Method and device for determining energy system operating scenario

1. A method comprising:

obtaining a plurality of performance measures of an energy system as a function of time corresponding to a plurality of sets of values of input variables;

clustering the plurality of sets of the values of the input variables and the performance measures associated therewith into groups;

defining a plurality of descriptors corresponding respectively to the groups;

using the plurality of descriptors to predict a performance measure of the energy system by determining a group of the groups to which a particular set of values of the input variables under evaluation are assigned and using the performance measure as indicative of a performance of the energy system; and

controlling a parameter of the energy system in accordance with controllable parameters that are associated with the group,

wherein each set of the plurality of sets of values of the input variables comprises a time series of an input variable over a predetermined time frame, the method further comprising determining the plurality of performance measures as a function of time over the predetermined time frame and aggregating the plurality of performance measures over the predetermined time frame,

wherein obtaining the plurality of performance measures comprises down-sampling the time series by partitioning an evolution of the input variable over the predetermined time frame into a plurality of discrete time segments,

wherein the down-sampling comprises performing an optimization process to determine the time segments of the discrete time segments having a non-uniform length, wherein the optimization process comprises an optimization of a cost function comprising a factor indicative of a goodness-of-fit of a down-sampled curve and a factor indicative of a goodness-of-fit of a statistical distribution of a down-sampled variable and a factor indicative of a smoothness of the variable or variables at its temporal resolution before down-sampling within each time segment.

2. The method of claim 1 , wherein the clustering comprises, for each set of the plurality of sets of values of the input variables and the performance measures associated therewith, determining summary statistics for the plurality of sets of values and/or the performance measure and defining a point in a multidimensional space in which the clustering is performed, where this coordinate in the multidimensional space of the point comprises the summary statistics.

3. The method of claim 2 , wherein the clustering comprises determining a similarity metric between points in the multidimensional space and determining the groups based on this similarity metric, the similarity metric comprising a hyper-distance measure in the multidimensional space.

4. The method of claim 2 , wherein the clustering comprises an iterative clustering, wherein the groups are iteratively created and/or updated, and wherein in each clustering step, points in the multidimensional space that were not assigned to a group are flagged for re-evaluation in a further clustering step.

5. The method of claim 4 , wherein, in the clustering, a number of points assigned to each group is, in each clustering step or in each clustering step except the last clustering step, constrained by a predetermined maximum number, and wherein, if the number exceeds the maximum when evaluating a point for adding to a group, a new group is created from the point.

6. The method of claim 4 , wherein each clustering step of the clustering comprises determining spatial properties of each group and taking the spatial properties into account to calculate a relative cost for deciding whether a point is included in the corresponding group, and wherein the point is flagged for the re-evaluation in the further clustering step if the relative cost exceeds a predetermined threshold.

7. The method of claim 1 , wherein the obtaining the plurality of performance measures comprises simulating a multi-physics model of the energy system for a set of the plurality of sets of values of input variables over time.

8. The method of claim 7 , wherein the obtaining the plurality of performance measures comprises, for a performance measure for which an occurrence probability of the corresponding set of values of input variables is below a predetermined threshold, applying an approximation to calculate the performance measure.

9. A device comprising:

an input configured for obtaining a plurality of performance measures of an energy system as a function of time corresponding to a plurality of sets of values of input variables; and

a processor configured to perform functions comprising:

clustering the plurality of sets of values of the input variables and the performance measures associated therewith into groups;

defining a plurality of descriptors corresponding respectively to the groups;

using the plurality of descriptors to predict a performance measure of the energy system by determining a group of the groups to which a particular set of values of the input variables under evaluation are assigned and using the performance measure as indicative of a performance of the energy system; and

controlling a parameter of the energy system in accordance with controllable parameters that are associated with the group,

wherein each set of the plurality of sets of values of the input variables comprises a time series of an input variable over a predetermined time frame, the functions further comprising determining the plurality of performance measures as a function of time over the predetermined time frame and aggregating the plurality of performance measures over the predetermined time frame,

wherein obtaining the plurality of performance measures comprises down-sampling the time series by partitioning an evolution of the input variable over the predetermined time frame into a plurality of discrete time segments,

wherein the down-sampling comprises performing an optimization process to determine the time segments of the discrete time segments having a non-uniform length, wherein the optimization process comprises an optimization of a cost function comprising a factor indicative of a goodness-of-fit of a down-sampled curve and a factor indicative of a goodness-of-fit of a statistical distribution of a down-sampled variable and a factor indicative of a smoothness of the variable or variables at its temporal resolution before down-sampling within each time segment.

10. The device of claim 9 , wherein the clustering comprises, for each set of the plurality of sets of values of the input variables and the performance measures associated therewith, determining summary statistics for the plurality of sets of values and/or the performance measure and defining a point in a multidimensional space in which the clustering is performed, where this coordinate in the multidimensional space of the point comprises the summary statistics.

11. The device of claim 10 , wherein the clustering comprises determining a similarity metric between points in the multidimensional space and determining the groups based on this similarity metric, the similarity metric comprising a hyper-distance measure in the multidimensional space.

12. The device of claim 10 , wherein the clustering comprises an iterative clustering, wherein the groups are iteratively created and/or updated, and wherein in each clustering step, points in the multidimensional space that were not assigned to a group are flagged for re-evaluation in a further clustering step.

13. The device of claim 12 , wherein, in the clustering, a number of points assigned to each group is, in each clustering step or in each clustering step except the last clustering step, constrained by a predetermined maximum number, and wherein, if the number exceeds the maximum when evaluating a point for adding to a group, a new group is created from the point.

14. The device of claim 12 , wherein each clustering step of the clustering comprises determining spatial properties of each group and taking the spatial properties into account to calculate a relative cost for deciding whether a point is included in the corresponding group, and wherein the point is flagged for the re-evaluation in the further clustering step if the relative cost exceeds a predetermined threshold.

15. The device of claim 9 , wherein the obtaining the plurality of performance measures comprises simulating a multi-physics model of the energy system for a set of the plurality of sets of values of input variables over time.

16. The device of claim 15 , wherein the obtaining the plurality of performance measures comprises, for a performance measure for which an occurrence probability of the corresponding set of values of input variables is below a predetermined threshold, applying an approximation to calculate the performance measure.

17. A non-transitory computer readable medium storing instructions that, when executed by a computing device, cause the computing device to perform functions comprising:

obtaining a plurality of performance measures of an energy system as a function of time corresponding to a plurality of sets of values of input variables;

clustering the plurality of sets of values of the input variables and the performance measures associated therewith into groups;

defining a plurality of descriptors corresponding respectively to the groups; and

using the plurality of descriptors to predict a performance measure of the energy system by determining a group of the groups to which a particular set of values of the input variables under evaluation are assigned and using the performance measure as indicative of a performance of the energy system; and

controlling a parameter of the energy system in accordance with controllable parameters that are associated with the group,

wherein each set of the plurality of sets of values of the input variables comprises a time series of an input variable over a predetermined time frame, the functions further comprising determining the plurality of performance measures as a function of time over the predetermined time frame and aggregating the plurality of performance measures over the predetermined time frame,

wherein obtaining the plurality of performance measures comprises down-sampling the time series by partitioning an evolution of the input variable over the predetermined time frame into a plurality of discrete time segments,

wherein the down-sampling comprises performing an optimization process to determine the time segments of the discrete time segments having a non-uniform length, wherein the optimization process comprises an optimization of a cost function comprising a factor indicative of a goodness-of-fit of a down-sampled curve and a factor indicative of a goodness-of-fit of a statistical distribution of a down-sampled variable and a factor indicative of a smoothness of the variable or variables at its temporal resolution before down-sampling within each time segment.

18. The non-transitory computer readable medium of claim 17 , wherein the clustering comprises, for each set of the plurality of sets of values of the input variables and the performance measures associated therewith, determining summary statistics for the plurality of sets of values and/or the performance measure and defining a point in a multidimensional space in which the clustering is performed, where this coordinate in the multidimensional space of the point comprises the summary statistics.

19. The non-transitory computer readable medium of claim 18 , wherein the clustering comprises determining a similarity metric between points in the multidimensional space and determining the groups based on this similarity metric, the similarity metric comprising a hyper-distance measure in the multidimensional space.

20. The non-transitory computer readable medium of claim 18 , wherein the clustering comprises an iterative clustering, wherein the groups are iteratively created and/or updated, and wherein in each clustering step, points in the multidimensional space that were not assigned to a group are flagged for re-evaluation in a further clustering step.