Kalman filter based capacity forecasting method, system and computer equipment

Patent metadata

Field	Value
Publication number	US-2018196891-A1
Application number	US-201515033604-A
Country	US
Kind code	A1
Filing date	Sep 7, 2015
Priority date	Apr 3, 2015
Publication date	Jul 12, 2018
Grant date	—

Abstract

In one embodiment, a Kalman filter based capacity forecasting method includes acquiring a capacity time sequence of an object to be forecasted; establishing a dynamical model for the capacity time sequence, and extracting a state transition parameter and a process noise parameter of the dynamical model; performing Kalman filter estimation on the capacity time sequence by using the state transition parameter and the process noise parameter to generate at least one state characteristic signal; segmenting the capacity time sequence according to the at least one state characteristic signal, and determining at least one corresponding segmentation point; and forecasting the capacity at future time according to the at least one segmentation point determined in the capacity time sequence. By adopting the technical solutions of the invention, accurate forecasting of the capacity growth is achieved, to facilitate operation and maintenance personnel making a reasonable capacity expansion plan.

First claim

1 . A Kalman filter based capacity forecasting method, comprising:
acquiring a capacity time sequence of an object to be forecasted; establishing a dynamical model for the capacity time sequence, and extracting a state transition parameter and a process noise parameter of the dynamical model; performing a Kalman filter estimation on the capacity time sequence by using the state transition parameter and the process noise parameter to generate at least one state characteristic signal; segmenting the capacity time sequence according to the at least one state characteristic signal, and determining at least one corresponding segmentation point; and forecasting the capacity at future time according to the at least one corresponding segmentation point determined in the capacity time sequence.

2 . The method of claim 1 , further comprising:
filtering the capacity time sequence by adopting a median filtering and/or a moving average filtering to generate a filtered capacity time sequence.

3 . The method of claim 1 , wherein the dynamical model comprises at least one of a constant velocity (CV) model, a constant acceleration (CA) model, a Singer model, a current statistical model and a Jerk model.

4 . The method of claim 3 , wherein establishing the dynamical model for the capacity time sequence, and extracting the state transition parameter and the process noise parameter of the dynamical model comprises:
constructing a system state equation under the Jerk model for the capacity time sequence, and extracting a state transition matrix A and a process noise variance matrix Q corresponding to the system state equation.

5 . The method of claim 4 , wherein performing the Kalman filter estimation on the capacity time sequence by using the state transition parameter and the process noise parameter to generate the at least one state characteristic signal comprises:
performing the Kalman filter estimation on the capacity time sequence by using the state transition matrix A and the process noise variance matrix Q, and generating at least one state characteristic signal of a velocity, an acceleration and an acceleration derivative.

6 . The method of claim 5 , wherein segmenting the capacity time sequence according to the at least one state characteristic signal, and determining at least one corresponding segmentation point comprises:
segmenting the capacity time sequence according to the at least one state characteristic signal of the velocity, the acceleration and the acceleration derivative obtained after the Kalman filter estimation on the capacity time sequence, and determining the at least one corresponding segmentation point.

7 . The method of claim 6 , wherein segmenting the capacity time sequence according to the at least one state characteristic signal of the velocity, the acceleration and the acceleration derivative obtained after the Kalman filter estimation on the capacity time sequence, and determining the at least one corresponding segmentation point comprises:
determining at least one phase change signal area for at least one of the signals of the velocity, the acceleration and the acceleration derivative obtained after the Kalman filter estimation on the capacity time sequence; and segmenting the capacity time sequence according to the at least one determined phase change signal area, and determining the at least one corresponding segmentation point.

8 . The method of claim 7 , wherein determining the at least one phase change signal area for the at least one of the signals of the velocity, the acceleration and the acceleration derivative obtained after the Kalman filter estimation on the capacity time sequence comprises:
calculating, for the at least one of the signals of the velocity, the acceleration and the acceleration derivative obtained after the Kalman filter estimation on the capacity time sequence, variances or means of the signals of the velocity within every two sliding windows according to the capacity time sequence by adopting two continuous sliding windows with an identical length; marking an intermediate point between the two continuous sliding windows as a suspected segmentation point, if a probability that the variances or means of the signals of the velocity within the two sliding windows are not identical is greater than a preset corresponding probability threshold; and determining a signal area constituted by a plurality of continuous suspected segmentation points as a phase change signal area of the capacity time sequence.

9 . The method of claim 8 , wherein segmenting the capacity time sequence according to the at least one determined phase change signal area, and determining the at least one corresponding segmentation point comprises:
determining, in each phase change signal area, the suspected segmentation point corresponding to a maximum of differences of the variances or the maximum of differences of the means of the signals of velocity within the two sliding windows as the segmentation point.

10 . The method of claim 8 , wherein segmenting the capacity time sequence according to the at least one determined phase change signal area, and determining the at least one corresponding segmentation point comprises:
determining the intermediate point in each phase change signal area as the segmentation point.

11 . The method of claim 9 , wherein forecasting the capacity at future time according to the at least one corresponding segmentation point determined in the capacity time sequence comprises:
taking a last segmentation point in the capacity time sequence as a starting point, linearly or nonlinearly fitting the data of the capacity time sequence thereafter to generate a capacity forecasting model; and forecasting the capacity at future time according to the capacity forecasting model.

12 . The method of claim 11 , further comprising:
performing a goodness-of-fit assessment on the generated capacity forecasting model; and triggering a processing operation of forecasting the capacity at future time according to the capacity forecasting model, if an assessment value obtained by the goodness-of-fit assessment is greater than a preset goodness-of-fit threshold.

13 . A Kalman filter based capacity forecasting system, comprising:
an acquisition module, configured to acquire a capacity time sequence of an object to be forecasted; an establishment and extraction module, configured to establish a dynamical model for the capacity time sequence, and extract a state transition parameter and a process noise parameter of the dynamical model; an estimation generating module, configured to perform a Kalman filter estimation on the capacity time sequence by using the state transition parameter and the process noise parameter to generate at least one state characteristic signal; a segmenting module, configured to segment the capacity time sequence according to the at least one state characteristic signal, and determining at least one corresponding segmentation point; and a forecasting module, configured to forecast the capacity at future time according to the at least one segmentation point determined in the capacity time sequence.

14 . The system of claim 13 , further comprising:
a preprocessing module, configured to filtering the capacity time sequence by adopting a median filtering and/or a moving average filtering to generate a filtered capacity time sequence.

15 . The system of claim 13 , wherein the dynamical model comprises at least one of a constant velocity (CV) model, a constant acceleration (CA) model, a Singer model, a current statistical model and a Jerk mod

Assignees

• Baidu online network technology beijing co ltd

Inventors

• Miao Beibei
• Chen Yu
• Jin Xuebo
• Qu Xianping
• Tao Shimin
• Zang Zhi
• Wang Bo

Classifications

• G06F16/23 Primary

  Updating · CPC title

• G06F2111/10

  Numerical modelling · CPC title

• G16Z99/00

  Subject matter not provided for in other main groups of this subclass · CPC title

• G06F30/20 Primary

  Design optimisation, verification or simulation (optimisation, verification or simulation of circuit designs G06F30/30) · CPC title

• H03H17/0257

  KALMAN filters · CPC title