# US 12,113,379 B2 · Machine learning-based method for increasing lifetime of a battery energy storage system > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,113,379 B2 - Document ID: US12113379B2 - Status: Granted - Kind: B2 - Application number: 17/452,340 - Filed: 2021-10-26 - Published: 2022-05-05 - Granted: 2024-10-08 - Earliest claimed: 2020-10-30 - Assignees: ABB Schweiz AG - Applicants: ABB Schweiz AG - Inventors: Carlos Nieto; Kacper Sowa; Piotr Oramus; Hemangi Borole; Tomasz Kuczek - Web page: https://patentgrep.com/patent/US12113379B2 ## CPC - H02J 3/38: Arrangements for feeding a single network from two or more generators or sources in parallel; Arrangements for feeding already energised networks from additional generators or sources in parallel - H02J 7/44: between battery management systems and power sources - G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] - G06N 3/045: Combinations of networks - G06N 3/047: Probabilistic or stochastic networks - G06N 3/0499: Feedforward networks - G06N 3/088: Non-supervised learning, e.g. competitive learning - G06N 3/09: Supervised learning - G06N 3/092: Reinforcement learning - G06N 20/00: Machine learning - H02J 3/322: {the battery being on-board an electric or hybrid vehicle, e.g. vehicle to grid arrangements [V2G], power aggregation, use of the battery for network load balancing, coordinated or cooperative battery charging} - H02J 7/92: with prioritisation of loads or sources - H02J 2101/24: Photovoltaics - H02J 2103/30: Simulating, planning, modelling, reliability check or computer assisted design [CAD] of electric power networks - Y02E 10/56: Power conversion systems, e.g. maximum power point trackers - Y02E 40/10: Flexible AC transmission systems [FACTS] - Y02E 60/00: Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation - Y04S 40/20: Information technology specific aspects, e.g. CAD, simulation, modelling, system security - H02J: ELECTRIC POWER NETWORKS; CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - Y02E: REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION - Y04S: SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS ## Abstract An apparatus for performing the following: the apparatus maintains, in a database, one or more trained machine learning algorithms for predicting an optimal charging strategy for a time interval based on one or more values of a set of prediction parameters relating to a point of common coupling and one or more electrical load devices and on a state of charge level of the battery energy storage system. The apparatus obtains one or more recent values of the set of prediction parameters relating to one or more previous time intervals and predicts, using the one or more trained machine learning algorithm, an optimal charging strategy for a next time interval based on the one or more recent values and a current state of charge level of the battery energy storage system. Finally, the apparatus operates the battery energy storage system using the predicted optimal charging strategy. ## Claims ### Claim 1 A first apparatus for managing a battery energy storage system for storing energy supplied via a point of common coupling of an electric power supply network for use by one or more electrical load devices, the first apparatus comprising: at least one processor, and at least one memory for storing instructions to be executed by the processor, wherein the at least one memory and the instructions are configured to, with the at least one processor, cause the first apparatus at least to perform: maintaining, in a first database, one or more trained machine learning algorithms for predicting an optimal charging strategy for a time interval of a pre-defined length based at least on one or more values of a set of prediction parameters and a state of charge level of at least one battery of the battery energy storage system, wherein the set of prediction parameters include at least one or more first parameters of the point of common coupling and one or more second parameters of the one or more electrical load devices, wherein the one or more trained machine learning algorithms comprise a first trained machine learning algorithm for predicting an expected load profile at the point of common coupling defining excess load to be covered by the battery energy storage system for the time interval of the pre-defined length based on the one or more values of the set of prediction parameters and a second trained machine learning algorithm for predicting the optimal charging strategy based on the expected load profile and a current state of charge level of said at least one battery of the battery energy storage system; obtaining one or more recent values of the set of prediction parameters relating to one or more previous time intervals of the pre-defined length; predicting, using the one or more trained machine learning algorithms, an optimal charging strategy for maximizing lifespan of said at least one battery of the battery energy storage system for a next time interval of the pre-defined length based at least on the one or more recent values of the set of prediction parameters and the current state of charge level of said at least one battery of the battery energy storage system, wherein the predicting using the one or more trained machine learning algorithms is performed by calculating, using the first trained machine learning algorithm, an expected load profile at the point of common coupling defining excess load to be covered by the battery energy storage system for the next time interval of the pre-defined length based at least one the one or more recent values of the set of prediction parameters, and predicting, using the second trained machine learning algorithm, the charging strategy for said next time interval based on a calculated expected load profile and the current state of charge level of said at least one battery of the battery energy storage system for maximizing lifespan of said at least one battery of the battery energy storage system; and operating the battery energy storage system using the predicted optimal charging strategy during said next time interval, wherein the at least one memory and the instructions are configured to, with the at least one processor, cause the first apparatus to perform, monitoring electrical power supplied by the power supply network via the point of common coupling and charging of the one or more electrical load devices during the operating of the battery energy storage system during said next time interval using the predictive charging strategy; calculating a realized load profile at the point of common coupling defining excess load covered by the battery energy storage system for a time interval of the pre-defined length based on results of the monitoring of the electrical power supplied by the power network via the point of common coupling and the charging of the one or more electrical load devices; and training the first trained machine learning algorithm using the one or more recent values of the set of prediction parameters and the realized load profile as training data. ### Claim 2 (depends on 1) The first apparatus according to claim 1, wherein the one or more values of the set of prediction parameters correspond, for at least one of the set of parameters, to a plurality of values defining a time series over at least one of the one or more previous time intervals of the pre-defined length. ### Claim 3 (depends on 1) The first apparatus according to claim 1, wherein the one or more previous time intervals comprise a most recent time interval of the pre-defined length directly preceding said next time interval and/or one or more-time intervals of the pre-defined length preceding said most recent time interval. ### Claim 4 (depends on 1) The first apparatus according to claim 1, wherein the one or more first parameters of the point of common coupling include one or more electric parameters associated with power transfer to the battery energy storage system and/or to the one or more electrical load devices selected from a group including: voltage, current, frequency, power, active power, reactive power, maximum available power, peak power and energy. ### Claim 5 (depends on 1) The first apparatus according to claim 1, wherein the one or more second parameters of the one or more electrical load devices include, for each of the one or more electrical load devices, at least load demand. ### Claim 6 (depends on 1) The first apparatus according to claim 1, wherein the one or more electrical load devices comprise one or more electric vehicle chargers and the one or more second parameters of the one or more electric vehicle chargers include one or more parameters selected from a group of: a type or model of a car charged by the one or more electric vehicle chargers or a distribution thereof covering a time interval of the pre-defined length, an initial and final state of charge level of a car charged by the one or more electric vehicle chargers or distributions thereof covering a time interval of the pre-defined length, energy transferred to cars, by the one or more electric vehicle chargers, over a time interval of the pre-defined length, charging power of the one or more electric vehicle chargers over the time interval of the pre-defined length and energy transferred from the BESS to the one or more electric vehicle chargers over the time interval of the pre-defined length. ### Claim 7 (depends on 1) The first apparatus according to claim 1, wherein the set of prediction parameters further comprise one or more third parameters of one or more photovoltaic systems connected to the point of common coupling and configured to supply power to the battery energy storage system and the one or more electrical load devices. ### Claim 8 (depends on 7) The first apparatus according to claim 7, wherein the one or more third parameters of the one or more photovoltaic systems include, for each of the one or more photovoltaic systems, one or more parameters selected from a group including: maximum available power and past energy generation during a past time interval of the pre-defined length and expected future energy generation during a future time interval of the pre-defined length. ### Claim 9 (depends on 1) The first apparatus according to claim 1, wherein the optimal charging strategy for the time interval of the pre-defined length predicted using the one or more machine-learning algorithms defines one or more of the following: one or more optimal periods for charging said at least one battery of the battery energy storage system; one or more optimal periods for discharging said at least one battery of the battery energy storage system; overall charging needed during said next time interval of the pre-defined length; a charging power during one or more optimal periods of charging; and a C-rate during one or more optimal periods of discharging. ### Claim 10 (depends on 1) The first apparatus according to claim 1, wherein the at least one memory and the instructions are configured to, with the at least one processor, cause the first apparatus to perform: monitoring one or more parameters of the battery energy storage system during the operating of the battery energy storage system during said next time interval using the predicted charging strategy, wherein the one or more parameters comprise at least a state of charge level; calculating an optimal charging strategy for the time interval of the pre-defined length or a part thereof based on results of the monitoring of the one or more parameters of the battery energy storage system and the realized load profile during that time interval; and training the second trained machine learning algorithm using the realized load profile, a state of charge level of the battery energy storage system prior to the time interval of the pre-defined length and the calculated optimal charging strategy as training data. ### Claim 11 (depends on 1) The first apparatus according to claim 1, wherein the time interval of the pre-defined length has a length selected from a range of 1-48 hours. ### Claim 12 (depends on 1) The first apparatus according to claim 1, wherein the one or more electric load devices comprise one or more electric vehicle chargers. ### Claim 13 (depends on 1) The first apparatus according to claim 1, wherein each of the one or more machine learning algorithms is based on statistical classification, regression analysis, reinforcement learning, multi-agent learning, decision tree-based learning or a combination thereof. ### Claim 14 A second apparatus comprising: at least one processor, and at least one memory for storing instructions to be executed by the processor, wherein the at least one memory and the instructions are configured to, with the at least one processor, cause the second apparatus at least to perform: maintaining, in a second database, a first training data set including a plurality of values of a set of prediction parameters corresponding to a plurality of time intervals of a pre-defined length and a plurality of load profiles at a point of common coupling of an electric power supply network defining excess load to be covered by a battery energy storage system corresponding to said plurality of time intervals, wherein the set of prediction parameters include at least one or more first parameters of the point of common coupling and one or more second parameters of one or more electrical load devices connected to the point of common coupling and the battery energy storage system; initializing a first machine learning algorithm; training the first machine learning algorithm using the first training data set, wherein the set of prediction parameters correspond to features of the first machine learning algorithm and the plurality of load profiles correspond to a label of the first machine learning algorithm; causing storing the first trained machine learning algorithm to one or more of said second database and a first database of a first apparatus for enabling applying of the first trained machine learning algorithm for online calculation of an expected load profile at the point of common coupling based at least on one or more recent values of the set of prediction parameters; maintaining, in the second database, a second training data set including the plurality of load profiles, a plurality of state of charge levels of at least one battery of the battery energy storage system, and a plurality of optimal charging strategies of the battery energy storage system connected to said point of common coupling and to one or more electrical load devices for said plurality of time intervals corresponding to the plurality of time intervals of the pre-defined length; initializing a second machine learning algorithm; training the second machine learning algorithm using the second training data set, wherein a load profile and a state of charge level of said at least one battery of the battery energy storage system correspond to features of the second machine learning algorithm and an optimal charging strategy corresponds to a label of the second machine learning algorithm; and causing storing the second trained machines learning algorithm to one or more of said second database and the first database for enabling applying of the trained second machine learning algorithm for online calculation of an optimal charging strategy for the battery energy storage system based on an expected load profile for a next time interval and a current state of charge level of at least one battery of the battery energy storage system. ### Claim 15 (depends on 14) The second apparatus according to claim 14, wherein the at least one memory and the instructions are configured to, with the at least one processor, cause the second apparatus to perform: maintaining, in the second database, one or more trained machine learning algorithms for predicting an optimal charging strategy for a time interval of a pre-defined length based at least on one or more values of a set of prediction parameters and a state of charge level of at least one battery of the battery energy storage system, wherein the one or more trained machine learning algorithms comprise the first and second trained machine learning algorithms, the set of prediction parameters comprise at least one or more first parameters of the point of common coupling and one or more second parameters of the one or more electrical load devices; obtaining one or more recent values of the set of prediction parameters relating to one or more previous time intervals of the pre-defined length; predicting, using the one or more trained machine learning algorithms, an optimal charging strategy for maximizing lifespan of said at least one battery of the battery energy storage system for a next time interval of the pre-defined length based at least on the one or more recent values of the set of prediction parameters and a current state of charge level of said at least one battery of the battery energy storage system; and operating the battery energy storage system using the predicted optimal charging strategy during said next time interval. ### Claim 16 A second apparatus comprising: at least one processor, and at least one memory for storing instructions to be executed by the processor, wherein the at least one memory and the instructions are configured to, with the at least one processor, cause the second apparatus at least to perform: maintaining, in a second database, a second training data set comprising a plurality of load profiles at a point of common coupling of an electric power supply network defining excess load to be covered by a battery energy storage system, a plurality of state of charge levels of at least one battery of the battery energy storage system and a plurality of optimal charging strategies of the battery energy storage system connected to said point of common coupling and to one or more electrical load devices for said plurality of time intervals, wherein the plurality of load profiles and the plurality of optimal charging strategies correspond to a plurality of time intervals of a pre-defined length; initializing a second machine learning algorithm; training the second machine learning algorithm using the second training data set, wherein a load profile and a state of charge level of said at least one battery of the battery energy storage system corresponds to features of the second machine learning algorithm and an optimal charging strategy correspond to a label of the second machine learning algorithm; and causing storing the second trained machine learning algorithm to one or more of said second database and a first database of a first apparatus for enabling applying of the trained second machine learning algorithm for online calculation of an optimal charging strategy for the battery energy storage system based on an expected load profile for a next time interval and a current state of charge level of at least one battery of the battery energy storage system. ### Claim 17 A method for managing a battery energy storage system for storing energy supplied via a point of common coupling of an electric power supply network for use by one or more electrical load devices, the method comprising: maintaining, in a first database, one or more trained machine learning algorithms for predicting an optimal charging strategy for a time interval of a pre-defined length based at least on one or more values of a set of prediction parameters and a state of charge level of at least one battery of the battery energy storage system, wherein the set of prediction parameters comprise at least one or more first parameters of the point of common coupling and one or more second parameters of the one or more electrical load devices, wherein the one or more trained machine learning algorithms comprise a first trained machine learning algorithm for predicting an expected load profile at the point of common coupling defining excess load to be covered by the battery energy storage system for the time interval of the pre-defined length based on the one or more values of the set of prediction parameters and a second trained machine learning algorithm for predicting the optimal charging strategy based on the expected load profile and a current state of charge level of said at least one battery of the battery energy storage system; obtaining one or more recent values of the set of prediction parameters relating to one or more previous time intervals of the pre-defined length; predicting, using the one or more trained machine learning algorithm, an optimal charging strategy for maximizing lifespan of said at least one battery of the battery energy storage system for a next time interval of the pre-defined length based at least on the one or more recent values of the set of prediction parameters and a current state of charge level of said at least one battery of the battery energy storage system, wherein the predicting using the one or more trained machine learning algorithms is performed by calculating, using the first trained machine learning algorithm, an expected load profile at the point of common coupling defining excess load to be covered by the battery energy storage system for the next time interval of the pre-defined length based at least one the one or more recent values of the set of prediction parameters, and predicting, using the second trained machine learning algorithm, the charging strategy for said next time interval based on a calculated expected load profile and the current state of charge level of said at least one battery of the battery energy storage system for maximizing lifespan of said at least one battery of the battery energy storage system; and operating the battery energy storage system using the predicted optimal charging strategy during said next time interval, wherein the method comprises: monitoring electrical power supplied by the power supply network via the point of common coupling and charging of the one or more electrical load devices during the operating of the battery energy storage system during said next time interval using the predictive charging strategy; calculating a realized load profile at the point of common coupling defining excess load covered by the battery energy storage system for a time interval of the pre-defined length based on results of the monitoring of the electrical power supplied by the power network via the point of common coupling and the charging of the one or more electrical load devices; and training the first trained machine learning algorithm using the one or more recent values of the set of prediction parameters and the realized load profile as training data. ### Claim 18 A computer program stored on a non-transitory computer readable medium, the computer program comprising instructions for performing at least the following: obtaining one or more recent values of a set of prediction parameters relating to one or more previous time intervals of a pre-defined length, wherein the set of prediction parameters comprise at least one or more first parameters of a point of common coupling and one or more second parameters of one or more electrical load devices, the point of common coupling being connected to a battery energy storage system, an electric power supply network and the one or more electrical load devices; predicting, using one or more trained machine learning algorithms, an optimal charging strategy for maximizing lifespan of said at least one battery of the battery energy storage system for a next time interval of the pre-defined length based at least on the one or more recent values of the set of prediction parameters and a current state of charge level of said at least one battery of the battery energy storage system, wherein the one ore more trained machine learning algorithms comprise a first trained machine learning algorithm for predicting an expected load profile at the point of common coupling defining excess load to be covered by the battery energy storage system for the time interval of the pre-defined length based on the one or more values of the set of prediction parameters and a second trained machine learning algorithm for predicting the optimal charging strategy based on the expected load profile and the current state of charge level of said at least one battery of the battery energy storage system, and wherein the predicting using the one or more trained machine learning algorithms is preformed by calculating, using the first trained machine learning algorithm, an expected load profile at the point of common coupling defining excess load to be covered by the battery energy storage system for the next time interval of the pre-defined length based at least on the one or more recent values of the set of prediction parameters, and predicting, using the second trained machine learning algorithm, the charging strategy for said next time interval based on a calculated expected load profile and the current state of charge level of said at least one battery of the battery energy storage system for maximizing lifespan of said at least one battery of the battery energy storage system; operating the battery energy storage system using the predicted optimal charging strategy during said next time interval; monitoring electrical power supplied by the power supply network via the point of common coupling and charging of the one or more electrical load devices during the operating of the battery energy storage system during said next time interval using the predictive charging strategy; calculating a realized load profile at the point of common coupling defining excess load covered by the battery energy storage system for a time interval of the pre-defined length based on results of the monitoring of the electrical power supplied by the power network via the point of common coupling and the charging of the one or more electrical load devices; and training the first trained machine learning algorithm using the one or more recent values of the set of prediction parameters and the realized load profile as training data. ### Claim 19 A method comprising: maintaining, in a second database, a first training data set comprising a plurality of values of a set of prediction parameters corresponding to a plurality of time intervals of a pre-defined length and a plurality of load profiles at a point of common coupling of an electric power supply network defining excess load to be covered by a battery energy storage system corresponding to said plurality of time intervals, wherein the set of prediction parameters comprise at least one or more first parameters of the point of common coupling and one or more second parameters of one or more electrical load devices connected to the point of common coupling and the battery energy storage system; initializing a first machine learning algorithm; training the first machine learning algorithm using the first training data set, wherein the set of prediction parameters correspond to features of the first machine learning algorithm and the plurality of load profiles correspond to a label of the first machine learning algorithm; causing storing the first trained machine learning algorithm to one or more of said second database and a first database of a first apparatus for enabling applying of the first trained machine learning algorithm for online calculation of an expected load profile at the point of common coupling based at least on one or more recent values of the set of prediction parameters; maintaining, in the second database, a second training data set including the plurality of load profiles, a plurality of state of charge levels of at least one battery of the battery energy storage system, and a plurality of optimal charging strategies of the battery energy storage system connected to said point of common coupling and to one or more electrical load devices for said plurality of time intervals corresponding to the plurality of time intervals of the pre-defined length; initializing a second machine learning algorithm; training the second machine learning algorithm using the second training data set, wherein a load profile and a state of charge level of said at least one battery of the battery energy storage system correspond to features of the second machine learning algorithm and an optimal charging strategy corresponds to a label of the second machine learning algorithm; and causing storing the second trained machines learning algorithm to one or more of said second database and the first database for enabling applying of the trained second machine learning algorithm for online calculation of an optimal charging strategy for the battery energy storage system based on an expected load profile for a next time interval and a current state of charge level of at least one battery of the battery energy storage system. ### Claim 20 A computer program stored on a non-transitory computer readable medium, the computer program comprising instructions for performing at least the following: initializing a first machine learning algorithm; training the first machine learning algorithm using a first training data set comprising a plurality of values of a set of prediction parameters corresponding to a plurality of time intervals of a pre-defined length and a plurality of load profiles at a point of common coupling of an electric power supply network defining excess load to be covered by a battery energy storage system corresponding to said plurality of time intervals, wherein the set of prediction parameters comprise at least one or more first parameters of the point of common coupling and one or more second parameters of one or more electrical load devices connected to the point of common coupling and the battery energy storage system and the set of prediction parameters correspond to features of the first machine learning algorithm and the plurality of load profiles correspond to a label of the first machine learning algorithm; and causing storing the first trained machine learning algorithm to one or more of said second database and a first database of a first apparatus for enabling applying of the first trained machine learning algorithm for online calculation of an expected load profile at the point of common coupling based at least on one or more recent values of the set of prediction parameters; initializing a second machine learning algorithm; training the second machine learning algorithm using a second training data set comprising the plurality of load profiles, a plurality of state of charge levels of at least one battery of the battery energy storage system, and a plurality of optimal charging strategies of the battery energy storage system connected to said point of common coupling and to one or more electrical load devices for said plurality of time intervals corresponding to the plurality of time interval of the pre-defined length, wherein a load profile and a state of charge level of said at least one battery of the battery energy storage system correspond to features of the second machine learning algorithm and an optimal charging strategy corresponds to a label of the second machine learning algorithm; and causing storing the second trained machine learning algorithm to one or more of said second database and the first database for enabling applying of the trained second machine learning algorithm for online calculation of an optimal charging strategy for the battery energy storage system based on an expected load profile for a next time interval and a current state of charge level of at least one battery of the battery energy storage system. ### Claim 21 A method comprising: maintaining, in a second database, a second training data set comprising a plurality of load profiles at a point of common coupling of an electric power supply network defining excess loads to be covered by a battery energy storage system, a plurality of state of charge levels of at least one battery of the battery energy storage system and a plurality of optimal charging strategies of the battery energy storage system connected to said point of common coupling and to one or more electrical load devices, wherein the plurality of load profiles, the plurality of state of charge levels and the plurality of optimal charging strategies correspond to a plurality of time intervals of a pre-defined length; initializing a second machine learning algorithm; training the second machine learning algorithm using the second training data set, wherein a load profile and a state of charge level of said at least one battery of the battery energy storage system corresponds to features of the second machine learning algorithm and an optimal charging strategy correspond to a label of the second machine learning algorithm; and causing storing the second trained machine learning algorithm to one or more of said second database and a first database of a first apparatus for enabling applying of the trained second machine learning algorithm for online calculation of an optimal charging strategy for the battery energy storage system based on a expected load profile for a next time interval and a current state of charge level of at least one battery of the battery energy storage system. ### Claim 22 A computer program stored on a non-transitory computer readable medium, the computer program comprising instructions for performing at least the following: initializing a second machine learning algorithm; training the second machine learning algorithm using a second training data set comprising a plurality of load profiles at a point of common coupling of an electric power supply network defining excess loads to be covered by a battery energy storage system, a plurality of state of charge levels of at least one battery of the battery energy storage system and a plurality of optimal charging strategies of the battery energy storage system connected to said point of common coupling and to one or more electrical load devices, wherein the plurality of load profiles, the plurality of state of charge levels and the plurality of optimal charging strategies correspond to a plurality of time intervals of a pre-defined length; causing storing the second trained machine learning algorithm to one or more of said second database and a first database of a first apparatus for enabling applying of the trained second machine learning algorithm for online calculation of an optimal charging strategy for the battery energy storage system based on an expected load profile for a next time interval and a current state of charge level of at least one battery of the battery energy storage system.