Auxiliary analysis method for observable, measurable and controllable digital investment of power distribution network

Abstract

The invention relates to an auxiliary analysis method for observable, measurable and controllable digital investment of a power distribution network, and the method comprises the steps: carrying out the sampling of model input data through employing a Latin hypercube sampling method based on a conventional digital investment effect analysis model of the power distribution network, and obtaining a group of input sample data sets; different effect indexes are calculated through a traditional power distribution network digital investment effect analysis model, so that a power distribution network digital investment related data set is formed; processing the data by adopting a z-score standardization method, and inputting the processed data into a feature selection model; performing feature selection on the multiple different effect targets to obtain a series of investment effect index high-correlation feature sets; and finally, a decision maker is assisted and supported to make an investment decision plan by using the high-correlation feature set. The method has the advantages that sample data representativeness is ensured, the model stability is good, and the decision-making efficiency and the decision-making accuracy are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of distribution network, and in particular relates to an auxiliary analysis method for observable, measurable and controllable digital investment in distribution network. Background Art

[0002] In the process of investment decision-making, power grid enterprises need to respond quickly to emergencies such as policy changes and market fluctuations. Power grid investment involves a variety of complex variables, and the amount of information required for decision-making is large. There may be complex nonlinear relationships and interactions between various factors. Traditional power grid investment analysis mainly relies on complex mathematical and physical models, including considerations such as the dispatch of power systems and safe and stable operation. These models simulate the operating status of the power grid by establishing detailed mathematical equations, thereby evaluating the feasibility and profitability of different investment plans. The Monte Carlo simulation method is one of the commonly used techniques in power grid investment risk analysis. It evaluates system performance through a large number of random sample simulations. Although it can handle uncertainty and provide reliable risk assessment, it requires a large amount of sample data and computing resources. The simulation process is complex and time-consuming, making it difficult to make quick decisions. Big data analysis methods are also widely used in power grid investment optimization. This method mines and analyzes historical data to identify key factors that affect investment returns and builds a prediction model based on these factors. However, it relies on a large amount of historical data, and the data collection and processing process is complex. The analysis results are easily affected by data noise and bias. The influence of the difference may lead to inaccurate predictions; machine learning technology has gradually been applied in power grid investment decisions. By training machine learning models, key factors affecting investment benefits can be automatically identified to assist decision makers in formulating investment strategies. Although it has advantages in automation and accuracy, model training requires a large amount of labeled data, and the data preparation cost is high. At the same time, the black box characteristics of the model lead to poor interpretability and difficulty in obtaining the full trust of decision makers; In summary, although the application of existing technologies in power grid enterprise investment decisions has certain advantages, there are still some obvious shortcomings compared to the needs of rapid response and efficient decision-making. Although traditional mathematical and physical models provide detailed system analysis, due to their complexity, the model solution speed is slow and cannot respond quickly to decision-making needs in emergency situations; Therefore, these existing technologies are difficult to quickly and accurately identify the key factors of investment effectiveness in emergency situations, and the decision-making efficiency and accuracy are subject to certain limitations; Therefore, it is very necessary to provide an auxiliary analysis method for observable, measurable and controllable digital investment in distribution networks that ensures the representativeness of sample data, good model stability, improves decision-making efficiency, and improves decision-making accuracy. Summary of the invention

[0003] The purpose of the present invention is to overcome the shortcomings of the prior art and to provide an auxiliary analysis method for observable, measurable and controllable digital investment in distribution networks that ensures representativeness of sample data, good model stability, improved decision-making efficiency and improved decision-making accuracy.

[0004] The object of the present invention is achieved by providing an auxiliary analysis method for observable, measurable and controllable digital investment in a distribution network, the method comprising the following steps:

[0005] Step 1: Use the Latin hypercube sampling method to sample the model input data to obtain a set of input sample data sets;

[0006] Step 2: Calculate different effectiveness indicators through the traditional distribution network digitalization investment effectiveness analysis model to form a distribution network digitalization investment related data set;

[0007] Step 3: Use the z-score standardization method to process the data and input the processed data into the feature selection model;

[0008] Step 4: Perform feature selection for multiple different performance targets to obtain a set of highly correlated features for a series of investment performance indicators;

[0009] Step 5: Finally, use the highly correlated feature set to assist decision makers in making investment decision plans.

[0010] The Latin hypercube sampling method in step 1 is specifically as follows: Assuming that there are d parameters in total, the value range of each parameter is [a i ,b i ](i=1,2,...,d), the number of samples is N, for each parameter x i The jth sample point x ij The calculation formula is: Where P ij is the parameter x i The interval number in the jth sample is a random sequence from 0 to N-1; U ij is a uniformly distributed random number in the interval [0, 1).

[0011] The step 1 of sampling the model input data using the Latin hypercube sampling method to obtain a set of input sample data sets includes the following steps:

[0012] Step 1.1: First, define the value range of each parameter;

[0013] Step 1.2: Then, the value range of each parameter is evenly divided into N intervals;

[0014] Step 1.3: Then generate interval numbers P for each parameter ij For each parameter x i Each sample point j generates a random number U in the interval [0, 1) ij ;

[0015] Step 1.4: Finally, use formula (1) to calculate the value of each sample point. LHS is suitable for sampling samples with complex types, large quantities and large differences in characteristics.

[0016] The z-score standardization method used in step 3 is used to process the data. Specifically, when normalizing the data, the present invention uses the z-score standardization method to subtract the mean from the data value of each feature and divide it by the standard deviation: Where: z is the original value of the data; μ is the mean of the data; σ is the standard deviation of the data.

[0017] In step 4, feature selection is performed for multiple different performance targets respectively. Specifically, the present invention adopts lasso Lasso to select features related to distribution network digital investment. Lasso regression adds an L1 regularization term with a penalty coefficient λ to the traditional linear regression loss function, and its cost function can be expressed as: The model solves the problem of minimizing the cost function Cost(β) and obtains the size of β: Among them, β is the regression coefficient; y i is the actual value of the i-th sample data; x ij is the jth feature of the i-th sample; λ is the penalty coefficient, which is used to control the strength of feature selection; Lasso regression achieves feature selection by shrinking the regression coefficients of unimportant features to 0.

[0018] The beneficial effects of the present invention are as follows: the present invention is an auxiliary analysis method for observable, measurable and controllable digital investment in distribution networks. Digital investment in distribution networks has certain potential benefits. The present invention aims at the situation where traditional investment analysis methods of power grids are difficult to cope with and require rapid decision-making. The present invention provides a method for power grid companies to respond quickly and make efficient decisions in emergency situations. During use, the present invention quickly identifies key influencing factors in digital investment in distribution networks through a feature selection method, which can effectively improve the efficiency and accuracy of decision-making and meet the rapid response needs of power grid companies in emergency situations; through a highly correlated feature set, direct and targeted guidance is provided to decision makers to assist them in making efficient investment decisions; the present invention has the advantages of ensuring the representativeness of sample data, good model stability, improving decision-making efficiency, and improving decision-making accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0020] The present invention will be further described below in conjunction with the accompanying drawings.

[0021] Example 1

[0022] like Figure 1 As shown in the figure, an auxiliary analysis method for observable, measurable and controllable digital investment in distribution network, in the process of digital investment in distribution network, it is necessary to analyze the distribution network by constructing a traditional mathematical model, which involves a complex power system operation mechanism, and also includes considerations such as power system scheduling, safe and stable operation, etc. The model solving speed is slow, and it is difficult to respond to the decision makers' needs for fast and efficient decision making; therefore, the method of the present invention is based on the traditional distribution network digital investment effectiveness analysis model, and includes the following steps:

[0023] Step 1: Use the Latin hypercube sampling method to sample the model input data to obtain a set of input sample data sets;

[0024] Step 2: Calculate different effectiveness indicators through the traditional distribution network digitalization investment effectiveness analysis model to form a distribution network digitalization investment related data set;

[0025] Step 3: Use the z-score standardization method to process the data and input the processed data into the feature selection model;

[0026] Step 4: Perform feature selection for multiple different performance targets to obtain a set of highly correlated features for a series of investment performance indicators;

[0027] Step 5: Finally, use the highly correlated feature set to assist decision makers in making investment decision plans.

[0028] The present invention is an auxiliary analysis method for observable, measurable and controllable digital investment in distribution networks. The present invention aims to solve the problems of slow response speed, low decision-making efficiency and insufficient accuracy faced by power grid companies in the process of investment decision-making in emergency situations. In use, the present invention aims to solve the problems of complex and time-consuming calculations of traditional mathematical and physical models. The Latin hypercube sampling method is used to generate input sample data sets to ensure the representativeness of sample data. At the same time, the zscore standardization is used to process data to eliminate data dimension differences and improve the stability of the model. The present invention uses a feature selection model to quickly identify the key factors that have the greatest impact on investment results, avoid wasting time on irrelevant or minor factors, and thus improve the accuracy of the model. High decision-making efficiency; in view of the disadvantage that a large amount of historical data is easily affected by data noise and deviation, the present invention constructs a distribution network digital investment-related data set, uses a feature selection method to extract a highly correlated feature set, and ensures the accuracy and reliability of the analysis results; the present invention provides clear and targeted decision-making guidance by directly analyzing the highly correlated feature set, and improves the transparency and interpretability of the decision; therefore, the present invention quickly identifies key influencing factors through a feature selection method, provides efficient and accurate decision support, and meets the rapid response needs of power grid companies in emergency situations; the present invention has the advantages of ensuring the representativeness of sample data, good model stability, improving decision-making efficiency, and improving decision-making accuracy.

[0029] Example 2

[0030] like Figure 1 As shown, the auxiliary analysis method of the observable, measurable and controllable digital investment of the distribution network comprises the following steps:

[0031] Step 1: Use the Latin hypercube sampling method to sample the model input data to obtain a set of input sample data sets;

[0032] Step 2: Calculate different effectiveness indicators through the traditional distribution network digitalization investment effectiveness analysis model to form a distribution network digitalization investment related data set;

[0033] Step 3: Use the z-score standardization method to process the data and input the processed data into the feature selection model;

[0034] Step 4: Perform feature selection for multiple different performance targets to obtain a set of highly correlated features for a series of investment performance indicators;

[0035] Step 5: Finally, use the highly correlated feature set to assist decision makers in making investment decision plans.

[0036] The Latin hypercube sampling method in step 1 is specifically as follows: in order to analyze the effectiveness of distribution network digitalization investment, it is necessary to generate multiple input samples in a multidimensional parameter space in order to fully understand the performance of the system under different conditions; it is not feasible to directly evaluate all possible input combinations in actual operation, so it is necessary to sample possible input change combinations; the Latin hypercube sampling (LHS) method is selected because it can evenly cover the entire parameter space while ensuring a limited sample size, improve sampling efficiency and representativeness of the results, reduce computational overhead, and ensure the accuracy and diversity of the results; through this method, more reliable analysis results can be obtained, providing solid data support for the optimization and investment decisions of the distribution network. The Latin hypercube sampling LHS method is a technique for generating samples in a multidimensional space to ensure that the samples in each dimension are evenly distributed. The specific mathematical formula and steps are as follows: Assume that there are d parameters in total, and the value range of each parameter is [a i ,b i ](i=1,2,...,d), the number of samples is N, for each parameter x i The jth sample point x ij The calculation formula is: Where P ij is the parameter x i The interval number in the jth sample is a random sequence from 0 to N-1; U ij is a uniformly distributed random number in the interval [0, 1).

[0037] The step 1 of sampling the model input data using the Latin hypercube sampling method to obtain a set of input sample data sets includes the following steps:

[0038] Step 1.1: First, define the value range of each parameter;

[0039] Step 1.2: Then, the value range of each parameter is evenly divided into N intervals;

[0040] Step 1.3: Then generate interval numbers P for each parameter ij For each parameter x i Each sample point j generates a random number U in the interval [0, 1) ij ;

[0041] Step 1.4: Finally, use formula (1) to calculate the value of each sample point. LHS is suitable for sampling samples with complex types, large quantities and large differences in characteristics.

[0042] The method of using z-score standardization in step 3 to process the data is specifically as follows: when performing data normalization, the present invention uses the z-score standardization method to subtract the mean from the data value of each feature and divide it by the standard deviation. The z-score standardization can convert the feature data into a set of standard normal distributions with a mean of zero and a standard deviation of one. By scaling features of different orders of magnitude to the same order of magnitude, the influence of dimensional differences on model calculations is eliminated, so that the feature selection algorithm can find the optimal solution faster and more stably, improve the efficiency of calculation, and prevent individual variables from dominating the feature selection due to a large range of numerical changes: Where: z is the original value of the data; μ is the mean of the data; σ is the standard deviation of the data.

[0043] In step 4, feature selection is performed for multiple different performance targets respectively. Specifically, the present invention adopts lasso Lasso to select features related to distribution network digital investment. Feature selection can effectively filter out noise and redundant information in the data, retain the most valuable variables for investment decisions, improve the accuracy and reliability of analysis, and systematically analyze and screen out key factors that have a significant impact on investment results, so that leaders can clearly understand which factors will bring significant changes in results by adjusting them. Lasso regression adds an L1 regularization term with a penalty coefficient λ to the traditional linear regression loss function, and its cost function can be expressed as: The model solves the problem of minimizing the cost function Cost(β) and obtains the size of β: Among them, β is the regression coefficient; y i is the actual value of the i-th sample data; xij is the jth feature of the i-th sample; λ is the penalty coefficient, which is used to control the strength of feature selection; Lasso regression achieves feature selection by shrinking the regression coefficients of unimportant features to 0.

[0044] The present invention is an auxiliary analysis method for observable, measurable and controllable digital investment in distribution networks. The present invention aims to solve the problems of slow response speed, low decision-making efficiency and insufficient accuracy faced by power grid companies in the process of investment decision-making in emergency situations. In use, the present invention aims to solve the problems of complex and time-consuming calculations of traditional mathematical and physical models. The Latin hypercube sampling method is used to generate input sample data sets to ensure the representativeness of sample data. At the same time, the zscore standardization is used to process data to eliminate data dimension differences and improve the stability of the model. The present invention uses a feature selection model to quickly identify the key factors that have the greatest impact on investment results, avoid wasting time on irrelevant or minor factors, and thus improve the accuracy of the model. High decision-making efficiency; in view of the disadvantage that a large amount of historical data is easily affected by data noise and deviation, the present invention constructs a distribution network digital investment-related data set, uses a feature selection method to extract a highly correlated feature set, and ensures the accuracy and reliability of the analysis results; the present invention provides clear and targeted decision-making guidance by directly analyzing the highly correlated feature set, and improves the transparency and interpretability of the decision; therefore, the present invention quickly identifies key influencing factors through a feature selection method, provides efficient and accurate decision support, and meets the rapid response needs of power grid companies in emergency situations; the present invention has the advantages of ensuring the representativeness of sample data, good model stability, improving decision-making efficiency, and improving decision-making accuracy.

Claims

1. An auxiliary analysis method for observable, measurable and controllable digital investment in distribution network, characterized by: The method comprises the following steps: Step 1: Use the Latin hypercube sampling method to sample the model input data to obtain a set of input sample data sets; Step 2: Calculate different effectiveness indicators through the traditional distribution network digitalization investment effectiveness analysis model to form a distribution network digitalization investment related data set; Step 3: Use the z-score standardization method to process the data and input the processed data into the feature selection model; Step 4: Perform feature selection for multiple different performance targets to obtain a set of highly correlated features for a series of investment performance indicators; Step 5: Finally, use the highly correlated feature set to assist decision makers in making investment decision plans.

2. The auxiliary analysis method for observable, measurable and controllable digital investment in distribution network according to claim 1, characterized in that: The Latin hypercube sampling method in step 1 is specifically as follows: Assuming that there are d parameters in total, the value range of each parameter is [a i ,b i ](i=1,2,...,d), the number of samples is N, for each parameter x i The jth sample point x ij The calculation formula is: Where P ij is the parameter x i The interval number in the jth sample is a random sequence from 0 to N-1; U ij is a uniformly distributed random number in the interval [0, 1).

3. The auxiliary analysis method for observable, measurable and controllable digital investment in distribution network according to claim 2, characterized in that: The step 1 of sampling the model input data using the Latin hypercube sampling method to obtain a set of input sample data sets includes the following steps: Step 1.1: First, define the value range of each parameter; Step 1.2: Then, the value range of each parameter is evenly divided into N intervals; Step 1.3: Then generate interval numbers P for each parameter ij For each parameter x i Each sample point j generates a random number U in the interval [0, 1) ij ; Step 1.4: Finally, use formula (1) to calculate the value of each sample point. LHS is suitable for sampling samples with complex types, large quantities and large differences in characteristics.

4. The auxiliary analysis method for observable, measurable and controllable digital investment in distribution network according to claim 1, characterized in that: The z-score standardization method used in step 3 is used to process the data. Specifically, when normalizing the data, the present invention uses the z-score standardization method to subtract the mean from the data value of each feature and divide it by the standard deviation: Where: z is the original value of the data; μ is the mean of the data; σ is the standard deviation of the data.

5. The auxiliary analysis method for observable, measurable and controllable digital investment in distribution network according to claim 1, characterized in that: In step 4, feature selection is performed for multiple different performance targets respectively. Specifically, the present invention adopts lasso Lasso to select features related to distribution network digital investment. Lasso regression adds an L1 regularization term with a penalty coefficient λ to the traditional linear regression loss function, and its cost function can be expressed as: The model solves the problem of minimizing the cost function Cost(β) and obtains the size of β: Among them, β is the regression coefficient; y i is the actual value of the i-th sample data; x ij is the jth feature of the i-th sample; λ is the penalty coefficient, which is used to control the strength of feature selection; Lasso regression achieves feature selection by shrinking the regression coefficients of unimportant features to 0.

Images