Convolutional neural network-based power grid production technical improvement investment analysis method and system

Abstract

The invention relates to a power grid production technical improvement investment analysis method and system based on a convolutional neural network, and belongs to the technical field of power industry analysis. The method comprises the following steps: performing factor analysis on influence factors through network production technical improvement historical data, and screening out a data type with the highest technical improvement investment benefit association degree from the influence factors by adopting a gray scale association degree-based method as the input of an analysis model; performing cleaning, standardization and normalization data processing on the selected historical index data of the power grid production technical improvement investment; dividing historical data into three types of data sets of testing, training and verification; secondly, sampling the test set in a putting-back manner by adopting a Bootstrap technology to construct a plurality of test subsets, constructing an initial model of the convolutional neural network, and selecting a kernel function for the initial model to optimize the model calculation efficiency; and determining a weighted value of each model by adopting a weighted fusion strategy according to prediction results and performance expressions of the plurality of convolutional neural network models to obtain a final fusion model, and predicting a final result according to the fusion model.

Description

Technical Field

[0001] The present invention relates to a method and system for analyzing power grid production technology transformation investment based on convolutional neural networks, belonging to the technical field of power industry analysis. Background Art

[0002] With the rapid development of the power industry and the continuous expansion of the power grid, power grid production technical transformation, as an important means to improve the stability and reliability of the power system, has the characteristics of complex project types, huge numbers, and small single investment amounts. Accurate investment is the key to significant value effects of production technical transformation.

[0003] The means currently adopted by power grid companies mainly rely on the historical experience of professional and technical personnel, select power grid technical transformation investment indicators relatively subjectively, and use simple mathematical formulas such as regression analysis to calculate the predicted value of production technical transformation investment. It is difficult to comprehensively and accurately evaluate the relationship and influence between indicators and investment values, resulting in investment decisions that have an adverse impact on the development of power grid enterprises. Power grid historical data involves financial data, equipment conditions, corporate operating efficiency and other factors. Investment decisions involve many complex factors such as investment scale, technology selection, economic benefits, and social benefits. How to construct a high-precision prediction model and system by selecting closely related value indicators for power grid technical transformation investment is an important issue in power grid technical transformation investment prediction. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for analyzing investment in power grid production technical transformation based on convolutional neural networks to solve the above-mentioned technical problems.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is: The power grid production technology transformation investment analysis method and system based on convolutional neural network, through the factor analysis of the influencing factors of the historical data of power grid production technology transformation, uses the gray-scale correlation method to select the data type with the highest correlation with the technology transformation investment benefit from many influencing factors as the input of the analysis model; Clean, standardize and normalize the historical indicator data of the selected power grid production technology transformation investment; Divide historical data into three types of data sets: testing, training, and validation; Then, Bootstrap technology is used to sample the test set with replacement to construct multiple test subsets, and the initial model of the convolutional neural network is constructed at the same time. The kernel function is selected for the initial model to optimize the model calculation efficiency; A weighted fusion strategy is adopted to determine the weighted values ​​of each model according to the prediction results and performance of multiple convolutional neural network models to obtain the final fusion model. The final result is predicted based on the fusion model, and a scientific basis for investment decision-making is provided based on the prediction results.

[0006] The further improvement of the technical solution of the present invention is: the data preprocessing process is as follows: Factor analysis: First, the historical data of power grid production technological transformation is comprehensively sorted out, and factor analysis technology is used to identify the key factors that affect the investment benefits of technological transformation.

[0007] Data type selection: Based on the grayscale correlation method, the data type with the highest correlation with the investment benefits of technological transformation is selected from a large number of influencing factors as the input of the analysis model.

[0008] Data cleaning and standardization: Clean, standardize and normalize the selected historical indicator data; data cleaning includes removing duplicate values, removing outliers, and filling missing values; standardization is to eliminate the impact of dimensions; normalization is to convert data to the same scale.

[0009] A further improvement of the technical solution of the present invention is: data set division is to divide the preprocessed historical data into a training set, a validation set and a test set according to a certain ratio, which are used for model training, model validation and final evaluation respectively.

[0010] A further improvement of the technical solution of the present invention is that Bootstrap sampling is to use Bootstrap technology to sample the test set multiple times with replacement to construct multiple test subsets to evaluate the stability and generalization ability of the model under different data distributions.

[0011] A further improvement of the technical solution of the present invention is as follows: the convolutional neural network model construction and optimization process is as follows: first, the convolutional neural network initial model is constructed, and the network structure and kernel function are determined by selecting the appropriate number of layers, convolution kernel size, and pooling method according to the characteristics of the power grid technical transformation investment problem; the model calculation efficiency and prediction accuracy are improved by adjusting hyperparameters and optimization algorithms.

[0012] The further improvement of the technical solution of the present invention is that the model fusion and prediction adopts a weighted fusion strategy, and the weighted value of each model is determined according to the prediction results and performance of multiple convolutional neural network models to obtain the final fusion model. The fusion model is used to predict the investment benefits of power grid production technology transformation, and a scientific investment decision-making basis is provided based on the prediction results.

[0013] Due to the adoption of the above technical solution, the technical effects achieved by the present invention are as follows: The technical solution of the present invention can more accurately evaluate the performance of the model and improve its actual application effect in power grid production technical transformation investment analysis by reasonably dividing the data set, using Bootstrap technology for multiple sampling, and constructing and optimizing the convolutional neural network model.

[0014] This technical solution adopts a weighted fusion strategy. According to the prediction results and performance of multiple convolutional neural network models, the weighted value of each model is determined to obtain the final fusion model, the final result is predicted based on the fusion model, and a scientific basis for investment decision-making is provided based on the prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is the basic structure diagram of the convolutional neural network of the present invention. DETAILED DESCRIPTION

[0016] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is further explained below in conjunction with specific implementation methods.

[0017] The present invention is a method and system for analyzing power grid production technology transformation investment based on convolutional neural network, which is used for accurately analyzing investment in the power industry.

[0018] The main idea of ​​this technical solution is to conduct factor analysis on the influencing factors of the historical data of power grid production technical transformation, use the grayscale correlation method to screen out the data type with the highest correlation with the technical transformation investment benefits from many influencing factors as the input of the analysis model, and perform data processing such as cleaning, standardization, and normalization on the selected historical indicator data of power grid production technical transformation investment. The historical data is divided into three data sets: test, training, and verification. Then, the Bootstrap technology is used to sample the test set with replacement to construct multiple test subsets, and at the same time, a convolutional neural network initial model is constructed, and the kernel function is selected for the initial model to optimize the model calculation efficiency; a weighted fusion strategy is used to determine the weighted value of each model according to the prediction results and performance of multiple convolutional neural network models to obtain the final fusion model, predict the final result according to the fusion model, and provide a scientific investment decision-making basis based on the prediction results.

[0019] 1. Data preprocessing: Factor analysis: First, the historical data of power grid production technological transformation is comprehensively sorted out, and factor analysis technology is used to identify the key factors that affect the investment benefits of technological transformation.

[0020] Data type selection: Based on the grayscale correlation method, the data type with the highest correlation with the investment benefits of technological transformation is selected from a large number of influencing factors as the input of the analysis model.

[0021] Data cleaning and standardization: The selected historical indicator data is cleaned (removing duplicates, outliers, filling missing values, etc.), standardized (eliminating the impact of dimensions) and normalized (converting data to the same scale) to improve data quality and the accuracy of analysis results.

[0022] 2. Dataset division: The preprocessed historical data is divided into training set, validation set and test set according to a certain ratio, which are used for model training, model validation and final evaluation respectively.

[0023] 3. Bootstrap Sampling: The Bootstrap technique is used to sample the test set multiple times with replacement to construct multiple test subsets to evaluate the stability and generalization ability of the model under different data distributions.

[0024] 4. Convolutional neural network model construction and optimization: Construct the initial model of the convolutional neural network, and select the appropriate number of layers, convolution kernel size, and pooling method to determine the network structure and kernel function according to the characteristics of the power grid technical transformation investment problem; improve the model's calculation efficiency and prediction accuracy by adjusting hyperparameters and optimizing algorithms.

[0025] 5. Model fusion and prediction: Using a weighted fusion strategy, according to the prediction results and performance of multiple convolutional neural network models, the weighted values ​​of each model are determined to obtain the final fusion model. The fusion model is used to predict the investment benefits of power grid production technology transformation, and a scientific basis for investment decision-making is provided based on the prediction results.

[0026] The following are specific embodiments (1) Factor analysis: Factor analysis explains the correlation or covariance relationship between variables by finding a few common factors hidden behind many variables. In this case, factor analysis technology was used to identify the key factors affecting the investment benefits of technical transformation from the historical data of power grid production technical transformation. These key factors may include but are not limited to the degree of technological innovation, the degree of equipment aging, changes in grid load demand, and the level of operation and maintenance costs. Through factor analysis, we can more clearly understand which factors have a significant impact on the investment benefits of technical transformation, which provides an important basis for subsequent data type selection and model construction.

[0027] (2) Data type selection: The gray correlation method is a mathematical tool based on gray system theory. It determines the degree of correlation between factors by analyzing the similarities or differences in the development trends of various factors in the system. In this case, we used the gray correlation method to calculate the correlation between each influencing factor and the benefits of technological transformation investment. By comparing the correlation between different influencing factors, we screened out the data types with the highest correlation with the benefits of technological transformation investment, such as equipment aging index, technological innovation index, and grid load demand change index. These data types, as inputs to subsequent analysis models, can more accurately reflect the actual situation of technological transformation investment benefits.

[0028] (3) Data cleaning and standardization: After determining the input data type of the analysis model, we need to clean and standardize the selected historical indicator data. Data cleaning is an important part of data preprocessing, which includes operations such as deduplication, removal of outliers, and filling of missing values. In this case, we adopted a variety of data cleaning strategies, such as statistical outlier detection methods and interpolation-based missing value filling methods, to ensure the accuracy and completeness of the data.

[0029] Data standardization is to eliminate the impact of different dimensions on the analysis results. In this case, we used the Z-score standardization method to convert the data into a standard normal distribution with a mean of 0 and a standard deviation of 1. In addition, we also normalized the data and converted the data to the same scale for subsequent analysis and calculation. Through data cleaning and standardization, we improved the data quality and the accuracy of the analysis results, providing reliable data support for subsequent analysis and decision-making.

[0030] (4) Dataset division: In the analysis of power grid production technology transformation investment, data set division is a crucial step. In order to effectively use the pre-processed historical data, we divide it into training set, validation set and test set according to a certain ratio.

[0031] The training set is used for the initial training of the model. By continuously adjusting the model parameters to minimize the loss function, the model can learn the underlying patterns in the data. The validation set is used to verify the model during the training process. By calculating indicators such as loss and accuracy on the validation set, we can evaluate the performance of the model and adjust the model structure and hyperparameters accordingly. The test set is used for the final evaluation after the model training is completed to verify the generalization ability and stability of the model.

[0032] When dividing the data sets, we took into account the diversity and representativeness of the data, ensuring that each data set contained data on technical transformation projects of different types and time periods, so as to fully reflect the actual situation of technical transformation investment in power grid production.

[0033] (5) Bootstrap sampling: In this case, we performed multiple Bootstrap sampling on the test set and constructed multiple test subsets. Each test subset contains the same number of data points as the original test set. By evaluating the performance of the model on the test subset, we can obtain the performance of the model under different data distributions, thereby more accurately evaluating the stability and generalization ability of the model.

[0034] (6) Convolutional neural network model: In building the basic structure of convolutional neural network, Figure 1As shown, it is a multi-layer neural network, each layer consists of multiple two-dimensional planes, and each plane is composed of multiple independent neurons.

[0035] After the input data passes through the convolution operation of the convolution layer C1, it is biased and activated and enters the downsampling layer S1. The downsampling layer can obtain feature maps through a variety of pooling methods. These maps are then convolved to obtain the C2 layer, and then downsampled to obtain the S2 layer. The S2 layer passes through a single-layer perceptron to obtain Fc and is fully connected to obtain the final output.

[0036] The convolution layer is used to extract the features of the input data. We selected the appropriate convolution kernel size and number based on the data dimension and number of features. The pooling layer is used to reduce the data dimension and reduce the amount of calculation. We selected the appropriate pooling method and step size. In addition, we also added a fully connected layer for classification or regression tasks, and selected the Sigmoid activation function to increase the nonlinear expression ability of the model.

[0037] In terms of model optimization, we adjusted the hyperparameters to further improve the computational efficiency and prediction accuracy of the model. The learning rate is a key hyperparameter that determines the step size of the model parameter update during training. We determined the appropriate learning rate range through experiments and adjusted it dynamically during training. The batch size affects the stability and convergence speed of model training. We selected the appropriate batch size based on hardware resources and data size. The number of iterations determines the length of time for model training. We determined the optimal number of iterations by observing the performance changes on the validation set.

[0038] In addition, we also use optimization algorithms (such as Adam, SGD, etc.) to accelerate the training process of the model and reduce the risk of overfitting. The Adam algorithm combines the momentum term and the adaptive learning rate adjustment strategy to automatically adjust the learning rate and accelerate convergence during training. The SGD algorithm is a simple and effective optimization algorithm that reduces the amount of calculation and increases the generalization ability of the model by randomly selecting a data point for gradient update in each iteration.

[0039] In summary, by reasonably dividing the data set, using Bootstrap technology for multiple sampling, and building and optimizing the convolutional neural network model, we can more accurately evaluate the performance of the model and improve its practical application effect in the investment analysis of power grid production technical transformation.

[0040] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A method and system for analyzing power grid production technology transformation investment based on convolutional neural network, characterized by: Through the factor analysis of the influencing factors of the historical data of power grid production technology transformation, the data type with the highest correlation with the technology transformation investment benefit is selected from the numerous influencing factors based on the gray correlation method as the input of the analysis model; Clean, standardize and normalize the historical indicator data of the selected power grid production technology transformation investment; Divide historical data into three types of data sets: testing, training, and validation; Then, Bootstrap technology is used to sample the test set with replacement to construct multiple test subsets, and the initial model of the convolutional neural network is constructed at the same time. The kernel function is selected for the initial model to optimize the model calculation efficiency; A weighted fusion strategy is adopted to determine the weighted values ​​of each model according to the prediction results and performance of multiple convolutional neural network models to obtain the final fusion model. The final result is predicted based on the fusion model, and a scientific basis for investment decision-making is provided based on the prediction results.

2. The method and system for analyzing power grid production technology transformation investment based on convolutional neural network according to claim 1, characterized in that: The data preprocessing process is as follows: Factor analysis: First, we comprehensively sorted out the historical data of power grid production technology transformation and used factor analysis technology to identify the key factors that affect the investment benefits of technology transformation; Data type selection: Based on the grayscale correlation method, the data type with the highest correlation with the investment benefits of technological transformation is selected from a large number of influencing factors as the input of the analysis model; Data cleaning and standardization: Clean, standardize and normalize the selected historical indicator data; data cleaning includes removing duplicate values, removing outliers, and filling missing values; standardization is to eliminate the impact of dimensions; normalization is to convert data to the same scale.

3. The method and system for analyzing power grid production technology transformation investment based on convolutional neural network according to claim 1, characterized in that: Dataset partitioning is to divide the preprocessed historical data into training set, validation set and test set according to a certain ratio, which are used for model training, model validation and final evaluation respectively.

4. The method and system for analyzing power grid production technology transformation investment based on convolutional neural network according to claim 1, characterized in that: Bootstrap sampling uses the Bootstrap technique to sample the test set multiple times with replacement to construct multiple test subsets to evaluate the stability and generalization ability of the model under different data distributions.

5. The method and system for analyzing power grid production technology transformation investment based on convolutional neural network according to claim 1, characterized in that: The convolutional neural network model construction and optimization process is as follows: first, the initial convolutional neural network model is constructed, and the network structure and kernel function are determined by selecting the appropriate number of layers, convolution kernel size, and pooling method according to the characteristics of the power grid technical transformation investment problem; the model calculation efficiency and prediction accuracy are improved by adjusting the hyperparameters and optimization algorithms.

6. The method and system for analyzing power grid production technology transformation investment based on convolutional neural network according to claim 1, characterized in that: Model fusion and prediction adopts a weighted fusion strategy. According to the prediction results and performance of multiple convolutional neural network models, the weighted values ​​of each model are determined to obtain the final fusion model. The fusion model is used to predict the investment benefits of power grid production technology transformation, and a scientific basis for investment decision-making is provided based on the prediction results.

Images