and uncertainty in grid asset cost forecasting, thereby improving the accuracy of the forecast. Secondly, through weight adjustment, the model can pay more attention to those features that have a greater impact on costs, thereby improving the relevance and practicality of the forecast. In addition, the neural network model has good generalization capabilities and can adapt to new data and conditions, providing a flexible and dynamic forecasting tool for grid asset management. By combining feature weights and neural network models, the present invention not only improves the accuracy of grid asset cost forecasts, but also enhances the interpretability and adaptability of the model, providing strong decision support for grid operators, helping to optimize asset management strategies, reduce operating costs, and improve the economic benefits and operating efficiency of the grid.

[0100] Optionally, obtaining the final predicted cost data of the asset according to the initial predicted cost data in combination with the current cost data includes:

[0101] Comparing the initial predicted cost data with the current cost data, obtaining the cost change difference of the asset within the preset future time period;

[0102] determining a forecast error based on the cost change difference;

[0103] Assigning weights to the initial predicted cost data and the current cost data according to the prediction error to obtain weight values ​​corresponding to the initial predicted cost data and the current cost data;

[0104] The final predicted cost data of the asset is obtained by taking a weighted average based on the weight values ​​corresponding to the initial predicted cost data and the current cost data.

[0105] Specifically, by combining the initial forecast cost data and the current cost data, a detailed cost analysis and forecast adjustment are performed. The initial forecast cost data is obtained through a neural network model, while the current cost data is a direct observation based on actual operating conditions. First, the two sets of data are compared to calculate the difference in cost changes in a preset future time period, which reflects the deviation between the forecast model and the actual situation. Next, the forecast error is determined based on the difference in cost changes. This error is a key indicator for evaluating forecast accuracy; the forecast error can be calculated using a variety of methods, such as mean square error (MSE) or mean absolute error (MAE), which can quantify the gap between the forecast value and the actual value. Then, weights are assigned to the initial forecast cost data and the current cost data based on the forecast error; in this step, data with smaller forecast errors are usually given higher weights to ensure the accuracy and reliability of the final forecast results; the weight assignment can be based on the size of the error or determined by a more complex optimization algorithm. Finally, the final forecast cost data of the asset is obtained by combining the weight values ​​of the initial forecast cost data and the current cost data through weighted averaging; this weighted averaging process takes into account the relative importance of the forecast data and the actual data, thereby providing a comprehensive cost forecast value.

[0106] In this optional embodiment, the accuracy and practicality of cost forecasting are improved by combining the forecasting model and actual data. By continuously adjusting the forecasting model and reducing the forecasting error, the cost change trend of the asset can be more realistically reflected. In addition, the weight distribution mechanism enables the model to adaptively adjust the degree of dependence on the forecast data and actual data, thereby enhancing the robustness of the model. Ultimately, the forecasted cost data obtained by weighted averaging provides a more accurate and reliable cost control and planning tool for power grid operators, which helps to optimize resource allocation, reduce operational risks, and improve asset management efficiency.

[0107] Optionally, the neural network-based power grid asset cost analysis method further includes:

[0108] Cost prediction is performed for each of the assets in the power grid through a distributed system, wherein the distributed system includes a plurality of nodes, and each of the nodes performs cost prediction for one of the assets.

[0109] Specifically, by building a distributed system, the cost prediction of each asset in the power grid is achieved. The system consists of multiple nodes, each of which is independently responsible for collecting and analyzing the basic data and external factor data of a specific asset. Each node performs data preprocessing, feature extraction, local model training and cost prediction, and then aggregates the prediction results to the central node or shared storage; this architecture allows the system to process large amounts of data in parallel, improve computing efficiency, and has good scalability and robustness, adapting to the specific needs of different assets, ensuring the flexibility of data processing and the protection of data privacy.

[0110] In this optional embodiment, firstly, through parallel processing, the overall time consumption of the prediction task is significantly reduced, and the efficiency of the prediction work is improved. Secondly, the scalability of the system allows nodes to be easily expanded as the number of grid assets increases, without the need for large-scale changes to the existing architecture. In addition, the distributed architecture enhances the robustness of the system, and the failure of a single node will not affect the operation of the entire system. In terms of flexibility, each node can update the model independently and quickly adapt to changes in asset characteristics. Finally, by aggregating the prediction results of each node, a comprehensive view of the grid cost forecast can be obtained, providing strong data support for the refined management and strategic planning of the grid.

[0111] like Figure 2 As shown, an embodiment of the present invention provides a power grid asset cost analysis system based on a neural network, comprising:

[0112] A data acquisition unit, used to acquire basic data and external factor data of each asset in the power grid within a current preset time period, wherein the basic data includes asset acquisition cost, operation and maintenance records, technical parameters and failure rate, and the external factor data includes geographic data of the asset within a preset range, current environmental data and asset usage data;

[0113] A correlation coefficient calculation unit, used to establish a relationship between the asset purchase cost, the operation and maintenance record, the technical parameters and the failure rate of each asset and the geographic data, the current environment data and the asset use data of the asset, to obtain a correlation coefficient between the basic data of the asset and the external factor data;

[0114] A feature extraction unit, used to extract features from the basic data and the external factor data to obtain a plurality of internal features and a plurality of external features of the asset;

[0115] A first data processing unit, configured to obtain current cost data of the asset according to the internal characteristics and the external characteristics of the asset;

[0116] A second data processing unit is used to assign weights to the internal features and the external features according to the correlation coefficient of the asset, and obtain weight values ​​corresponding to each of the internal features and the external features;

[0117] A prediction unit, used for inputting the internal features and the weight values ​​corresponding to the internal features and the external features and the weight values ​​corresponding to the external features into a neural network model to obtain initial predicted cost data of the asset in a preset future time period;

[0118] A correction unit is used to obtain final predicted cost data of the asset based on the initial predicted cost data in combination with the current cost data.

[0119] The neural network-based power grid asset cost analysis system of the present invention first systematically collects basic data and external factor data of each asset, such as purchase cost, maintenance records, technical parameters, failure rate, geography, environment and usage, etc., to ensure that the data covers the whole picture of the asset. Secondly, through feature extraction and correlation coefficient calculation, the key factors affecting the cost are excavated, which not only enhances the model's ability to capture data diversity and complexity, but also provides a basis for weight allocation; the weight allocation mechanism further optimizes the influence of the features and ensures the pertinence of model training. Subsequently, the weighted features are input into the neural network model, and the initial predicted cost data of the asset in the future time period is generated by using its powerful nonlinear fitting ability. Finally, by integrating with the current cost data, adjusting and optimizing the prediction results, more accurate and reliable final predicted cost data is generated. The present invention obtains the basic data and external factor data of the power grid assets, uses feature extraction technology to obtain the internal and external features of the assets, and calculates the correlation coefficients between these features and the asset cost. Further, through weight allocation, the importance of features to cost prediction is strengthened, and the weighted features are input into the neural network model for training to obtain the initial predicted cost data of the asset. Finally, the initial forecast results are adjusted in combination with the current cost data to obtain more accurate final forecast cost data. This effectively solves the problem that the traditional model is insufficiently generalized in the cost analysis of power grid assets and cannot adapt to the dynamic changes in the environment and the differences in regional characteristics, improves the accuracy and reliability of the forecast, and provides a more advanced and flexible solution for power grid asset management; it also improves the accuracy and adaptability of power grid asset cost forecasts, providing a powerful decision-making support tool for power grid operators, helping to optimize asset management, reduce operating costs, and improve the economic benefits and operating efficiency of the power grid.

[0120] An embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for analyzing power grid asset costs based on a neural network is implemented, including: obtaining basic data and external factor data of each asset in the power grid within a current preset time period, wherein the basic data includes asset acquisition cost, operation and maintenance records, technical parameters, and failure rate, and the external factor data includes geographic data of the asset within a preset range, current environmental data, and asset usage data;

[0121] Establishing a relationship between the asset purchase cost, the operation and maintenance record, the technical parameters and the failure rate of each asset and the geographic data, the current environment data and the asset usage data of the asset to obtain a correlation coefficient between the basic data of the asset and the external factor data;

[0122] Extracting features from the basic data and the external factor data to obtain a plurality of internal features and a plurality of external features of the asset;

[0123] Obtaining current cost data of the asset based on the internal characteristics and the external characteristics of the asset;

[0124] According to the correlation coefficient of the asset, weights are assigned to the internal features and the external features to obtain weight values ​​corresponding to each of the internal features and the external features;

[0125] Inputting the internal features and the weight values ​​corresponding to the internal features and the external features and the weight values ​​corresponding to the external features into a neural network model to obtain initial predicted cost data of the asset in a preset future time period;

[0126] The final predicted cost data of the asset is obtained based on the initial predicted cost data and the current cost data.

[0127] The computer-readable storage medium of the present invention, first, systematically collects basic data and external factor data of each asset, such as purchase cost, maintenance records, technical parameters, failure rate, geography, environment and usage, etc., to ensure that the data covers the whole picture of the asset. Secondly, through feature extraction and correlation coefficient calculation, the key factors affecting the cost are excavated, which not only enhances the model's ability to capture data diversity and complexity, but also provides a basis for weight allocation; the weight allocation mechanism further optimizes the influence of the features and ensures the pertinence of model training. Subsequently, the weighted features are input into the neural network model, and its powerful nonlinear fitting ability is used to generate the initial predicted cost data of the asset in the future time period. Finally, by integrating with the current cost data, adjusting and optimizing the prediction results, more accurate and reliable final predicted cost data are generated. The present invention obtains the basic data and external factor data of the power grid assets, uses feature extraction technology to obtain the internal and external features of the assets, and calculates the correlation coefficients between these features and the asset cost. Further, through weight allocation, the importance of features to cost prediction is strengthened, and the weighted features are input into the neural network model for training to obtain the initial predicted cost data of the asset. Finally, the initial forecast results are adjusted in combination with the current cost data to obtain more accurate final forecast cost data. This effectively solves the problem that the traditional model is insufficiently generalized in the cost analysis of power grid assets and cannot adapt to the dynamic changes in the environment and the differences in regional characteristics, improves the accuracy and reliability of the forecast, and provides a more advanced and flexible solution for power grid asset management; it also improves the accuracy and adaptability of power grid asset cost forecasts, providing a powerful decision-making support tool for power grid operators, helping to optimize asset management, reduce operating costs, and improve the economic benefits and operating efficiency of the power grid.

[0128] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0129] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0130] The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for analyzing power grid asset cost based on neural network, characterized in that: include: Obtaining basic data and external factor data of each asset in the power grid within a current preset time period, wherein the basic data includes asset acquisition cost, operation and maintenance records, technical parameters and failure rate, and the external factor data includes geographic data of the asset within a preset range, current environmental data and asset usage data; Establish a relationship between the asset purchase cost, the operation and maintenance record, the technical parameters and the failure rate of each asset and the geographic data, the current environment data and the asset usage data of the asset to obtain the correlation coefficient between the basic data of the asset and the external factor data; specifically including: performing data analysis on the asset purchase cost, the operation and maintenance record, the technical parameters and the failure rate of the asset to obtain the change trends of the asset purchase cost, the operation and maintenance record, the technical parameters and the failure rate within the current preset time period; quantifying the linear relationship of the change trend in the geographic data, the current environment data and the asset usage data of the asset by the Pearson correlation coefficient; and calculating the correlation coefficient between the basic data of the asset and the external factor data according to the linear relationship; Extracting features from the basic data and the external factor data to obtain a plurality of internal features and a plurality of external features of the asset; Obtaining current cost data of the asset based on the internal characteristics and the external characteristics of the asset; According to the correlation coefficient of the asset, weights are assigned to the internal features and the external features to obtain weight values ​​corresponding to each of the internal features and the external features; Inputting the internal features and the weight values ​​corresponding to the internal features and the external features and the weight values ​​corresponding to the external features into a neural network model to obtain initial predicted cost data of the asset in a preset future time period; The final predicted cost data of the asset is obtained based on the initial predicted cost data and the current cost data.

2. The neural network-based power grid asset cost analysis method according to claim 1, characterized in that: The feature extraction of the basic data and the external factor data to obtain a plurality of internal features and a plurality of external features of the asset includes: Extracting features from the basic data to obtain the number of maintenance times, maintenance costs, component replacement frequency, equipment capacity, voltage level, and line length of the asset within the current preset time period, and using them as the internal features of the asset; Feature extraction is performed on the external factor data to obtain the altitude, terrain type, geographical location, average temperature and humidity, pollution level and usage requirements of the asset within the current preset time period, and use them as the external features of the asset.

3. The neural network-based power grid asset cost analysis method according to claim 2, characterized in that: The obtaining of current cost data of the asset according to the internal characteristics and the external characteristics of the asset comprises: The number of maintenance, the repair cost, the frequency of component replacement, the equipment capacity, the voltage level, the line length, the altitude, the terrain type, the geographical location, the average temperature and humidity, the pollution level, and the usage demand of the asset are standardized to obtain standardized data of the same magnitude; The standardized data is input into a cost prediction model to obtain the current cost data of the asset within the current preset time period.

4. The neural network-based power grid asset cost analysis method according to claim 1, characterized in that: The step of assigning weights to the internal features and the external features according to the correlation coefficient of the asset to obtain weight values ​​corresponding to each of the internal features and the external features includes: The internal features and the external features are sorted according to the correlation coefficients corresponding to the asset acquisition cost, the operation and maintenance record, the technical parameters, and the failure rate of the asset in the asset relative to the geographic data, the current environment data, and the asset usage data to obtain a correlation sequence; According to the association sequence, the initial weights of the asset acquisition cost, the operation and maintenance record, the technical parameters, the failure rate, the geographic data, the current environmental data and the asset usage data of the asset are adjusted to obtain the weight values ​​corresponding to each of the internal characteristics and the external characteristics.

5. The neural network-based power grid asset cost analysis method according to claim 1, characterized in that: The step of inputting the internal features and the weight values ​​corresponding to the internal features and the external features and the weight values ​​corresponding to the external features into a neural network model to obtain initial predicted cost data of the asset in a preset future time period includes: Inputting the internal features and the weight values ​​corresponding to the internal features and the external features and the weight values ​​corresponding to the external features into a neural network model; According to the internal feature, obtaining a feature value corresponding to the internal feature; Multiplying the weight value corresponding to the internal feature by the feature value corresponding to the internal feature to obtain the weighted internal feature; According to the external feature, obtaining a feature value corresponding to the external feature; Multiplying the weight value corresponding to the external feature by the feature value corresponding to the external feature to obtain the weighted external feature; The weighted internal features are predicted by combining the neural network model with the weighted external features to obtain the initial predicted cost data of the asset in the preset future time period.

6. The neural network-based power grid asset cost analysis method according to claim 1, characterized in that: The step of obtaining the final predicted cost data of the asset based on the initial predicted cost data and the current cost data includes: Comparing the initial predicted cost data with the current cost data, obtaining the cost change difference of the asset within the preset future time period; determining a forecast error based on the cost change difference; Assigning weights to the initial predicted cost data and the current cost data according to the prediction error to obtain weight values ​​corresponding to the initial predicted cost data and the current cost data; The final predicted cost data of the asset is obtained by taking a weighted average based on the weight values ​​corresponding to the initial predicted cost data and the current cost data.

7. The neural network-based power grid asset cost analysis method according to claim 1, characterized in that: Also includes: Cost prediction is performed for each of the assets in the power grid through a distributed system, wherein the distributed system includes a plurality of nodes, and each of the nodes performs cost prediction for one of the assets.

8. A power grid asset cost analysis system based on neural network, characterized in that: include: A data acquisition unit, used to acquire basic data and external factor data of each asset in the power grid within a current preset time period, wherein the basic data includes asset acquisition cost, operation and maintenance records, technical parameters and failure rate, and the external factor data includes geographic data of the asset within a preset range, current environmental data and asset usage data; A correlation coefficient calculation unit is used to establish a relationship between the asset purchase cost, the operation and maintenance record, the technical parameters and the failure rate of each asset and the geographic data, the current environment data and the asset use data of the asset, so as to obtain the correlation coefficient between the basic data of the asset and the external factor data; wherein, it specifically includes: performing data analysis according to the asset purchase cost, the operation and maintenance record, the technical parameters and the failure rate of the asset, respectively, to obtain the change trends of the asset purchase cost, the operation and maintenance record, the technical parameters and the failure rate within the current preset time period; quantifying the linear relationship of the change trend in the geographic data, the current environment data and the asset use data of the asset by the Pearson correlation coefficient; and calculating according to the linear relationship to obtain the correlation coefficient between the basic data of the asset and the external factor data; A feature extraction unit, used to extract features from the basic data and the external factor data to obtain a plurality of internal features and a plurality of external features of the asset; A first data processing unit, configured to obtain current cost data of the asset according to the internal characteristics and the external characteristics of the asset; A second data processing unit is used to assign weights to the internal features and the external features according to the correlation coefficient of the asset, and obtain weight values ​​corresponding to each of the internal features and the external features; A prediction unit, used for inputting the internal features and the weight values ​​corresponding to the internal features and the external features and the weight values ​​corresponding to the external features into a neural network model to obtain initial predicted cost data of the asset in a preset future time period; A correction unit is used to obtain final predicted cost data of the asset based on the initial predicted cost data in combination with the current cost data.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the neural network-based power grid asset cost analysis method described in any one of claims 1 to 7 is implemented.

Images