CN111581883B - Method for calculating and predicting load on automation device - Google Patents
Method for calculating and predicting load on automation device Download PDFInfo
- Publication number
- CN111581883B CN111581883B CN202010387688.9A CN202010387688A CN111581883B CN 111581883 B CN111581883 B CN 111581883B CN 202010387688 A CN202010387688 A CN 202010387688A CN 111581883 B CN111581883 B CN 111581883B
- Authority
- CN
- China
- Prior art keywords
- data
- period
- load
- fault
- time
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/04—Power grid distribution networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Probability & Statistics with Applications (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method for calculating and predicting a load on an automation device, which comprises the steps of collecting and calculating the load, storing historical data and predicting the load. When the automatic device normally operates, load data are collected in real time, invalid data are identified, the average value of the load within a period of time is rapidly calculated through a weighting method, and the reliability of the load data is guaranteed. Meanwhile, historical data samples are periodically stored according to the periodicity of time and the number of load data, and the samples can be conveniently extracted for analysis and prediction in case of faults. When a fault occurs and load transfer is needed, a matching history period is preliminarily screened out through load analysis at the moment of the fault. And further screening out the optimal matching period according to the load data trend in a plurality of time before the fault by combining a recursive thought and a clustering analysis method. The invention can reasonably predict the fault load, thereby finishing the accurate operation of the automatic device for specially supplying the load, ensuring the resource maximization of the power supply system and improving the power supply stability and reliability.
Description
Technical Field
The invention relates to the technical field of power system relay protection, in particular to a method for calculating and predicting load on an automation device.
Background
With the development of power system automation, more and more automation devices are applied in the system, so that the stability and reliability of power supply are rapidly improved. In the prior art, automatic power supply is often recovered by using an automatic device, wherein measures such as connection and disconnection, overload and load reduction are generally adopted for processing loads.
However, the development of the economic society is increasingly deep, the dependence of life and work of people on electric power is rapidly increased, and the requirement on the stability of power supply is gradually increased. When various automation strategies are researched, further research is needed on calculation and prediction of loads, and the method is favorable for maximum recovery and stability maintenance when power supply is recovered, so that the efficiency of the automation device is improved, and the power supply in the maximum range is stable, safe and reliable.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a method for calculating and predicting a load on an automatic device, which realizes the maximum recovery of power supply and maintains the stability of the power supply according to the calculation and prediction of the load, realizes the calculation and prediction of the load by using a computer algorithm, and is matched with other recovery steps to fully improve the efficiency of transfer.
One technical scheme for achieving the above purpose is as follows: a method of load calculation and prediction on an automated device, comprising the steps of:
(3) analyzing the time data of the fault time to find a period set similar to the trend of the fault time; an approximate model is built by the fault moment and the recent historical data in the sample,
in a history database
In the same time as the fault time in each period, namely the data of the t-2 th column in the array, and respectively calculating respective coefficients according to the data of the t-1 th column
For all model coefficients C d ＝(C 0 ,C 1 … D-1 ,C t ) Carrying out data processing according to the principle of cluster analysis to find out an optimal cluster combination; these and fault time model coefficients C t The period corresponding to the similar model coefficient is the period similar to the load data trend at the fault moment;
according to the previous screening, the historical data corresponding to the matching cycle number is recombined and analyzed;
respectively establishing models for a plurality of data segments before the fault, and listing out the fault time and n model coefficients before the fault time, wherein the model coefficients are represented as:
they can represent the comprehensive trend of each period load data in a period before the fault moment;
firstly, a cycle with the highest trend coincidence degree with the fault cycle load data needs to be found, namely, the cycle with the closest model coefficient with the fault cycle is found; a row closest to each data in the last row of the two-dimensional array needs to be found, and the corresponding row number (cycle number) is recorded; for the analysis and processing of the two-dimensional array, the invention adopts a recursive method to screen layer by layer, and adopts clustering to analyze each layer until finding the periodicity which is most matched with the fault period, and reasonably predicting the load according to the periodicity;
calculating the corresponding time of the current fault in the period, and calculating the model coefficients of a plurality of backward-deducing times as follows;
……
according to the model coefficient C _ D ^ (t-2) at the fault occurrence time, the subsequent model coefficients are calculated and respectively:
……
the maximum value of the load in a period of time in the future is predicted according to the load value at the fault moment, and the amount of the special supply load and the specific scheme are determined according to the maximum load value and the maximum available load value, so that the accurate switching of the load transfer is realized.
Furthermore, the calculation of the average value of the real-time load data is realized by introducing a weight method;
meanwhile, invalid data are identified and processed in the data processing process; when the sampled data is compared with the previous data, the fluctuation is much larger than the maximum fluctuation value of the previous M data, namely max a (I a - I a-1 ) (a ═ 1,2,3, … …, M), determining that the data is invalid data;
when invalid data is encountered, the processing method mainly refers to the values of the previous moment and the next moment and takes the average value to replace, namely
The average value of the load at two moments in time is given at the next moment of invalid data, as follows:
further, analyzing the load data of the fault time point, firstly calculating a fault time model coefficient and a model coefficient at the same time of each period, and secondly finding out the period number similar to the trend of the fault time by adopting a clustering analysis method;
randomly taking K points as initial focusing center and then matching model coefficient C d Calculating the nearest clustering centers of all the points in the array, and dividing the nearest clustering centers into corresponding clusters to obtain K initial clusters, wherein the K initial clusters are expressed as f (x) (x is 1,2, … …, K);
finding out the point nearest to all the numerical values in the cluster among the K clusters as a new cluster center, namely obtaining the point by calculation
Min＝Min(f(x))
Max＝Max(f(x))
The new core is
Repeat pair model coefficient C d Calculating the nearest clustering centers of all the points in the array, and dividing the nearest clustering centers into corresponding clusters to obtain K new clusters, wherein the K new clusters are expressed as l (x) (x is 1,2, … …, K);
continuously repeating the process, performing closest point analysis on all numerical values in the K clusters, finding a new clustering center, and re-dividing the clusters of the whole array according to the new clustering center; until the latest convergence coincides with the previous convergence, the cluster is the optimized cluster combination, namely the cycle number corresponding to the cluster represents the cycle similar to the trend of the fault moment.
Further, comprehensively analyzing data in a period of time before the fault moment, namely analyzing the trend of the load data in a period of time before the fault moment, and finding a period matched with the trend in a period of time in the fault period;
respectively establishing models for a plurality of data segments before the fault, and finding out the period with similar coefficients, namely the period with the most matched trend in the period, through coefficient comparison; the method can be converted into analysis of a two-dimensional array, and needs to combine a recursive idea and a clustering analysis method;
firstly, respectively analyzing each column of the two-dimensional array, randomly taking 2 points for each column as initial clustering centers, repeatedly calculating by the clustering analysis method until finding out clusters corresponding to each column, and expressing the 2(n +1) clusters as f t-2-n (x)、f t-1-n (x)、……、f t-2 (x) (wherein x is 1, 2);
reading clusters where all columns of data of the last row are located, wherein the clusters are (n +1) in total and are respectively represented as f t-2-n (1)、f t-1-n (1)、……、f t-2 (1) (ii) a Reading out the corresponding cycle number of the data in the clusters to obtain n +1 groups of cycle numbers, namely d t-2-n (1)、d t-1-n (1)、……、d t-2 (1)；
By pair d t-2-n (1)、d t-1-n (1)、……、d t-2 (1) Carrying out next analysis, and finding out the intersection of the n +1 groups of the cycle numbers to find out the matching cycle number screened this time;
counting the number of cycles in all the clusters by using a counting method, wherein the number of cycles with the largest count is the maximum intersection number, and the number of cycles is a fault cycle matching item after primary screening; the specific method comprises the following steps:
counting each period number in the i-1 period numbers preliminarily matched in the previous step respectively; that is, it is calculated at d from the 1 st to i-1 t-2-n (1)、d t-1-n (1)、……、d t-2 (1) The number of occurrences in the n +1 arrays;
if the count of the z-th cycle is n +1, the number of cycles is saved in time n+1 (i) In the array, if the cycle count is n, the number of cycles is saved in time n (i) In an array; and so on, if the repeated count value of the period number is stored in the array corresponding to the maximum value, d is a matching item 2 The model coefficients corresponding to the period are rearranged into a new two-dimensional array C z][n+1]；
The number of rows in the two-dimensional array is unchanged and is still n +1 rows; the number of lines is reduced to z (z-1 is a plurality of numerical values of the matching period after the previous round of screening);
repeating the steps, continuously analyzing each row of the two-dimensional array, taking two clustering centers for clustering analysis, and carrying out periodicity screening on n +1 clusters until the analysis is finished when the screened periodicity is 1, namely, completing the matching when z is 2;
and finding out the optimal matching period through the steps, reading the number of the periods, and finding out the original historical data of the corresponding row as the optimal matching historical data.
Compared with the prior art, the method for calculating and predicting the load on the automatic device has the following beneficial effects:
1. the load is calculated by adopting a half-hour average value through a weight method, so that the influence of errors on data is effectively avoided, the burden of the operation speed is effectively reduced, and the stability of the automatic operation of the electric power is improved.
2. The load prediction adopts a recursion method and combines with cluster analysis, so that the load prediction can be expressed through an automatic system language, and the optimal prediction result can be quickly found, thereby realizing automatic and accurate switching of the load.
Drawings
FIG. 1 is a logic diagram of the present invention for real-time load handling and forecasting;
FIG. 2 is a schematic diagram of a load prediction method according to the present invention;
FIG. 3 is a schematic modeling diagram of the load handling of the present invention.
Detailed Description
In order to better understand the technical solution of the present invention, the following detailed description is made by specific examples:
the present invention is further illustrated by the following description in conjunction with the accompanying drawings and the specific embodiments, it is to be understood that these examples are given solely for the purpose of illustration and are not intended as a definition of the limits of the invention, since various equivalent modifications will occur to those skilled in the art upon reading the present invention and fall within the limits of the appended claims.
Referring to fig. 1, the automation device collects and processes the in-station load data in real time through the collection unit. First, assuming we read N load data in half an hour, then the average is:
wherein the content of the first and second substances,
mean value of load data at time d and time t, I i (i is more than or equal to 1 and less than or equal to N) refers to the ith data value read half an hour before the time t.
The weight is decomposed by adopting a weight method to obtain,
wherein the content of the first and second substances,
the date is d, and the time is before t
Average of load data in minutes, I 0 Then it means
The first data read half an hour before the moment.
Is a weight value.
By the method, calculation of cyclic summation can be avoided, the average value at the moment can be rapidly obtained according to the data at the previous moment, and the operation efficiency of the automatic device is improved.
It is noted that during data processing, individual invalid data may occur, which need to be identified and processed accordingly.
This requires us to calculate the fluctuation upper bound of the first M data in real time during normal operation and calculation:
f(a)＝max a (I a -I a-1 ) a＝1,2,3,……,M
when the sampling data at the M +1 th moment is far larger than the fluctuation upper limit, the data is judged to be invalid data. When the data fluctuation is within the acceptable range, the fluctuation of the first M data is continuously processed.
And if the maximum value of the sampling fluctuation at the M-th moment is f (1), f (1) needs to be removed at the M + 1-th moment, and the maximum value of the fluctuation is recalculated.
If the maximum value of the sampling fluctuation at the M-th time is not f (1), the M + 1-th time only needs to be (I) n -I n-1 ) Comparing with f (a).
When invalid data is encountered, it is processed by referring mainly to the values of the previous and next time instants and taking the average value thereof, that is,
therefore, we can simultaneously give the average value of the loads at two moments in time next to the invalid data as follows:
wherein
Refers to the average value of the load data at the moment of invalidity,
means average value of load data at the next moment of invalid data, I N-1 It is the original value, I, read at the moment before the invalid data N+1 It means that the invalid data is read at the later timeThe original value is taken.
And respectively storing the data by adopting a plurality of two-dimensional arrays. And the orgin [2] [ i ] is the original data which is pushed forward by 2N, wherein the orgin [0] [ i ] and the orgin [1] [ i ] are the original data at the first N moments and the second N moments respectively. average [2] [ i ] is average value data advanced by 2N, wherein average [0] [ i ] and average [1] [ i ] are average value data of the first N time points and the last N time points respectively. Wherein N is the number of samples in the set sample storage period.
history data is stored in history [ d ] [ t ]. The history [ d ] is t data in one period, and the history [ d ] [ t ] is data at the t-th time in a certain period.
After real-time sampling, the samples are firstly stored in an origin array, an average value is calculated through the method, and average value data are put into average.
When the remaining capacity of the data storage space is zero, overwriting storage from the beginning.
While calculating the average value, the device time is read. And when the current time is judged to meet the sample storage period, verifying whether the number of the average values which are not stored at the time accords with the sample period, namely whether the current average array is positioned at the average [0] [ N-1] and the average [1] [ N-1] or not. If both of them are in accordance with the condition, the data is stored in the historical database, so that the sampling analysis can be conveniently carried out at any time.
In addition, some space for temporarily storing data such as model coefficients, cluster analysis, prediction results and the like is also needed, the space does not need to exist for a long time, and the space is called when prediction analysis is carried out after a fault occurs and can be released after the calling is finished, so that detailed description is omitted.
When a fault occurs, accurate load switching can be realized according to load data at the previous moment and prediction of the load, wherein the prediction idea of the load refers to fig. 2.
1) Analysis of recent data
Now, let us assume that the load average value at the latest moment is I at the moment T of the fault occurrence T The load data in the latest history data is I (D, H), and the previous load data in the latest history data is I (D, H)The load data at one time is I (D, H-1), and the processing procedure is as follows.
When the time T is not close to the storage cycle time of the sample, an approximate model can be established through the data at the time and the recent historical data of the sample,
the model is built similarly as described above when time T is approaching the next sample storage cycle time.
Wherein C is t Are model coefficients.
In the historical database history [ d ] [ t ], the model is assumed to be
Wherein
Load data in recent historical samples.
Finding t-2 th data in different history [ d ] and respectively calculating respective coefficients according to the t-1 data (namely the column where the recent history sample data is located and the data in the next column)
Wherein T is the sample storage period, C d And d period of fault moment model coefficients.
When the time T just exceeds the storage cycle time of the sample, an approximate model can be built by the data at the time and the previous data of the latest historical data of the sample, as shown in fig. 3,
then the t-2 data are found in different history [ d ] and respective coefficients are calculated from the t-3 data
For visual presentation, the model coefficients are listed as follows:
C d ＝(C 0 ,C 1 …C D-1 ,C t )
according to the principle of cluster analysis, K points are randomly taken as initial clustering centers, and then C is subjected to d All points in the array are assigned K initial clusters, which are represented by f (x) (1, 2, … …, K), by calculating the nearest center to each point and assigning the centers to the corresponding clusters.
Finding out the point nearest to all the numerical values in the cluster among the K clusters as a new cluster center, namely obtaining the point by calculation
Min＝Min(f(x))
Max＝Max(f(x))
The new core is
Where x is 1,2, … …, k.
Repeat pair C d All points in the array are assigned K new clusters by calculating the nearest cluster center to each point and assigning it to the corresponding cluster, which is denoted by l (x) (1, 2, … …, K).
And continuously repeating the process, performing closest point analysis on all the numerical values in the K clusters, finding a new clustering center, and re-dividing the clusters of the whole array according to the new clustering center. Until the latest center coincides with the previous center, the cluster is the optimized cluster combination.
In the optimal clustering combination, a fault moment model coefficient C is found t The cluster where the cluster is located and the period (d value) corresponding to the cluster, thereby preliminarily screening out the approximationA matching historical sample.
2) Analysis of historical data
According to the previous screening, we find the number of sample cycles of the initial match, assumed to be i-1, and recombine and analyze the corresponding historical data.
Wherein d is 0 、d 1 ……d i-2 The number of the sample cycles primarily matched in the previous step is i-1 in total; the corresponding line contents are history data in the matching period respectively. Last row I D Is historical data of the cycle in which the fault is located. The total of t load data per cycle.
And respectively establishing models for a plurality of data segments before the fault, wherein the combination of the series of data also represents the general trend of the load data in the time.
Suppose that the fault occurs at time t-2, i.e. in the array
To (3). According to the above, the model coefficient at time t is obtained from the data at time t and time t + 1, in this case, the model coefficient at time t-2 needs to be calculated, and the fault time and n model coefficients before the fault time are calculated and listed, which are expressed as:
wherein
Denotes d 0 And (4) the model coefficients at n moments are pushed forward from the fault occurrence moment (t-2 moment) in the period.
For the analysis and processing of the two-dimensional array, the invention adopts a recursive method to screen layer by layer, and adopts clustering to analyze each layer until finding the period number which is most matched with the fault period. And the load trend of the period is referred to, and the load of the fault period is predicted.
First, the cycle with the highest degree of trend coincidence with the fault cycle load data needs to be found, namely, the cycle with the closest model coefficient to the fault cycle needs to be found. It is worth noting that the comprehensive trend of several time points before the fault time needs to be considered, namely, a balance scheme that multiple points are close to each other needs to be considered.
From the viewpoint of processing analysis method by an automation device, the requirements of practical application are arranged into mathematical requirements, namely, a line closest to each data in the last line of the two-dimensional array needs to be found, and the corresponding line number (d) is recorded i Value), a detailed description of the specific steps is presented below.
First, each column of the two-dimensional array is analyzed. Respectively randomly taking 2 points for each column as initial clustering centers, repeating the calculation by the clustering analysis method until the corresponding cluster of each column is found, and representing the 2(n +1) clusters as f t-2-n (x)、f t-1-n (x)、……、f t-2 (x) (wherein x is 1, 2).
Reading clusters where all columns of data of the last row are located, wherein the clusters are (n +1) in total and are respectively represented as f t-2-n (1)、f t-1-n (1)、……、f t-2 (1). Reading out the corresponding cycle number of the data in the clusters to obtain n +1 groups d i Number, i.e. d t-2-n (1)、d t-1-n (1)、……、d t-2 (1)。
By pair d t-2-n (1)、d t-1-n (1)、……、d t-2 (1) And (5) carrying out next analysis, and finding out the intersection of the n +1 groups of cycles so as to find out the matching cycles of the screening.
Counting the number of cycles in all clusters using a counting method, wherein d is counted as n +1 i The number is the maximum intersection number, and the cycle number of the group is the fault cycle matching item after the preliminary screening. The specific method is as follows,
i-1 cycles of preliminary matching in the previous stepEach cycle of the number is counted separately. That is, it is calculated at d from the 1 st to i-1 t-2-n (1)、d t-1-n (1)、……、d t-2 (1) The number of occurrences in these n +1 arrays.
If the count of the z-th cycle is n +1, the number of cycles is saved in time n+1 (i) In the array, if the cycle count is n, the number of cycles is saved in time n (i) In an array. And so on, if the repeated count value of the period number is stored in the array corresponding to the maximum value, d z The period is the matching item, d z The model coefficients corresponding to the period are rearranged into a new two-dimensional array C [ z ]][n+1]。
The number of rows in the two-dimensional array is unchanged and is still n +1 rows; the number of rows is reduced to z (z-1 is the number of matching cycles after the last round of screening).
And repeating the steps, continuously analyzing each column of the two-dimensional array, taking two clustering centers for clustering analysis, and carrying out cycle number screening on n +1 clusters until the analysis is finished when the screened cycle number is 1, namely, finishing the matching when z is 2.
And finding out the best matching period through the steps, reading the number of the periods, and finding out the original historical data of the corresponding row through the period to carry out the next analysis.
3) Prediction of future data
Automation device automatically finds the most matched historical data
This historical data is then further analyzed.
The model coefficients of the corresponding time of the current fault in the period and the later several times are calculated as follows.
……
Model coefficient according to fault occurrence time
Calculating to obtain the following model coefficients:
thereby predicting the maximum value of the load in a future period of time as
The automation device can determine the amount of the dedicated load and the specific scheme according to the maximum load value and the maximum available load value. Therefore, accurate switching during load transfer is realized, the running stability is improved, and resources are utilized to the maximum extent.
It should be understood by those skilled in the art that the above embodiments are only for illustrating the present invention and are not to be used as a limitation of the present invention, and that changes and modifications to the above described embodiments are within the scope of the claims of the present invention as long as they are within the spirit and scope of the present invention.
Claims (4)
1. A method for load calculation and prediction on an automated device, comprising the steps of:
step 1, acquiring and calculating load data in normal operation, acquiring and calculating the average value of N data in half an hour in real time in the normal operation, calculating the average value by a weight method, identifying invalid data, and performing substitution processing on the invalid data;
step 2, storing the data by adopting a plurality of two-dimensional arrays respectively, wherein orgin [2] [ i ] is 2N pieces of original data, average [2] [ i ] is 2N pieces of average value data, and N is the number of samples in a set sample storage period; history data are stored in history [ d ] [ t ]; the history [ d ] is t data in one period, and the history [ d ] [ t ] is data at the t-th moment in a certain period; when the residual capacity of the data storage space is zero, overwriting and storing from the beginning; calculating the average value and reading the device time; when the current time is judged to meet the sample storage period, verifying whether the number of the average values which are not stored at the moment meets the sample period; if both the data are in accordance with the conditions, the data are stored in a historical database, so that sampling and analysis can be conveniently carried out at any time;
(3) analyzing the time data of the fault time to find a period set similar to the trend of the fault time; an approximate model is built by the fault moment and the recent historical data in the sample,
in a history database
In the same time as the fault time in each period, namely the data of the t-2 th column in the array, and respectively calculating respective coefficients according to the data of the t-1 th column
For all model coefficients C d ＝(C 0 ，C 1 …C D-1 ，C t ) Carrying out data processing according to the principle of cluster analysis to find out an optimal cluster combination; these and fault time model coefficients C t The period corresponding to the similar model coefficient is the period similar to the load data trend at the fault moment;
according to the previous screening, the historical data corresponding to the matching cycle number is recombined and analyzed;
respectively establishing models for a plurality of data segments before the fault, listing out the fault time and n model coefficients before the fault time, and expressing as follows:
they can represent the comprehensive trend of each period load data in a period before the fault moment;
firstly, a cycle with the highest trend coincidence degree with the fault cycle load data needs to be found, namely, the cycle with the closest model coefficient with the fault cycle is found; a row closest to each data in the last row of the two-dimensional array needs to be found, and the corresponding row number (cycle number) is recorded; for the analysis and processing of the two-dimensional array, the invention adopts a recursive method to screen layer by layer, and adopts clustering to analyze each layer until finding the periodicity which is most matched with the fault period, and reasonably predicting the load according to the periodicity;
calculating the corresponding time of the current fault in the period and model coefficients of a plurality of backward moments as follows;
……
according to the model coefficient C _ D ^ (t-2) at the fault occurrence time, the subsequent model coefficients are calculated and respectively:
……
the maximum value of the load in a period of time in the future is predicted according to the load value at the fault moment, and the amount of the special supply load and the specific scheme are determined according to the maximum load value and the maximum available load value, so that the accurate switching of the load transfer is realized.
2. The method of claim 1, wherein the load calculation and prediction is performed on an automated device by: the calculation of the average value of the real-time load data is realized by introducing a weight method;
meanwhile, invalid data is identified and processed in the data processing process; when the sampled data is compared with the previous data, the fluctuation is much larger than the maximum fluctuation value of the previous M data, namely max a (I a -I a-1 ) (a ═ 1,2,3, … …, M), and this data is judged to be invalid data;
when invalid data is encountered, the processing method mainly refers to the values of the previous moment and the next moment and takes the average value to replace, namely
The average value of the load at two moments in time is given at the next moment of invalid data, as follows:
3. the method of claim 1, wherein the load calculation and prediction is performed on an automated device by: analyzing the load data of the fault time point, firstly calculating a fault time model coefficient and a model coefficient at the same time of each period, and secondly finding out the period number similar to the trend of the fault time by adopting a clustering analysis method;
randomly taking K points as initial focusing center and then matching model coefficient C d Calculating the nearest clustering centers of all the points in the array, and dividing the nearest clustering centers into corresponding clusters to obtain K initial clusters, wherein the K initial clusters are expressed as f (x) (x is 1,2, … …, K);
finding out the point nearest to all the values in the K clusters as a new center of convergence, namely obtaining the point by calculation
Min＝Min(f(x))
Max＝Max(f(x))
The new core is
Repeat pair model coefficient C d Calculating the nearest clustering centers of all the points in the array, and dividing the nearest clustering centers into corresponding clusters to obtain K new clusters, wherein the K new clusters are expressed as l (x) (x is 1,2, … …, K);
continuously repeating the process, performing closest point analysis on all numerical values in the K clusters, finding a new clustering center, and re-dividing the clusters of the whole array according to the new clustering center; until the latest center coincides with the previous center, the cluster is the optimized cluster combination, namely, the cycle number corresponding to the cluster represents the cycle similar to the trend of the fault moment.
4. The method of claim 1, wherein the load calculation and prediction is performed on an automated device by: comprehensively analyzing data in a period of time before the fault moment, namely analyzing the trend of the load data in the period of time before the fault moment, and finding a period matched with the trend in the period of time in the fault period;
respectively establishing models for a plurality of data segments before the fault, and finding out the period with similar coefficients, namely the period with the most matched trend in the period, through coefficient comparison; the method can be converted into analysis of a two-dimensional array, and needs to combine a recursive idea and a clustering analysis method;
firstly, respectively analyzing each column of the two-dimensional array, randomly taking 2 points for each column as initial clustering centers, repeatedly calculating by the clustering analysis method until finding out clusters corresponding to each column, and expressing the 2(n +1) clusters as f t-2-n (x)、f t-1-n (x)、......、f t-2 (x) (wherein x is 1, 2);
reading clusters where all columns of data of the last row are located, wherein the clusters are (n +1) in total and are respectively represented as f t-2-n (1)、f t-1-n (1)、......、f t-2 (1) (ii) a Reading out the corresponding cycle number of the data in the clusters to obtain n +1 groups of cycle numbers, namely d t-2-n (1)、d t-1-n (1)、......、d t-2 (1)；
By pair d t-2-n (1)、d t-1-n (1)、......、d t-2 (1) Carrying out next analysis, and finding out the intersection of the n +1 groups of cycles, thus finding out the matching cycles of the current screening;
counting the number of cycles in all the clusters by using a counting method, wherein the number of cycles with the largest count is the maximum intersection number, and the number of cycles is a fault cycle matching item after primary screening; the specific method comprises the following steps:
counting each period number in the i-1 period numbers preliminarily matched in the previous step respectively; that is, it is calculated at d from the 1 st to i-1 t-2-n (1)、d t-1-n (1)、......、d t - 2 (1) The number of occurrences in the n +1 arrays;
if the count of the z-th cycle is n +1, the number of cycles is saved in time n+1 (i) In the array, if the cycle count is n, the number of cycles is saved in time n (i) In an array; and so on, if the repeated count value of the period number is stored in the array corresponding to the maximum value, the matching item is the matching item, and d is used for matching the matching item z The model coefficients corresponding to the period are rearranged into a new two-dimensional array C [ z ]][n+1]；
The number of rows in the two-dimensional array is unchanged and is still n +1 rows; the number of lines is reduced to z (z-1 is a plurality of numerical values of the matching period after the previous round of screening);
repeating the steps, continuously analyzing each row of the two-dimensional array, taking two clustering centers for clustering analysis, and carrying out periodicity screening on n +1 clusters until the analysis is finished when the screened periodicity is 1, namely, completing the matching when z is 2;
and finding out the optimal matching period through the steps, reading the number of the periods, and finding out the original historical data of the corresponding row through the period, namely the historical data of the optimal matching.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010387688.9A CN111581883B (en) | 2020-05-09 | 2020-05-09 | Method for calculating and predicting load on automation device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010387688.9A CN111581883B (en) | 2020-05-09 | 2020-05-09 | Method for calculating and predicting load on automation device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111581883A CN111581883A (en) | 2020-08-25 |
| CN111581883B true CN111581883B (en) | 2022-09-23 |
Family
ID=72124701
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010387688.9A Active CN111581883B (en) | 2020-05-09 | 2020-05-09 | Method for calculating and predicting load on automation device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111581883B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112749920B (en) * | 2021-01-29 | 2023-05-23 | 南方电网科学研究院有限责任公司 | Non-fault area power supply recovery method, device, equipment and storage medium |
| CN115269593B (en) * | 2022-07-20 | 2023-04-18 | 清云智通(北京)科技有限公司 | Industrial data acquisition and storage method, device and system |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105844371A (en) * | 2016-05-19 | 2016-08-10 | 北京中电普华信息技术有限公司 | Electricity customer short-term load demand forecasting method and device |
| CN106485262B (en) * | 2016-09-09 | 2020-02-07 | 国网山西省电力公司晋城供电公司 | Bus load prediction method |
| CN107944594B (en) * | 2017-09-30 | 2021-11-23 | 华南理工大学 | Short-term load prediction method based on spearman grade and RKELM microgrid |
| CN109636059A (en) * | 2018-12-24 | 2019-04-16 | 国网北京市电力公司 | Electric heating distribution transformer load forecasting method and device |
| CN109888800B (en) * | 2019-03-07 | 2022-11-29 | 南京工程学院 | Power distribution area three-phase unbalanced load adjustment method based on load prediction and commutation strategy |
| CN110866633B (en) * | 2019-10-25 | 2023-11-24 | 上海电气集团股份有限公司 | Micro-grid ultra-short-term load prediction method based on SVR support vector regression |
-
2020
- 2020-05-09 CN CN202010387688.9A patent/CN111581883B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN111581883A (en) | 2020-08-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111581883B (en) | Method for calculating and predicting load on automation device | |
| CN110232203B (en) | Knowledge distillation optimization RNN short-term power failure prediction method, storage medium and equipment | |
| CN108022014B (en) | Power system load prediction method and system | |
| CN106909933A (en) | A kind of stealing classification Forecasting Methodology of three stages various visual angles Fusion Features | |
| CN115647687B (en) | Welding abnormity real-time detection method integrating residual error network and self-attention mechanism | |
| CN114757432A (en) | Future execution activity and time prediction method and system based on flow log and multi-task learning | |
| CN112447265A (en) | Lysine acetylation site prediction method based on modular dense convolutional network | |
| CN111160626A (en) | Power load time sequence control method based on decomposition and fusion | |
| CN114239718A (en) | High-precision long-term time sequence prediction method based on multivariate time sequence data analysis | |
| CN114884828A (en) | Space-based networking efficiency evaluation system and method based on analytic hierarchy process | |
| CN112861071A (en) | High-speed rail traction system anomaly detection method based on deep self-coding | |
| CN117278313A (en) | Security detection method and system for computer network structure optimization | |
| CN117668691A (en) | Power system transient stability judging method and system based on deep learning | |
| CN111539558A (en) | Power load prediction method adopting optimized extreme learning machine | |
| CN111311318A (en) | User loss early warning method, device, equipment and storage medium | |
| CN109242304B (en) | Method for evaluating reliability of small-probability event of power system | |
| CN113158134B (en) | Method, device and storage medium for constructing non-invasive load identification model | |
| CN114239999A (en) | Element reliability parameter optimization analysis method based on cross entropy important sampling | |
| CN113298298B (en) | Short-term load prediction method and system for charging pile | |
| CN114399407A (en) | Power dispatching monitoring data anomaly detection method based on dynamic and static selection integration | |
| CN114840505A (en) | Main transformer big data preprocessing method based on big data analysis platform | |
| CN114154567A (en) | Wind power plant station operation data anomaly identification method based on machine learning | |
| CN113989070A (en) | Short-term energy consumption prediction method and system for power utilization end of smart power grid | |
| CN113821419A (en) | Cloud server aging prediction method based on SVR and Gaussian function | |
| CN113592064A (en) | Ring polishing machine process parameter prediction method, system, application, terminal and medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |