Computational framework for modeling of physical process

1. A method performed by one or more computing system generating a set of system variables to represent a physical power system having a stochastic behavior, comprising:

receiving a set of measurements relating to one or more power grids in the physical power system, the set of measurements following a non-standard probability distribution and being non-linearly correlated, wherein a non-linear relationship exists between the set of system variables that describes the stochastic behavior and a corresponding set of measurements;

determining, based on the set of measurements, a numerical model of the stochastic behavior of the physical power system, wherein the numerical model comprises a feature space comprising non-correlated features corresponding to the stochastic behavior, wherein the non-correlated features have a dimensionality that is less than the dimensionality of the set of measurements;

sampling the numerical model based on a statistical inference approach to generate a set of approximated system variables corresponding to the set of measurements based on the numerical model, wherein the statistical inference approach comprises combining a Langevin Markov Chain Monte Carlo (MCMC) approach and an adaptive MCMC that calibrates a covariance of a distribution function based on a history of a Markov Chain; and

applying the set of approximated system variables to the physical power system for controlling actions.

2. The method of claim 1 , further comprising:

determining a second numerical model of the stochastic behavior, the second numerical model comprising a space having a same dimensionality of N, N being an integer,

wherein the stochastic behavior is associated with one or more target domains, and

wherein generating the set of approximated system variables comprises:

sampling the numerical model using a first statistical inference approach; and

reevaluating a subset of samples in the one or more target domains based on the second numerical model using a second statistical inference approach.

3. The method of claim 2 , wherein the one or more target domains corresponds to one or more failure events with corresponding probabilities of occurrences.

4. The method of claim 1 , further comprising:

determining a set of weighted measurements based on evaluating weights for the set of measurements, wherein the numerical model of the stochastic behavior is determined based on the set of weighted measurements.

5. The method of claim 4 , wherein the weights are evaluated iteratively for the set of measurements.

6. The method of claim 4 , wherein the weights are evaluated in an importance sampling procedure performed based on the statistical inference approach.

7. The method of claim 1 , wherein generating the set of approximated system variables comprises:

expanding the non-correlated features having the dimensionality of M to the set of approximated system variables having the dimensionality of N, N and M being integers.

8. The method of claim 1 , wherein the feature space is based on a statistical transformation that converts a first set of correlated values having a first dimensionality into a second set of uncorrelated values having a second dimensionality that is lower than the first dimensionality.

9. The method of claim 8 , wherein the statistical transformation comprises a manifold-learning based Isomap method.

10. The method of claim 1 , wherein the numerical model of the stochastic behavior of the physical power system comprises a Gaussian process emulator.

11. A system for generating model of a power system exhibiting a stochastic behavior, comprising:

a plurality of sensors configured to collect a set of measurements of the power system, the set of measurements following a non-standard probability distribution and being non-linearly correlated, and wherein a non-linear relationship exists between a set of system variables that describes the stochastic behavior and a corresponding set of measurements;

a processor coupled to the plurality of sensors; and

a memory including processor executable code, execution of which by the processor causes the system to:

determine, based on the set of measurements, a numerical model of the stochastic behavior, wherein the numerical model comprises a feature space comprising non-correlated features corresponding to the stochastic behavior, wherein the non-correlated features have a dimensionality of M and the set of measurements has a dimensionality of N, wherein M and N are integers, M being smaller than N;

sample the numerical model based on a statistical inference approach to generate a set of approximated system variables according to the set of measurements based on the numerical model, wherein the statistical inference approach comprises combining a Langevin Markov Chain Monte Carlo (MCMC) approach and an adaptive MCMC that calibrates a covariance of a distribution function based on a history of a Markov Chain; and

apply the set of approximated system variables to the power system for controlling actions.

12. The system of claim 11 , wherein the stochastic behavior is associated with one or more target domains, wherein the processor is configured to:

determine a second numerical model of the stochastic behavior, the second numerical model comprising a space having a same dimensionality of N, and

wherein the processor is configured to generate the set of approximated system variables based on:

sampling the numerical model using a first statistical inference approach; and

reevaluating a subset of samples in the one or more target domains based on the second numerical model using a second statistical inference approach.

13. The system of claim 12 , wherein the one or more target domains corresponds to one or more failure events with corresponding probabilities of occurrences.

14. The system of claim 11 , wherein the processor is configured to:

determine a set of weighted measurements based on evaluating weights for the set of measurements, wherein the numerical model of the stochastic behavior is determined based on the set of weighted measurements.

15. The system of claim 14 , wherein the weights are evaluated iteratively for the set of measurements.

16. The system of claim 14 , wherein the weights are evaluated in an importance sampling procedure performed based on the statistical inference approach.

17. The system of claim 11 , wherein the processor is configured to:

generate the set of approximated system variables based on expanding the non-correlated features having the dimensionality of M to the set of approximated system variables having the dimensionality of N.

18. The system of claim 11 , wherein the feature space is based on a statistical transformation that converts a first set of correlated values having a first dimensionality into a second set of uncorrelated values having a second dimensionality that is lower than the first dimensionality.

19. The system of claim 18 , wherein the statistical transformation comprises a manifold-learning based Isomap method.

20. The system of claim 18 , wherein the numerical model of the stochastic behavior of the power system comprises a Gaussian process emulator.