A multiple interpolation method for distributed photovoltaic data based on chain equations

By adopting a multi-interpolation method based on chain equations in an intelligent distributed photovoltaic system, using Bayesian regression model and iterative filling mechanism, the data loss problem is solved, the accuracy and consistency of data processing is improved, and the error is reduced .

CN111949937BInactive Publication Date: 2024-11-12SOUTHEAST UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202010690310.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-07-17
Publication Date
2024-11-12
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In intelligent distributed photovoltaic systems, data loss caused by weather and equipment affects the accuracy of data processing, and existing multivariate interpolation methods are difficult to effectively solve this problem.

Method used

A distributed photovoltaic data multiple interpolation method based on chain equations is adopted, and the Bayesian regression model and iterative fill mechanism are used to gradually adjust the regression model and the number of interpolations until the setting requirements are met, and the final interpolation result is generated.

Benefits of technology

It improves the accuracy and consistency of data interpolation, reduces errors compared with traditional multiple interpolation methods, and provides more reliable data processing results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN111949937B_ABST
    Figure CN111949937B_ABST
Patent Text Reader

Abstract

The present invention discloses a distributed photovoltaic data multiple interpolation method based on chain equations, the steps are as follows: substituting distributed photovoltaic data with missing data into the interpolation process, adopting the interpolation method based on chain equations to obtain multiple different distributed photovoltaic interpolation data sets, and adopting the Bayesian regression model in the interpolation method; analyzing the statistical results of each distributed photovoltaic interpolation data set, and proceeding to the next step if the analysis result meets the set requirements; if the set requirements cannot be met, the regression model or the number of interpolation times needs to be adjusted until the analysis result meets the set requirements; finally, all distributed photovoltaic interpolation data sets are integrated to obtain the final interpolation result, and the result is used as the final distributed photovoltaic data value. The present invention uses the interpolation method of chain equations as the interpolation step in the multiple interpolation structure, thereby improving the error of the traditional multiple interpolation algorithm.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of distributed photovoltaics, and particularly designs a distributed photovoltaic data multiple interpolation method. Background Art

[0002] With the popularization of sensors and intelligent distributed photovoltaics, more and more intelligent applications such as intelligent fault detection systems have emerged. However, the application of intelligent systems must be based on data collection and processing technology. In the process of sensor data collection, data loss is often caused by weather, equipment and other reasons. Missing data will inevitably lead to adverse effects on data processing in the later stage. Therefore, processing missing data is a very important and necessary step for later data processing.

[0003] The interpolation methods for missing data are mainly divided into univariate interpolation and multivariate interpolation. Simple interpolation considers missing values ​​to be a constant, while multiple interpolation considers missing values ​​to be variables that conform to a certain probability distribution. Therefore, for multiple interpolation, the interpolation results can reflect the randomness of the variables, and multivariate interpolation can often obtain more accurate results than univariate interpolation. Summary of the invention

[0004] In order to solve the technical problems mentioned in the above background technology, the present invention proposes a distributed photovoltaic data multiple interpolation method based on chain equations.

[0005] In order to achieve the above technical objectives, the technical solution of the present invention is:

[0006] A distributed photovoltaic data multiple interpolation method based on chain equations includes the following steps:

[0007] (1) Substituting the distributed photovoltaic data with missing data into the interpolation process, a chain equation-based interpolation method is used to obtain multiple different distributed photovoltaic interpolation data sets. The regression model in the interpolation method adopts the Bayesian regression model;

[0008] (2) Analyze the statistical results of each distributed photovoltaic interpolation data set. If the analysis results meet the set requirements, proceed to the next step; if not, adjust the regression model or the number of interpolation times and return to step (1) until the analysis results meet the set requirements;

[0009] (3) All distributed photovoltaic interpolation data sets are integrated to obtain the final interpolation result, and this result is used as the final distributed photovoltaic data value.

[0010] Furthermore, the specific process of step (1) is as follows:

[0011] (1-1) Assume that the data collected by the distributed photovoltaic sensors is a data set matrix X with missing data;

[0012] (1-2) Fill each missing value with the initial imputation value;

[0013] (1-3) Sort the features of matrix X according to the missing rate;

[0014] (1-4) Start the iteration, first take the feature S with the smallest missing rate, and divide the matrix X into 4 parts: the missing part of feature S is defined as y(s)mis, the non-missing part of feature S is defined as y(s)obs, the part of all variables except feature S corresponding to the missing index of feature S is defined as X(s)mis, and the part of all variables except feature S corresponding to the non-missing index of feature S is defined as X(s)obs;

[0015] (1-5) Fit y(s)obs and X(s)obs through the Bayesian regression model, and then predict y(s)mis corresponding to X(s)mis;

[0016] (1-6) Update the predicted value y(s)mis to the matrix X, then take the feature S with the second smallest missing rate, loop all the features S with missing values, and fill the predicted value of the Bayesian regression model into the matrix X;

[0017] (1-7) Compare the updated matrix X with the initial matrix X to determine whether the convergence value of the convergence function meets the set threshold. If not, return to step (1-4) to start the next iteration until the convergence value meets the set threshold or the number of iterations reaches the set maximum number of iterations.

[0018] Furthermore, in step (1-2), the mean or median of the original data set is used as the initial interpolation value.

[0019] Furthermore, in step (2), each distributed photovoltaic interpolation data set is analyzed using the ologit regression analysis method based on complex sampling conditions.

[0020] Furthermore, in step (3), the average value of all distributed photovoltaic interpolation data sets is taken to obtain the final interpolation result.

[0021] The beneficial effects brought by adopting the above technical solution are:

[0022] The present invention is an improved chain equation interpolation method based on a multiple interpolation framework. Different from the traditional multiple interpolation, the present invention uses the chain equation interpolation method as the interpolation step in the multiple interpolation structure, thereby improving the error of the traditional interpolation algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a framework diagram of multiple interpolation of the present invention;

[0024] Figure 2 It is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0025] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings.

[0026] The present invention designs a distributed photovoltaic data multiple interpolation method based on chain equations, such as Figure 1-2 As shown, the steps are as follows:

[0027] Step 1: Substitute the distributed photovoltaic data with missing data into the interpolation process, and use the interpolation method based on chain equations to obtain m different distributed photovoltaic interpolation data sets. The regression model in the interpolation method adopts the Bayesian regression model;

[0028] Step 2: Analyze the statistical results of each distributed photovoltaic interpolation data set. If the analysis results meet the set requirements, proceed to the next step; if not, adjust the regression model or the number of interpolation times and return to step 1 until the analysis results meet the set requirements;

[0029] Step 3: Integrate all distributed photovoltaic interpolation data sets to obtain the final interpolation result, and use this result as the final distributed photovoltaic data value.

[0030] In this embodiment, preferably, the above step 1 is implemented using the following preferred solution:

[0031] 1-1. Assume that the data collected by the distributed photovoltaic sensor is a data set matrix X with missing data;

[0032] 1-2. Fill each missing value with the initial interpolation value;

[0033] 1-3. Sort the features of matrix X according to the missing rate;

[0034] 1-4. Start the iteration, first take the feature S with the smallest missing rate (i.e., the column of matrix X), and divide the matrix X into 4 parts: the missing part of feature S is defined as y(s)mis, the non-missing part of feature S is defined as y(s)obs, the part of all variables except feature S that corresponds to the missing index of feature S is defined as X(s)mis, and the part of all variables except feature S that corresponds to the non-missing index of feature S is defined as X(s)obs;

[0035] 1-5. Fit y(s)obs and X(s)obs through the Bayesian regression model, and then predict y(s)mis corresponding to X(s)mis;

[0036] 1-6. Update the predicted value y(s)mis to the matrix X, then take the feature S with the second smallest missing rate, loop all the features S with missing values, and fill the predicted value of the Bayesian regression model into the matrix X;

[0037] 1-7. Compare the updated matrix X with the initial matrix X to determine whether the convergence value of the convergence function meets the set threshold. If not, return to step 1-4 to start the next iteration until the convergence value meets the set threshold or the number of iterations reaches the set maximum number of iterations γ.

[0038] In this embodiment, preferably, in the above steps 1-2, the mean or median of the original data set is used as the initial interpolation value.

[0039] In this embodiment, preferably, in the above step 2, an ologit regression analysis method based on complex sampling conditions is used to analyze each distributed photovoltaic interpolation data set.

[0040] In this embodiment, preferably, in the above step 3, the average value of all distributed photovoltaic interpolation data sets is taken to obtain the final interpolation result.

[0041] The embodiments are only for illustrating the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the present invention.

Claims

1. A distributed photovoltaic data multiple interpolation method based on chain equations, characterized in that: The following steps are involved: (1) Substituting the distributed photovoltaic data with missing data into the interpolation process, a chain equation-based interpolation method is used to obtain multiple different distributed photovoltaic interpolation data sets. The regression model in the interpolation method adopts the Bayesian regression model; (2) Analyze the statistical results of each distributed photovoltaic interpolation data set. If the analysis results meet the set requirements, proceed to the next step; if not, adjust the regression model or the number of interpolation times and return to step (1) until the analysis results meet the set requirements; (3) All distributed photovoltaic interpolation data sets are integrated to obtain the final interpolation result, and this result is used as the final distributed photovoltaic data value; The specific process of step (1) is as follows: (1-1) Assume that the data collected by the distributed photovoltaic sensors is a data set matrix X with missing data; (1-2) Fill each missing value with the initial imputation value; (1-3) Sort the features of matrix X according to the missing rate; (1-4) Start the iteration, first take the feature S with the smallest missing rate, and divide the matrix X into 4 parts: the missing part of feature S is defined as y(s)mis, the non-missing part of feature S is defined as y(s)obs, the feature part of matrix X except S is defined as X(s)mis, and the part of all variables except feature S that corresponds to the non-missing index of feature S is defined as X(s)obs; (1-5) Fit y(s)obs and X(s)obs through the Bayesian regression model, and then predict y(s)mis corresponding to X(s)mis; (1-6) Update the predicted value y(s)mis to the matrix X, then take the feature S with the second smallest missing rate, loop all the features S with missing values, and fill the predicted value of the Bayesian regression model into the matrix X; The specific process of the above step (2) is as follows: Compare the updated matrix X with the initial matrix X to determine whether the convergence value of the convergence function meets the set threshold. If not, return to step (1-4) to start the next iteration until the convergence value meets the set threshold or the number of iterations reaches the set maximum number of iterations γ; In the step (1-2), the mean or median of the original data set is used as the initial interpolation value; In the step (2), each distributed photovoltaic interpolation data set is analyzed using an ologit regression analysis method based on complex sampling conditions; In the step (3), the average value of all distributed photovoltaic interpolation data sets is taken to obtain the final interpolation result.