identify.timevarying

TimevaryingIdentificier

Identify components of which the flow results differ between softwares.

Identify submodule holding identification tools for timevarying results.

class tessif.identify.timevarying.TimevaryingIdentificier(data, error_value='nmae', error_value_threshold=0.1, correlation='pearson', correlation_threshold=0.7, conditions_dict=None, reference=None)[source]

Bases: Identificier

Identify components of which the flow results differ between softwares.

Flows are identified using following logic:

High interest / significantly different:

CORR < corr (e.g. CORR < 0.7)

ERROR_VALUE > error_value: (e.g. ERROR_VALUE > 0.1)

Medium interest / borderline significantly different:

CORR < corr (e.g. CORR < 0.7)

ERROR_VALUE < error_value: (e.g. ERROR_VALUE < 0.1)

And:

CORR > corr (e.g. CORR > 0.7)

ERROR_VALUE > error_value: (e.g. ERROR_VALUE > 0.1)

Low Interest / not significantly different:

CORR > corr (e.g. CORR > 0.7)

ERROR_VALUE < error_value: (e.g. ERROR_VALUE < 0.1)

Parameters:

data¶ (dict) –
Dictionairy of multiindexed dataframes of all of all timevarying results of all components keyed inside the dict by the respective software specifier

Top-level index represents the individual components, while the second-level index represent the respective outflow targets.

Meaning for an energy system like:
```
A -> B -> C
     |
     v
     D
```
The dataframe would look something like:
```
    A  B
    B  C  D
0  10  8  2
1   0  0  0
2  20  2 18
```
Usually returned by something like tessif.analyze.ComparativeResultier.all_loads.
error_value¶ (str) –
String abbrevating the error value calculated. Currently supported are:
- nmae for Normalized Mean Average Error (default)
- nmbe for Normalized Mean Biased Error
- nrmse for Normalized Root Mean Square Error
error_value_threshold¶ (float, default = 0.1) – Threshold value used for clustering data interest. Clustering is done according to cluster_conditions.
correlation¶ ({'pearson', 'kendall', 'spearman'} or callable) –
Method of correlation:
- pearson : standard correlation coefficient
- kendall : Kendall Tau correlation coefficient
- spearman : Spearman rank correlation
- callable: callable with input two 1d ndarrays and returning a float.
correlation_threshold¶ (float, default = 0.7) – Threshold value used for clustering data interest. Clustering is done according to cluster_conditions.

conditions_dict¶ (dict, None, default=None) –

Dictionairy keying container(s) of dicts by the respective cluster labels “high”, “medium” and “low”. The dictionairies inside the tuples need to have following keywords:

thres specyfying the threshold used

oprt specifying the operator used.

If None is used, following default conditions are applied:

conditions = {
    "high": (
        {"oprt": "lt", "thres": correlation_coefficient_threshold},
        {"oprt": "ge", "thres": error_value_threshold},
     ),
    "medium1": (
        {"oprt": "ge", "thres": correlation_coefficient_threshold},
        {"oprt": "ge", "thres": error_value_threshold},
    ),
    "medium2": (
        {"oprt": "lt", "thres": correlation_coefficient_threshold},
        {"oprt": "lt", "thres": error_value_threshold},
    ),
    "low": (
        {"oprt": "ge", "thres": correlation_coefficient_threshold},
        {"oprt": "lt", "thres": error_value_threshold},
    ),
}

reference¶ (str, None, default=None) –
Defines the reference results to be used for calculating the statistical error values and pearson correlation coeficients.

In case None is used (default), the dataframes average is used as returned by average_timevarying_dataframe_results().

property mutuals: Mutual components between softwares

property non_mutuals: Non mutual components between softwares

property error_value: String specifying the error value used.

property correlation: String specyfying the correlation used.

property error_value_threshold: Error value threshold used.

property correlation_coefficient_threshold: Correlation value threshold used.

property error_values: Calculated Normalized Medium Average Errors.

property corrs: Calculated Pearson Correlation Coefficients.

property high_interest_averaged_results: Mean and high interest results differing more than threshold.

property medium_interest_averaged_results: Mean and medium interest results differing more than threshold.

property low_interest_averaged_results: Mean and low interest results differing more than threshold.

property high_interest_timeframes: Significant results identified as highly interesting.

property medium_interest_timeframes: Significant results identified as medium interesting.

property low_interest_timeframes: Significant results identified as not very interesting.

cluster_interest()[source]: Cluster inter component results by interest.

map_interest_results(data)[source]: Map data to identified interest categories.