CN-122022518-A - Novel load prediction influence factor association analysis algorithm

CN122022518ACN 122022518 ACN122022518 ACN 122022518ACN-122022518-A

Abstract

The invention discloses a novel load prediction influence factor association analysis algorithm which comprises a frequency spectrum identification step, a parameter inversion step and a main mode locking step, wherein the frequency spectrum identification step is used for acquiring dynamic time sequence data of influence factors and forming a factor frequency spectrum, and identifying a first peak frequency and a second peak frequency at the high frequency side of the factor frequency spectrum to determine a frequency band bandwidth, the parameter inversion step is used for obtaining factor equivalent association rigidity and slow-change basic association strength based on constraint conditions of core factors and association factors. Relates to the technical field of load prediction. According to the method, dynamic time sequence data of influencing factors are obtained through a frequency spectrum identification step, factor frequency spectrums are generated, a first peak frequency and a second peak frequency at a high frequency side are identified to determine a frequency band width, a precise frequency boundary is provided for subsequent association characteristic analysis, and then the action mechanism of the influence of the factor association characteristic on the load is defined by reversely solving the equivalent association stiffness of the factors and the association strength of the slow-change foundation based on constraint conditions of core factors and association factors through a parameter inversion step.

Inventors

Liang Maolong
LIN MEIYING

Assignees

广东南泰能源科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260129

Claims (8)

1. A novel load prediction influence factor association analysis algorithm, comprising: A frequency spectrum identification step, which is used for acquiring dynamic time sequence data of influence factors and forming a factor frequency spectrum, and identifying a first peak frequency and a second peak frequency at the high frequency side of the factor frequency spectrum so as to determine the frequency band width; A parameter inversion step, which is used for obtaining factor equivalent association stiffness and slow-change basic association strength based on constraint conditions of core factors and association factors; A main mode locking step, which is used for generating a correlation mode set according to the factor equivalent correlation stiffness and the slow-change basic correlation strength, determining a main correlation mode based on the factor correlation slope energy concentration, and determining the center frequency of a locking frequency band by the main correlation mode and the slow-change basic correlation strength; The association increment calculating step is used for carrying out band-pass extraction on the influence factor dynamic time sequence data according to the locking frequency band center frequency and the frequency band bandwidth to obtain a locking band factor dynamic component, and converting the locking band factor dynamic component into a load influence increment according to the relationship between the association increment and the load influence increment in the association relation geometry and data association analysis; A mode comparison step, which is used for setting an analysis time window according to the center frequency of the locking frequency band and the bandwidth of the locking frequency band, and comparing the main association mode in the current analysis time window with the main association mode in the previous analysis time window to obtain a main association mode comparison result; and a decision generation step, which is used for issuing a dyne check instruction according to the main association mode comparison result and determining factor retesting priority based on the load influence increment.
2. The novel load prediction influence factor correlation analysis algorithm of claim 1, wherein the acquiring of the influence factor dynamic time series data and the forming of the factor spectrum, the identifying of the first peak frequency and the second peak frequency on the high frequency side of the factor spectrum to determine the frequency band bandwidth, comprises: Acquiring dynamic monitoring signals of influence factors, and performing constraint double-integration processing on the dynamic monitoring signals to obtain dynamic time sequence data of the influence factors; Performing short-time Fourier transform on the influence factor dynamic time sequence data to generate a factor spectrum; Identifying a first peak frequency and a second peak frequency at the high frequency side of the factor spectrum, and taking the nearest high-frequency side local minimum frequency and low-frequency side local minimum frequency of each peak frequency as a frequency band boundary respectively; the difference between the high-frequency side local minimum frequency and the low-frequency side local minimum frequency is used as the band bandwidth.
3. The novel load prediction influence factor association analysis algorithm according to claim 1, wherein the obtaining of the factor equivalent association stiffness and the slowly varying base association strength based on the constraint condition of the core factor and the association factor comprises: according to the geometric association of the factor association change and the dynamic data and the physical equivalent relation affecting the strength and the association degree, the factor association rigidity is equivalent to the association rigidity of the core association factor; Based on a correlation system model of the core factors and the correlation factors, establishing boundary conditions of the core factors at a direct correlation end and an indirect correlation end, and forming a mapping from an intrinsic relation and frequency to a correlation wave number; substituting the first peak frequency and the second peak frequency into the intrinsic relation respectively to form a combined constraint on the factor equivalent association rigidity and the slow-change basic association strength; And carrying out residue minimization solving under the combined constraint to obtain the factor equivalent associated stiffness and the slow-change basic associated strength.
4. A novel load prediction influence factor correlation analysis algorithm as claimed in claim 3 wherein generating a set of correlation modes from factor equivalent correlation stiffness and slow base correlation strength, and determining a primary correlation mode based on factor correlation slope energy concentration, and determining a lock band center frequency from the primary correlation mode and the slow base correlation strength comprises: according to the mapping of the frequency and the associated wave number, obtaining an associated wave number set meeting the intrinsic relation and sorting the associated wave number set according to descending order to obtain an associated mode set; Calculating factor association slope energy concentration degree one by one for the association mode set, and selecting the mode with the largest concentration degree as a main association mode; And calculating the center frequency of the locking frequency band based on the main association mode, the slow-change basic association strength and the factor association density.
5. The novel load prediction influence factor association analysis algorithm according to claim 1, wherein the band-pass extraction of the influence factor dynamic time series data according to the locking frequency band center frequency and the frequency band bandwidth to obtain a locking band factor dynamic component, and the conversion of the locking band factor dynamic component into the load influence increment according to the relation between the association increment and the load influence increment in the association relation geometry and the data association analysis comprises: Calculating a left band edge and a right band edge with the locking frequency band center frequency and the frequency band bandwidth, wherein the left band edge is equal to the locking frequency band center frequency minus half of the frequency band bandwidth, and the right band edge is equal to the locking frequency band center frequency plus half of the frequency band bandwidth; band-pass filtering is carried out on the dynamic time sequence data of the influence factors between the left band edge and the right band edge, so that dynamic components of the locking band factors are obtained; under the weak interference condition, extrapolating the locking band factor dynamic component into the dynamic distribution of the full factor association by using the space association function of the main association mode to obtain an association increment which is half of the integral of the dynamic distribution in the effective association range along the square of the first derivative of the association dimension; And converting the association increment into a load influence increment according to the factor association elasticity coefficient, the association influence area and the effective association range to obtain a load influence increment time sequence.
6. The novel load prediction influence factor association analysis algorithm according to claim 1, wherein the analysis time window is set according to the locked frequency band center frequency and the locked frequency band bandwidth, the main association mode in the current analysis time window is compared with the main association mode in the previous analysis time window, and a main association mode comparison result is obtained, and the novel load prediction influence factor association analysis algorithm comprises the following steps: taking the inverse of the bandwidth of the locking frequency band as the analysis time window length, and taking half of the inverse of the center frequency of the locking frequency band as the step pitch of the analysis time window to generate an analysis time window sequence which is arranged in time sequence; Selecting a current analysis time window and a last analysis time window from the analysis time window sequence, respectively reading main association modes in the two windows, and calculating a main association mode difference value; And when the difference value of the main association modes is not zero, determining that the main association mode comparison result is changed, and when the difference value of the main association mode is zero, determining that the main association mode comparison result is unchanged.
7. The novel load prediction influence factor association analysis algorithm according to claim 6, wherein the method for issuing a dyne check instruction according to the main association mode comparison result and determining the factor retest priority based on the load influence increment comprises the steps of: When the main association mode comparison result is changed, a factor checking instruction is issued, wherein the checking instruction comprises suspending the training of the current load prediction model or the prediction execution procedure, checking the association relation between the influence factor data and the verification, and when the main association mode comparison result is unchanged, the checking instruction is continuous monitoring and data recording; When the main association mode comparison result is changed, the root mean square value of the load influence increment time sequence is used as the influence intensity measurement, and the factor retest priority is determined based on the ratio of the influence intensity measurement to the slowly-varying basic association intensity.
8. The novel load prediction influence factor correlation analysis algorithm according to claim 1, wherein the analysis evidence package is formed based on factor equivalent correlation stiffness, slow-change base correlation strength, main correlation mode, locking frequency band center frequency, load influence increment and main correlation mode comparison results.

Description

Novel load prediction influence factor association analysis algorithm Technical Field The invention relates to the technical field of load prediction, in particular to a novel load prediction influence factor correlation analysis algorithm. Background The load prediction is widely applied to a plurality of fields such as power system dispatching, energy resource optimal allocation, industrial production planning and the like, and the prediction accuracy directly influences the operation efficiency, cost control and safety stability of related systems. In the load prediction process, factors influencing load change are complex and various, including historical load data fluctuation, environmental parameter change, user behavior mode adjustment and the like, a complex association relationship exists among the factors, and part of the factors can generate dynamic fluctuation so as to indirectly influence the accuracy of a load prediction result. In the prior art, a single-dimension factor analysis method is mostly adopted, the dynamic characteristics and the mutual correlation of each influence factor are not fully considered, the analysis result deviation is easily caused by the interference factors, the basic correlation strength and the instantaneous dynamic correlation increment of the core influence factors cannot be accurately distinguished, the influence of the dynamic fluctuation factors on the correlation relationship is ignored, the input parameters of the load prediction model are unreasonable to select, the prediction precision is insufficient, and the requirements of actual application scenes are difficult to meet. Disclosure of Invention The invention aims to solve the defect that the association relation between a core influence factor and a load is difficult to accurately analyze due to dynamic fluctuation of the influence factor and association characteristic change among factors in the load prediction process in the prior art, and provides a novel load prediction influence factor association analysis algorithm. In order to solve the problems existing in the prior art, the invention adopts the following technical scheme: A novel load prediction influence factor association analysis algorithm, comprising: A frequency spectrum identification step, which is used for acquiring dynamic time sequence data of influence factors and forming a factor frequency spectrum, and identifying a first peak frequency and a second peak frequency at the high frequency side of the factor frequency spectrum so as to determine the frequency band width; A parameter inversion step, which is used for obtaining factor equivalent association stiffness and slow-change basic association strength based on constraint conditions of core factors and association factors; A main mode locking step, which is used for generating a correlation mode set according to the factor equivalent correlation stiffness and the slow-change basic correlation strength, determining a main correlation mode based on the factor correlation slope energy concentration, and determining the center frequency of a locking frequency band by the main correlation mode and the slow-change basic correlation strength; The association increment calculating step is used for carrying out band-pass extraction on the influence factor dynamic time sequence data according to the locking frequency band center frequency and the frequency band bandwidth to obtain a locking band factor dynamic component, and converting the locking band factor dynamic component into a load influence increment according to the relationship between the association increment and the load influence increment in the association relation geometry and data association analysis; A mode comparison step, which is used for setting an analysis time window according to the center frequency of the locking frequency band and the bandwidth of the locking frequency band, and comparing the main association mode in the current analysis time window with the main association mode in the previous analysis time window to obtain a main association mode comparison result; and a decision generation step, which is used for issuing a dyne check instruction according to the main association mode comparison result and determining factor retesting priority based on the load influence increment. Preferably, acquiring the influence factor dynamic time series data and forming a factor spectrum, identifying a first peak frequency and a second peak frequency at a high frequency side of the factor spectrum to determine a frequency band bandwidth, including: Acquiring dynamic monitoring signals of influence factors, and performing constraint double-integration processing on the dynamic monitoring signals to obtain dynamic time sequence data of the influence factors; Performing short-time Fourier transform on the influence factor dynamic time sequence data to generate a factor spectrum; Identifying a first peak frequency and a se