CN-121997165-A - Intermittent energy output modeling prediction method and system based on Gaussian process

CN121997165ACN 121997165 ACN121997165 ACN 121997165ACN-121997165-A

Abstract

Modeling intermittent energy output based on a Bernstein polynomial, calculating the mean value and covariance function of the intermittent energy output, and constructing a Gaussian process priori form; the method comprises the steps of obtaining a historical observation set of intermittent energy output, optimizing super parameters in a Gaussian process prior form through maximum likelihood estimation according to noise observation points of each track sample in the historical observation set in discrete time, solving a covariance matrix of a Bernstein coefficient, and calculating a mean value and the covariance matrix of posterior distribution of the Bernstein coefficient according to the observed intermittent energy output at a plurality of moments, so that the joint distribution of the intermittent energy output in future test time is obtained. Compared with the prior art, the method and the device can effectively utilize the latest observed data to predict the power generation amount of the next time period, and realize the energy scheduling and optimization of quick response.

Inventors

WEI XINCHI
LI MINGXUAN
LIU SHU
XU YIN
WEI WEI
WANG YING
SHI SHANSHAN
WU XIANGYU
ZHANG QIQI
CUI YONG

Assignees

国网上海市电力公司

Dates

Publication Date: 20260508
Application Date: 20251217

Claims (10)

1. The intermittent energy output modeling prediction method based on the Gaussian process is characterized by comprising the following steps of: Modeling the intermittent energy output based on a Bernstein polynomial, and constructing a Gaussian process prior form of the intermittent energy output based on a mean value and a covariance function of the intermittent energy output obtained by modeling; Acquiring a historical observation set of intermittent energy output, optimizing super parameters in the prior form of the Gaussian process through maximum likelihood estimation according to noise observation points of each track sample in the historical observation set in discrete time, and solving a covariance matrix of a Bernstein coefficient; According to the observed intermittent energy output at a plurality of moments, the mean value and the covariance matrix of the posterior distribution of the Bernstein coefficient are calculated by solving the obtained covariance matrix of the hyper-parameters and the Bernstein coefficient, so that the joint distribution of the intermittent energy output in the future test time is obtained.
2. The method for modeling and predicting intermittent energy output based on a gaussian process according to claim 1, wherein the expression of the intermittent energy output obtained by modeling the intermittent energy output based on a bernstein polynomial is: in the formula, As a result of modeling the intermittent energy source output, Is one A vitamin random variable representing a bernstein coefficient; for a reduced-order bernstein function space, In the case of a linear transformation, Is that The space of the order bernstein function, , To be a time interval Divided into The number of intervals between which the first and second electrodes are spaced, For the index parameter(s), Is that An order bernstein base function.
3. The method for modeling and predicting intermittent energy output based on a gaussian process according to claim 2, wherein the prior form of the gaussian process of intermittent energy output is: in the formula, Is that Is used for the average value of (a), Is that Is used as a function of the covariance of (a), And For different points in time; the said The specific computational expression of the covariance function of (2) is: in the formula, Is that Is used to determine the covariance function of the model, And The variance of the periodic and smooth components respectively, And As a feature length scale, a feature is provided, Representing a period.
4. The method for modeling and predicting intermittent energy output based on a gaussian process according to claim 3, wherein the expression of the noise observation point of each track sample in the historical observation set in discrete time is: in the formula, All track samples for power-out process The vector of dimensions is used to determine, Is a historical observation set of intermittent energy source output, For the number of noise observation points in the trace sample, In the last track sample in the set of historical observations of intermittent energy source output The output value of each noise observation point, In the last track sample in the history observation set for intermittent energy source output model The value of the individual noise observation points, Is the output vector of the intermittent energy output model, As a vector of the noise it is, In the last track sample in the set of historical observations of intermittent energy source output The noise vector of each noise observation point, Each element in the system is an independent and equidistributed zero mean value and variance is Is a gaussian noise of (c).
5. The method for modeling and predicting intermittent energy output based on a gaussian process according to claim 4, wherein the optimizing the super-parameters in the prior form of the gaussian process by maximum likelihood estimation is specifically: Order the Represents a mean vector, The covariance function of the power output process calculated on the noise observation point is represented by the following specific expression: in the formula, As a sum covariance function The associated super-parameter vector is used to determine, In the last track sample in the set of historical observations of intermittent energy source output Model prediction mean values of the noise observation points; Assume that 1, Wherein Is the mean value, 1 is And (3) a dimension unit vector, wherein the expression of an optimization target for optimizing the super parameters in the prior form of the Gaussian process through maximum likelihood estimation is as follows: Wherein the optimization parameters are as follows 、 And 。
6. The method for modeling and predicting intermittent energy output based on a gaussian process according to claim 5, wherein the solving process of the covariance matrix of the bernstein coefficients comprises: in a time range Selecting an evaluation time set At the evaluation time set Calculating covariance functions at various time points in (a) Constructed to The component is a linear equation of unknown quantity: in the formula, Is that Is used to determine the optimal estimate of (a), , ; By selecting the time point of evaluation to enable the following The component is the linear equation full rank of the unknown quantity, thereby solving the covariance matrix of the Bernstein coefficient 。
7. The method for modeling and predicting intermittent energy output based on a gaussian process according to claim 6, wherein the solving process of the joint distribution of intermittent energy output in the future test time is specifically as follows: on the day of prediction of intermittent energy source output, according to what has been observed The intermittent energy output at each moment predicts the intermittent energy output distribution at the future moment, and specifically comprises the following steps: record the observed The power output at each moment is Calculating the mean value of the posterior distribution of the Bernstein coefficients Sum covariance matrix Thereby calculating and obtaining future test time according to the modeling model of intermittent energy output Is a joint distribution of intermittent energy source output 。
8. The method for modeling and predicting intermittent energy output based on Gaussian process according to claim 7, wherein the mean value of posterior distribution of the Bernstein coefficients Sum covariance matrix The calculated expression of (2) is: in the formula, 。
9. The gaussian process-based intermittent energy output modeling prediction method according to claim 1, wherein the intermittent energy source is wind power energy source and/or photovoltaic energy source.
10. An intermittent energy source output modeling prediction system for implementing the intermittent energy source output modeling prediction method based on a gaussian process according to any of claims 1 to 9, comprising: the intermittent energy output modeling module is used for modeling the intermittent energy output based on the Bernstein polynomial, and constructing a Gaussian process prior form of the intermittent energy output based on a mean value and a covariance function of the intermittent energy output obtained by modeling; the parameter estimation module is used for acquiring a historical observation set of intermittent energy output, optimizing the super parameters in the prior form of the Gaussian process through maximum likelihood estimation according to the noise observation points of each track sample in the historical observation set on discrete time, and solving a covariance matrix of the Bernstein coefficient; The actual measurement module is used for calculating the mean value and the covariance matrix of posterior distribution of the Bernstein coefficient according to the observed intermittent energy output at a plurality of moments through solving the obtained covariance matrix of the super-parameters and the Bernstein coefficient, so as to obtain the joint distribution of the intermittent energy output in the future test time.

Description

Intermittent energy output modeling prediction method and system based on Gaussian process Technical Field The invention relates to the field of intermittent energy output prediction, in particular to an intermittent energy output modeling prediction method and system based on a Gaussian process. Background In the emergency recovery process of a power grid, randomness of factors such as intermittent power supply output, equipment repair time and mobile emergency resource allocation time has important influence on the establishment and implementation of a recovery plan of a system, but a traditional static modeling method is difficult to cope with, the invention disclosed by the publication No. CN110795841A discloses a mathematical modeling method for uncertainty of intermittent power supply output, and the method comprises the steps of 1, initializing a maximum expectation algorithm, adopting a hard clustering algorithm to solve model parameter initial values, 2, calculating expectation, carrying out probability calculation on hidden variables of each data, 3, maximizing, deriving a Gaussian mixture model parameter iteration formula with weight data, and 4, repeating the steps 2-3 until convergence. According to the scheme, modeling of the intermittent energy output uncertainty is performed through a Gaussian mixture model based on weight data and a maximum expected algorithm, but model parameters are calculated based on probability of prior distribution parameters, data volatility is high, repeated iteration is needed to be performed on the whole scheme in a parameter updating process, and data calculation amount is large. Disclosure of Invention The invention aims to overcome the defects of the prior art and provide the intermittent energy output modeling prediction method and the system based on the Gaussian process, which effectively utilize the latest observed data to predict the generated energy in the next time period and realize the energy scheduling and the optimization of quick response. The aim of the invention can be achieved by the following technical scheme: A modeling and predicting method for intermittent energy output based on a Gaussian process comprises the following steps: Modeling the intermittent energy output based on a Bernstein polynomial, and constructing a Gaussian process prior form of the intermittent energy output based on a mean value and a covariance function of the intermittent energy output obtained by modeling; Acquiring a historical observation set of intermittent energy output, optimizing super parameters in the prior form of the Gaussian process through maximum likelihood estimation according to noise observation points of each track sample in the historical observation set in discrete time, and solving a covariance matrix of a Bernstein coefficient; According to the observed intermittent energy output at a plurality of moments, the mean value and the covariance matrix of the posterior distribution of the Bernstein coefficient are calculated by solving the obtained covariance matrix of the hyper-parameters and the Bernstein coefficient, so that the joint distribution of the intermittent energy output in the future test time is obtained. Further, the expression of the intermittent energy output obtained by modeling the intermittent energy output based on the Bernstein polynomial is as follows: in the formula, As a result of modeling the intermittent energy source output,Is oneA vitamin random variable representing a bernstein coefficient; for a reduced-order bernstein function space, In the case of a linear transformation,Is thatThe space of the order bernstein function,,To be a time intervalDivided intoThe number of intervals between which the first and second electrodes are spaced,For the index parameter(s),Is thatAn order bernstein base function. Further, the prior form of the gaussian process of intermittent energy source output is as follows: in the formula, Is thatIs used for the average value of (a),Is thatIs used as a function of the covariance of (a),AndFor different points in time; the said The specific computational expression of the covariance function of (2) is: in the formula, Is thatIs used to determine the covariance function of the model,AndThe variance of the periodic and smooth components respectively,AndAs a feature length scale, a feature is provided,Representing a period. Further, the expression of the noise observation point of each track sample in the historical observation set at discrete time is: in the formula, All track samples for power-out processThe vector of dimensions is used to determine,Is a historical observation set of intermittent energy source output,For the number of noise observation points in the trace sample,In the last track sample in the set of historical observations of intermittent energy source outputThe output value of each noise observation point,In the last track sample in the history observation set for intermittent ene