LOAD COLLECTIVES FOR ENERGY STORAGE SYSTEMS
Methods, systems, apparatuses, and non-transitory computer-readable media are provided for using load collectives in energy storage systems. In one implementation, the computer-readable media includes instructions to cause a processor to: receive operational data associated with an energy storage unit for a time period; determine one or more cycles of charging and discharging of the energy storage unit during the time period; generate, based on the operational data, a plurality of load collectives; determine one or more operational parameters of the energy storage unit for the time period; provide, to a machine learning model, the one or more operational parameters and one or more load collectives of the plurality of load collectives; generate, based on the machine learning model, a predicted capacity of the energy storage unit at an end of the time period; and configure, based on the predicted capacity, one or more energy storage units.
1 . A system comprising:
one or more energy storage units; and
a computing device comprising at least one processor and at least one memory storing instructions that, when executed by the at least one processor, cause the computing device to:
receive operational data associated with an energy storage unit of the one or more energy storage units for a period of time;
determine one or more cycles of charging and discharging of the energy storage unit during the period of time;
generate, based on the operational data, a plurality of load collectives, wherein each load collective of the plurality of load collectives includes:
one or more criteria associated with operation of the energy storage unit; and
a quantity of cycles, of the one or more cycles, that satisfy the one or more criteria;
determine one or more operational parameters of the energy storage unit for the period of time;
provide, to a machine learning model, the one or more operational parameters and one or more load collectives of the plurality of load collectives;
generate, based on the machine learning model, a predicted capacity of the energy storage unit at an end of the period of time; and
configure, based on the predicted capacity, the one or more energy storage units.
2 . The system of claim 1 , wherein the operational data includes one or more of:
a temperature of the energy storage unit during each of a plurality of intervals of the period of time;
a voltage level of the energy storage unit during each of the plurality of intervals of the period of time;
a state of charge of the energy storage unit during each of the plurality of intervals of the period of time; or
an amount of current of the energy storage unit during each of the plurality of intervals of the period of time.
3 . The system of claim 1 , wherein the instructions, when executed by the at least one processor, further cause the computing device to:
determine the one or more cycles of charging and discharging of the energy storage unit during the period of time using a rain-flow counting algorithm.
4 . The system of claim 1 , wherein the one or more criteria associated with operation of the energy storage unit comprise one or more of:
a range of a temperature of the energy storage unit;
a range of a voltage level of the energy storage unit;
a range of a minimum state of charge of the energy storage unit;
a range of a maximum state of charge of the energy storage unit;
a range of a sum of amounts of current of the energy storage unit;
a range of a sum of current squared of the energy storage unit during a time segment;
a range of an amount of current of the energy storage unit; or
a range of a depth of charge of the energy storage unit.
5 . The system of claim 1 , wherein one or more criteria associated with operation of the energy storage unit for a first load collective of the plurality of load collectives are different from one or more criteria associated with operation of the energy storage unit for a second load collective of the plurality of load collectives.
6 . The system of claim 1 , wherein the one or more operational parameters of the energy storage unit comprise one or more of:
a quantity of cycles, of charging and discharging of the energy storage unit during the period of time, determined using a rain-flow counting algorithm;
a quantity of equivalent full cycles of charging and discharging of the energy storage unit during the period of time;
a sum of amounts of current of the energy storage unit during the period of time;
an average state of charge of the energy storage unit during the period of time;
a length of the period of time; or
a capacity of the energy storage unit at a beginning of the period of time.
7 . The system of claim 1 , wherein the machine learning model comprises a support vector machine, a relevance vector machine, or a model based on extreme gradient boosting.
8 . The system of claim 1 , wherein the one or more load collectives comprise a subset of the plurality of load collectives.
9 . The system of claim 1 , wherein the instructions, when executed by the at least one processor, further cause the computing device to:
determine, based on a feature reduction technique, the one or more load collectives of the plurality of load collectives.
10 . The system of claim 1 , wherein the instructions, when executed by the at least one processor, further cause the computing device to:
train the machine learning model using historical data including one or more of:
load collectives of the energy storage unit for a prior period of time before the period of time;
operational parameters of the energy storage unit for the prior period of time; or
a measured capacity of the energy storage unit at an end of the prior period of time.
11 . The system of claim 1 , wherein the instructions, when executed by the at least one processor, further cause the computing device to:
configure, based on the predicted capacity, the one or more energy storage units by one or more of: adjusting a pattern for the one or more energy storage units to dispatch electricity, or augmenting a capacity of the one or more energy storage units.
12 . The system of claim 11 , wherein the pattern includes a plurality of time intervals during which the one or more energy storage units are configured to charge or discharge at a particular rate.
13 . The system of claim 1 , wherein the instructions, when executed by the at least one processor, further cause the computing device to:
calculate a degree of influence of each input item of a plurality of input items to the machine learning model on an output of the machine learning model.
14 . The system of claim 1 , wherein the instructions, when executed by the at least one processor, further cause the computing device to:
determine a degree of similarity between the energy storage unit and another energy storage unit; and
based on determining that the degree of similarity satisfies a threshold, use training data for the machine learning model to train a machine learning model for the other energy storage unit.
15 . A method comprising:
receiving, by a computing device, operational data associated with an energy storage unit of the one or more energy storage units for a period of time;
determining one or more cycles of charging and discharging of the energy storage unit during the period of time;
generating, based on the operational data, a plurality of load collectives, wherein each load collective of the plurality of load collectives includes:
one or more criteria associated with operation of the energy storage unit; and
a quantity of cycles, of the one or more cycles, that satisfy the one or more criteria;
determining one or more operational parameters of the energy storage unit for the period of time;
providing, to a machine learning model, the one or more operational parameters and one or more load collectives of the plurality of load collectives;
generating, based on the machine learning model, a predicted capacity of the energy storage unit at an end of the period of time; and
configuring, based on the predicted capacity, the one or more energy storage units.
16 . The method of claim 15 , wherein the machine learning model comprises a support vector machine, a relevance vector machine, or a model based on extreme gradient boosting.
17 . The method of claim 15 , further comprising:
determining, based on a feature reduction technique, the one or more load collectives of the plurality of load collectives.
18 . A non-transitory computer-readable medium storing instructions that, when executed by at least one processor, cause the at least one processor to:
receive operational data associated with an energy storage unit of the one or more energy storage units for a period of time;
determine one or more cycles of charging and discharging of the energy storage unit during the period of time;
generate, based on the operational data, a plurality of load collectives, wherein each load collective of the plurality of load collectives includes:
one or more criteria associated with operation of the energy storage unit; and
a quantity of cycles, of the one or more cycles, that satisfy the one or more criteria;
determine one or more operational parameters of the energy storage unit for the period of time;
provide, to a machine learning model, the one or more operational parameters and one or more load collectives of the plurality of load collectives;
generate, based on the machine learning model, a predicted capacity of the energy storage unit at an end of the period of time; and
configure, based on the predicted capacity, the one or more energy storage units.
19 . The non-transitory computer-readable medium of claim 18 , wherein the machine learning model comprises a support vector machine, a relevance vector machine, or a model based on extreme gradient boosting.
20 . The non-transitory computer-readable medium of claim 18 , wherein the instructions, when executed by the at least one processor, further cause the at least one processor to:
determine, based on a feature reduction technique, the one or more load collectives of the plurality of load collectives.