Sensitive meteorological factor extraction method based on multivariable unified information coefficient and used for micro-grid source and load power prediction
The multivariate unified information coefficient analysis method extracts sensitive meteorological factors in the microgrid, which solves the problem of failure to consider the impact of weather characteristics on power supply and load in the existing technology, and improves the accuracy and model generalization of the source load power prediction of microgrids ability.
Patent Information
- Application Number
- CN202311598314.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-11-28
- Publication Date
- 2024-12-13
AI Technical Summary
The existing technology fails to effectively consider the simultaneous impact of weather characteristics on the output and load power of distributed power of microgrids, resulting in poor generalization capabilities and overfitting of prediction models, and fails to extract sensitive meteorological factors to reduce the prediction dimension.
By using a method based on multivariate unified information coefficient, we can obtain and normalize the meteorological information and power and load power data in the microgrid area, and use mutual information coefficients to analyze the correlation between meteorological information and power and load power to extract sensitivity Meteorological factors reduce the prediction dimension.
The accuracy of microgrid source charge power prediction is improved, and variable redundancy is reduced by selecting high-correlation weather characteristics as input, and the generalization ability of the prediction model is enhanced.
Smart Images
Figure CN119134263A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of microgrid power source and load power prediction, and in particular to a sensitive meteorological factor extraction method based on multivariate unified information coefficient for microgrid source and load power prediction. Background Art
[0002] Distributed power sources in microgrids include wind power generation, photovoltaic power generation, ocean power generation and other power generation forms, and their output fluctuates randomly with changes in meteorological factors. With the progress of society, the types and numbers of active loads and weather-sensitive loads in microgrids are gradually increasing, and the load power also changes with changes in meteorological factors. Therefore, the source and load power of microgrids are uncertain and change with changes in meteorological factors. When the microgrid is running, it is necessary to optimize the scheduling according to the size of the source and load power of the microgrid at the same time and under the same environmental conditions. However, the area to which the microgrid belongs is relatively small, and the weather characteristics, surrounding environment, social factors and other factors in the area to which the microgrid belongs affect the output of distributed power sources and load power changes at the same time. If all influencing factors are input into the prediction model for training, the training set will be extremely large, affecting the generalization ability of the prediction model and prone to overfitting. Therefore, it is necessary to consider the correlation between various meteorological characteristics and source and load power at the same time, and obtain sensitive characteristic factors to reduce the dimension of the source and load power prediction input.
[0003] At present, the univariate correlation coefficient method is only used to extract the correlation between weather characteristics and the distributed power output and load power of microgrids, thereby reducing the dimension of input variables. The impact of weather characteristics on the distributed power output and load power of microgrids is not considered at the same time, and sensitive meteorological factors are extracted to reduce the prediction dimension.
[0004] Therefore, there is a need for a sensitive meteorological factor extraction method based on multivariate unified information coefficient for microgrid source and load power prediction. Summary of the invention
[0005] The main purpose of the present invention is to provide a sensitive meteorological factor extraction method based on multivariate unified information coefficient for microgrid source and load power prediction, so as to solve the problem in the prior art of lacking simultaneous consideration of the impact of weather characteristics on the microgrid distributed power output and load power.
[0006] To achieve the above object, the present invention provides a sensitive meteorological factor extraction method based on multivariate unified information coefficient for microgrid source and load power prediction, which specifically includes the following steps:
[0007] S1, obtain the data set D = {(P DG , P L ), H}, where H is the meteorological information of the microgrid area for a period of time, P DGis the historical data of power output in the microgrid in the same area and the same time period, P L It is the historical data of load power in the microgrid in the same area and the same time period.
[0008] S2, normalizes the meteorological information, power output and load power.
[0009] S3, based on the multivariate unified information coefficient I MV-UIC Conduct correlation analysis between meteorological information and power output and load power.
[0010] Furthermore, step S2 specifically includes the following steps:
[0011] S2.1, put the regional meteorological characteristic vector H on the X axis, expressed as X = [x i ], i = 1, 2, ... n, n is the sequence length; the power output and load power (P DG , P L ) is placed on the Y axis, expressed as P = [y i ], i=1, 2,…n.
[0012] S2.2, according to the uniform division method, the length of each segment after the X and Y axes are evenly divided into r and s segments is:
[0013]
[0014] Where: d x With d y are the partition unit lengths of the X-axis and Y-axis respectively; max With x min are the maximum and minimum values of the eigenvector X respectively; y max and ymin are the maximum and minimum values of the eigenvector P respectively; n 0.6 Represents the partition grid size, which is usually taken as 0.6 of the data volume.
[0015] S2.3, the ratio of the number of points falling into the corresponding grid to the total number of points is defined as the approximate probability density of the grid, and the marginal probability density p(x) of the meteorological information H when it is evenly divided into r grids is obtained. The power output and load power (P DG , P L ) is evenly divided into marginal probability densities p(y) under s grid partitions, and the data set D is divided into joint probability densities p(xy) under r×s grids.
[0016] Furthermore, step S3 specifically includes the following steps:
[0017] S3.1, Substitute p(x), p(y) and p(xy) into equation (1) to obtain the meteorological characteristic variable H and power supply power P DG and load power P L The mutual information coefficient I MI (r,s,(P DG , P L ), H);
[0018]
[0019] S3.2, use formula (2) to calculate the relationship between meteorological information, power output and load power:
[0020]
[0021] In the formula, log2(min(r, s)) is the normalization processing, min(r, s) is the minimum value of r and s, and B is the maximum number of grid divisions, that is, r, s<B.
[0022] The present invention has the following beneficial effects:
[0023] The present invention simultaneously considers the influence of weather characteristics on the output of distributed power sources and load power of microgrids, extracts sensitive meteorological factors, and improves prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] In order to more clearly illustrate the specific implementation of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the specific implementation or the prior art description. Obviously, the drawings described below are some implementations of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:
[0025] Figure 1 The flowchart of the sensitive meteorological factor extraction method based on multivariate unified information coefficient for microgrid source and load power prediction of the present invention is shown. DETAILED DESCRIPTION
[0026] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0027] like Figure 1 A sensitive meteorological factor extraction method based on multivariate unified information coefficient for microgrid source and load power prediction is shown, which specifically includes the following steps:
[0028] S1, obtain the data set D = {(P DG , P L ), H}, where H is the meteorological information of a certain period of time in the area where the microgrid is located (i.e., sensitive meteorological feature sequence), P DG is the historical data of power output in the microgrid in the same area and the same time period, P L It is the historical data of load power in the microgrid in the same area and the same time period.
[0029] S2, normalizes the meteorological information, power output and load power.
[0030] S3, based on the multivariate unified information coefficient I MV-UIC Conduct correlation analysis between meteorological information and power output and load power.
[0031] Specifically, in step S1, the meteorological information of the area where the microgrid is located for a certain period of time is obtained, including albedo, snow albedo, high cloud cover, low cloud cover, 2m relative humidity, 10m wind speed, 30m wind speed, 50m wind speed, 70m wind speed, temperature, wind direction, surface pressure, and visibility.
[0032] Specifically, step S2 includes the following steps:
[0033] S2.1, put the regional meteorological characteristic vector H on the X axis, expressed as X = [x i ], i = 1, 2, ... n, n is the sequence length; the power output and load power (P DG , P L ) is placed on the Y axis, expressed as P = [y i ], i=1, 2,…n.
[0034] S2.2, according to the uniform division method, the length of each segment after the X and Y axes are evenly divided into r and s segments is:
[0035]
[0036] Where: d x With d y are the partition unit lengths of the X-axis and Y-axis respectively; max With x min are the maximum and minimum values of the eigenvector X respectively; y max With y min are the maximum and minimum values of the eigenvector Y respectively; n 0.6 Represents the partition grid size, which is usually taken as 0.6 of the data volume.
[0037] S2.3, the ratio of the number of points falling into the corresponding grid to the total number of points is defined as the approximate probability density of the grid, and the marginal probability density p(x) of the meteorological information H when it is evenly divided into r grids is obtained. The power output and load power (P DG , P L ) is evenly divided into marginal probability densities p(y) under s grid partitions, and the data set D is divided into joint probability densities p(xy) under r×s grids.
[0038] Specifically, step S3 includes the following steps:
[0039] S3.1, Substitute p(x), p(y) and p(xy) into equation (1) to obtain the meteorological characteristic variable H and power supply power P DG and load power P L The mutual information coefficient I MI (r,s,(P DG , P L ), H);
[0040]
[0041] S3.2, use formula (2) to calculate the relationship between meteorological information, power output and load power:
[0042]
[0043] In the formula, log2(min(r, s)) is the normalization processing, min(r, s) is the minimum value of r and s, and B is the maximum number of grid divisions, that is, r, s<B.
[0044] Based on equations (1) and (2), the power output and load power of the microgrid and the meteorological environment I are calculated respectively. MV-UIC The relationship value, I MV-UIC The larger the value, the stronger the correlation between the corresponding meteorological characteristic variables and the output of distributed power sources and load power of the microgrid. The weather characteristic sequence with high correlation is selected from the meteorological characteristics as the input sequence of the state prediction model, and the remaining characteristics are filtered out to avoid problems such as too many variables and redundant calculations, thereby improving the prediction accuracy.
[0045] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by technicians in this technical field within the essential scope of the present invention should also fall within the protection scope of the present invention.
Claims
1. A sensitive meteorological factor extraction method based on multivariate unified information coefficient for microgrid source and load power prediction, characterized in that: The specific steps include: S1, obtain the data set D = {(P DG ,P L ),H}, where H is the meteorological information of the microgrid area for a period of time, P DG is the historical data of power output in the microgrid in the same area and the same time period, P L The historical data of load power in the microgrid in the same area and the same time period; S2, normalizes the meteorological information, power output and load power; S3, based on the multivariate unified information coefficient I MV-UIC Conduct correlation analysis between meteorological information and power output and load power.
2. According to claim 1, a sensitive meteorological factor extraction method based on multivariate unified information coefficient for microgrid source and load power prediction is characterized in that: Step S2 specifically includes the following steps: S2.1, put the regional meteorological characteristic vector H on the X axis, expressed as X = [x i ], i = 1, 2, ... n, n is the sequence length; the power output and load power (P DG ,P L ) is placed on the Y axis, expressed as Y = [y i ], i=1,2,…n; S2.2, according to the uniform division method, the length of each segment after the X and Y axes are evenly divided into r and s segments is: Where: d x With d y are the partition unit lengths of the X-axis and Y-axis respectively; max With x min are the maximum and minimum values of the eigenvector X respectively; y max With y min are the maximum and minimum values of the eigenvector Y respectively; n 0.6 Represents the partition grid size, usually taken as 0.6 of the data volume; S2.3, the ratio of the number of points falling into the corresponding grid to the total number of points is defined as the approximate probability density of the grid, and the marginal probability density p(x) of the meteorological information H when it is evenly divided into r grids is obtained. The power output and load power (P DG ,P L ) is evenly divided into marginal probability densities p(y) under s grid partitions, and the data set D is divided into joint probability densities p(xy) under r×s grids.
3. The sensitive meteorological factor extraction method based on multivariate unified information coefficient for microgrid source and load power prediction according to claim 1 is characterized in that: Step S3 specifically includes the following steps: S3.1, Substitute p(x), p(y) and p(xy) into equation (1) to obtain the meteorological characteristic variable H and power supply power P DG and load power P L The mutual information coefficient I MI (r,s,(P DG ,P L ),H); S3.2, use formula (2) to calculate the relationship between meteorological information, power output and load power: Where log2(min(r,s)) is the normalization process, min(r,s) is the minimum value of r and s, and B is the maximum number of grid divisions, that is, r, s <B。