Reliant monotonic charging controllers for smoothing power fluctuations and battery degradation reduction

Abstract

A reliant monotonic charging / discharging controller RMCC can allocate charging and discharging based on state of charge estimates received from a battery management system. A battery storage system (BESS) can include a first group of first battery units. The BESS can further include a second group of second battery units. Additionally, the BESS can include a controller. The controller can receive state of charge estimates associated with the first group and the second group. Additionally, the control can control, based on a comparison of at least one of the state of charge estimates with a charging limit and absent additional sensor data, a first operation of the first group and a second operation of the second group. The first operation can involve charging the first group. The second operation can involve discharging the second group and can be concurrent with the first operation.

Description

 CROSS-REFERENCES TO RELATED APPLICATIONS

[0001] This application claims the benefit of U.S. Provisional Patent Application No. 63 / 558,843, filed February 28, 2024, the entire contents of which are hereby incorporated by reference for all purposes in its entirety.BACKGROUND OF THE INVENTION

[0002] Globally, carbon footprints of conventional power plants can cause concern. An increasing share of renewable energy (RE) resources like photovoltaic (PV) power systems can curb greenhouse-gas emissions. But intermittent solar irradiance can create rapid power drops and power fluctuations, resulting in voltage and frequency deviations, wear and tear of on-load tap changers, and voltage flickers. Power ramp rate (RR) limits can be introduced into power grid regulations of various utilities around to reduce consequences of RE intermittency. For example, commercial power grids can define a maximum RR as a positive incremental of power up to some value (e.g. 1 MW / min, 30 MW / min, 10% / min, etc.). Integrating renewable energy sources with energy storage systems can enable flexible energy management to mitigate negative impacts of intermittency before power is dispatched into the grid.BRIEF SUMMARY OF THE INVENTION

[0003] A reliant monotonic charging / discharging controller RMCC can allocate charging and discharging based on state of charge estimates received from a battery management system. For example, a battery energy storage system (BESS) described herein can include a first group of first battery units. The BESS can further include a second group of second battery units. Additionally, the BESS can include a controller. The controller can receive state of charge estimates associated with the first group and the second group. Additionally, the control can control, based on a comparison of at least one of the state of charge estimates with a charging limit and absent additional sensor data, a first operation of the first group and a second operation of the second group. The first operation can involve charging the first group. The second operation can involve discharging the second group and can be concurrent with the first operation.

[0004] In another example, a method described herein can involve receiving state of charge estimates associated with first battery units of a first battery unit group and with second battery units for a battery energy storage system (BESS). The method can further involve comparing at least one of the state of charge estimates with a charging limit. Additionally, the method can involve controlling, based on the comparison, a first operation of the first battery unit group. The first operation can involve charging the first battery unit group. The method can involve controlling, based on the comparison, a second operation of the second battery unit group. The second operation can involve discharging the second battery unit group and can be concurrent with the first operation.

[0005] In another example, a non-transitory computer-readable medium described herein can include instructions that upon execution on a system, can cause the system to perform operations. The operations can involve receiving a state of charge estimates associated with a first group of first battery units and a second group of second battery units for a battery energy storage system (BESS). The operations can further involve comparing at least one of the state of charge estimates with a charging limit. Additionally, the operations can involve controlling, based on the comparison, a first operation of the first group. The first operation can include charging the first group. The operations can involve controlling, based on the comparison, a second operation of the second group. The second operation can involve discharging the second group and can be concurrent with the first operation.BRIEF DESCRIPTION OF THE DRAWINGS

[0006] FIG. 1 is a schematic of PV power smoothing with a power sharing model and parallel operation of battery storage units according to some aspects of the present application.

[0007] FIG. 2A and 2B are exemplary graphs of solar irradiance and variability index throughout a year according to some aspects of the present application.

[0008] FIG. 3 is a schematic of a power smoothing and a RMCC-based power allocation model according to some aspects of the present application.

[0009] FIG. 4 is a flowchart of a charging / discharging operation performed by a RMCC according to some aspects of the present application.

[0010] FIG. 5 is a schematic of a framework of a smart monotonic charging controller according to some aspects of the present application.

[0011] FIG. 6 is a flow chart for a power allocation model of a S-RMCC according to some aspects of the present application.

[0012] FIG. 7 is a flow chart of an operation mode selector of a S-RMCC according to some aspects of the present application.

[0013] FIGs. 8A, 8B, 8C, and 8D are exemplary graphs of state of charge and power for battery units in an alpha mode of a S-RMCC according to some aspects of the present application.

[0014] FIGs. 9A and 9B are exemplary schematics of a long short-term memory (LSTM) memory cell and stacked LSTM memory architecture according to some aspects of the present application.

[0015] FIG. 10 is an exemplary graph of cycle number versus depth of discharge for a battery according to some aspects of the present application.

[0016] FIGs. 11A, 11B, and 11C are exemplary graphs illustrating smoothing characteristics of a RMCC with equivalent battery units according to some aspects of the present application.

[0017] FIG. 12 is an exemplary graph depicting a time dependence of variability reduction using a RMCC with equivalent battery units according to some aspects of the present application.

[0018] FIGs. 13A, 13B, and 13C are exemplary graphs illustrating a performance of an independent monotonic charging controller according to some aspects of the present application.

[0019] FIG. 14 is an exemplary graph depicting a time dependence of variability reduction using an independent monotonic charging controller according to some aspects of the present application.

[0020] FIGs. 15A, 15B, and 15C are exemplary graphs illustrating smoothing characteristics of a RMCC with battery units of different capacities according to some aspects of the present application.

[0021] FIG. 16 is an exemplary graph depicting a time dependence of variability reduction using a RMCC with battery units of different capacities according to some aspects of the present application.

[0022] FIGs. 17A, 17B, 17C, and 17D are exemplary graphs of energy profiles and operation mode selections for battery units operating under a S-RMCC scheme according to some aspects of the present application.

[0023] FIGs. 18A, 18B, 18C, 18D, 18E, 18F, 18G, and 18H are exemplary graphs of charge level and power sharing for battery units operating under a S-RMCC scheme according to some aspects of the present application.

[0024] FIG. 19 is an exemplary graph depicting a time dependence of variability reduction using a S-RMCC according to some aspects of the present application.

[0025] FIGs. 20A, 20B, and 20C are exemplary histograms illustrating frequency distributions of cycle numbers at various discharge levels for battery units according to some aspects of the present application.

[0026] FIGs. 21A, 21B, 21C, and 21D are exemplary graphs of cycle counts for battery units operating under a S-RMCC scheme according to some aspects of the present application.

[0027] FIGs. 22A and 22B are exemplary graphs of cycle counts for battery units operating under a RMCC scheme according to some aspects of the present application.

[0028] FIG. 23 is an exemplary graph of ramp rate compliance evaluation of a RMCC compared to an independent monotonic charging controller according to some aspects of the present application.

[0029] FIG. 24 is a flow chart of a process 2400 for controlling battery units of a BESS according to certain aspects of the present disclosure.

[0030] FIG. 25 is a block diagram of a controller 2500 for a BESS according to certain aspects of the present disclosure.DETAILED DESCRIPTION OF THE INVENTION

[0031] Smoothing techniques to smooth out power fluctuations can vary according to operation, resulting in more or less battery operation depending on a charge / discharge power profile sent to a battery energy storage system (BESS). A wide range of solutions can reduce variability of PV power generation typically using filtering techniques like, low pass filters (LPFs), moving average (MA) filters, or using ramp rate control, or rule-based algorithms. Smoothing efficiency of MA filters and LPFs can be enhanced by increasing a window length of the filters. However, MA filters can rely on previous values in calculating a present value, known as a memory effect, which can cause a target output (e.g., a reference power) to lag an original PV power resulting in more charging-discharging cycles for a battery system. RR control can have no memory effect when compared to MA control. Nevertheless, battery charge level or a so-called State-of-Charge (SOC) may not return to an initial value after smoothing, implying that a battery does not dispatch all stored power for smoothing or dispatch more than the stored power. Although most of smoothing solutions can exhibit good smoothing performance and minimize ramp rate of the dispatched power, the smoothing solutions may yield a repetitive change in SOC or up-and-down behavior before reaching maximum () or minimum charging limits .

[0032] SOC feedback control (SFC) can be employed to govern battery operation between charginglimits during power smoothing. A BESS can be used to mitigate wind power fluctuations using SFC to keep a charge level of a battery within a desired range during smoothing. The smoothing performance can be compromised to maintain the battery between 30% and 70% of its capacity. An implementation of a control strategy based on SFC can modulate a power output of a wind-PV hybrid power generation in combination with multiple energy storage systems (ESS). Power smoothing can be achieved by limiting battery operation to - limits and allowing a target power regulation between connected battery storage systems to prevent overcharging or depletion of the BESS. However, a controller can evenly distribute target power among BESS units according to charge level. As a result, all BESS units can have a single charging and discharging patterns if the charge level of the BESS device is between permissible limits. An adaptive cutoff frequency for filter-based smoothing can modulate shared reference power between parallel-connected BESS units to avoid overcharging or draining of energy. When all BESS systems are discharging, the controller can instruct a BESS unit with a highest level of charge to supply more energy; when the BESS system is charging, the controller can instruct a BESS unit with a lowest SOC to charge more. Metaheuristic optimization can also be used to find an optimal cutoff frequency that provides a balance between smoothing efficiency and power storage capacity. An optimal capacity of the BESS (e.g., a number of BESS units and a capacity of individual units) can be coupled with a wind farm to reduce fluctuations of wind power. However, the optimal capacity can be determined under an assumption that parallel-connected battery units can have a same capacity and that the battery units can be independently controlled. Thus, while a first group is charging, a second group can be discharging regardless of a state of the first group. Also, if the charging group reaches a limit, the second group may continue to charge. Typically, BESSs connected to REs can undergo multiple charge / discharge cycles over a short period to compensate for fluctuated power profiles which might affect life span due to a low power density. In contrast, a parallel-connected energy storage system may help to avoid a problem of repetitive charging and discharging of BESS units by controlling the BESS units monotonically. Monotonic charging / discharging can refer to a use of parallel BESS units, some of which can operate in a charge mode while others operate in a discharge mode. When charging and discharging limits are reached, states of the BESS devices can change from charging to discharging and vice versa. Monotonic charging / discharging control can reduce a number of battery cycles and can extend a life of the BESS.

[0033] Certain aspects and examples of the present disclosure relate to a reliant monotonic charging / discharging controller (RMCC) framework integrated with a smoothing controller for a BESS that includes multiple units (e.g, battery units). The RMCC can monitor a charging status of each unit and can allocate charging and discharging power profiles to the units to reduce a variability of PV power generation. and values can govern a state transition of the BESS units as one unit is converted from one state to another or if an SOC value of the first or second unit reaches charging level boundaries. Additionally, a smart RMCC (or S-RMCC) is disclosed that can manage an operation of a BESS made up of an even number of battery units greater than two. The S-RMCC can use a ramp event detector to detect a presence of power ramps after utilizing a long-short-term memory (LSTM) model to predict irradiance for an upcoming predetermined duration of time (e.g., 30 minutes). The S-RMCC can determine a number of battery units to activate to reduce power fluctuations. An effectiveness of the RMCC and S-RMCC can be demonstrated using a variety of scenarios and an assessment of an impact on battery degradation.

[0034] The RMCC can act as a supervisory controller for a BESS that includes two subunits to reduce degradation associated with a single battery unit. The RMCC can manage an operation of the two subunits and allocate power to the subunits based on charge levels of the subunits. The RMCC can maintain a monotonic operation to reduce a number of battery cycles while smoothing out power fluctuations. An event driven S-RMCC can manage an operation of a BESS made up of an even number of battery units greater than two (e.g., four, six, eight, etc.). The S-RMCC can maintain a balance between smoothing, which can enable an operation of a complete BESS system and minimizing a number of units involved in handling an upcoming ramp event to lower battery aging. The RMCC and S-RMCC frameworks can be used for a variety of smoothing controllers, including filtering techniques and ramp rate controllers, whereas other approaches (e.g., adaptive smoothing) may only be specifically designed for one particular filter to slow battery degradation.

[0035] Operations implemented by the RMCC or the S-RMCC can enhance many types of applications. For example, the operations can be applied to large utility-scale battery storage systems, where battery packs can be interconnected through inverter units. For behind-the-meter applications, homeowners with access to solar panels and multiple battery systems can leverage the RMCC to optimize utilization of solar energy and extend battery system lifespans. Indendent system owners (ISOs) with multiple battery systems that can offer ancillary services to electrical utilities for financial rewards can use the operations to boost efficiency and longevity of energy storage systems. These operations can result in improved management, lower expenses due to decreased battery aging and fewer replacements, and enhanced cost-saving measures. Network towers can also benefit from the operations implemented by the RMCC or S-RMCC. Network towers can use RMCC or S-RMCC supported battery systems to ensure operational continuity during power outages, which can be critical for maintaining communication services.

[0036] Additionally, operations implemented by the RMCC or the S-RMCC can be broadly applied to manage and control various types of energy storage systems. An RMCC or S-RMCC can oversee an operation of multiple hydrogen tanks, enhancing an efficiency of systems with multiple hydrogen tanks. The RMCC or S-RMCC can manage functionality of multiple fuel cells, ensuring optimal operation. In data centers, the operations can enhance efficiency of multiple Uninterrupted Power Supply (UPS) systems, where reliability can be important, or manage backup battery systems more effectively, ensuring continuous operation and energy efficiency.

[0037] The RMCC or S-RMCC can be implemented in electric vehicle (EV) smart charging stations that are powered by DC microgrids, incorporating PV systems and BESSs. In such a setup, an energy storage system can be charged via a PV system, which can discharge to provide power to charge EVs. An operation of the RMCC can reduce charging and discharging cycles of connected battery units. An alpha mode operation of the RMCC or S-RMCC can be deployed in an EV smart charging station environment, for example when multiple EVs are connected. Under the alpha mode, a charging protocol for each EV can be dynamically adjusted based on individual states of charge. EVs with lower states of charge can be assigned a priority and can be charged at higher rates that EVs with higher states of charge.

[0038] In the following description, various embodiments will be described. For purposes of explanation, specific configurations and details are set forth in order to provide a thorough understanding of the embodiments. However, it will also be apparent to one skilled in the art that the embodiments may be practiced without the specific details. Furthermore, well-known features may be omitted or simplified in order not to obscure the embodiment being described.

[0039] The following description will include the following sections: Section I provides an overview and description of system modeling. Section II describes an RMCC-based power allocation model. A framework and concept for a S-RMCC is presented in Section III. A battery degradation assessment is presented in Section IV. An evaluation of the RMCC and S-RMCC is performed in Section V.

[0040] Section I. Modeling of PV power smoothing

[0041] FIG. 1 is a schematic of PV power smoothing with a power sharing model and parallel operation of battery storage units according to some aspects of the present application. An RMCC system can include a BESS that includes many units connected to a utility-scale PV power plant. FIG. 1 shows a BESS made up of two packs or units, a first unit is charging, and a second unit is discharging. Other examples can include any number of packs or units. The BESS units can be connected to a common DC bus via a bidirectional DC / DC converter. A role of the BESS units can be to provide a compensating power profile to reduce PV power fluctuations based on commands of a power-sharing controller (e.g., the RMCC). A filtering controller can generate a reference signal, known as a target output, to the BESS system, then a state machine-based power-sharing model can split power into charging and discharging power profiles and can monitor a charging status of each unit. The RMCC can serve as a supervisory controller, keeping track of a charge level of each unit and managing an operation of the BESS.

[0042] Solar irradiance data can be collected from a network database (e.g., a Brazilian Environmental Data Organization System (SONDA) network database) to simulate a generated PV power and evaluate a smoothing efficacy of the RMCC and S-RMCC controllers and investigate battery degradation. Data can be measured at a 1-minute time resolution and recorded over 24 hours. Measured irradiance data can be imported to the PV model to simulate a generated photovoltaic power profile that can be smoothed with assistance from parallel-connected battery storage units. Modeling of PV power variability can be caused by variations in an average incident irradiance at a PV power plant and by weather variations. Simulating PV output variability can differ depending on input data, which can be satellite-driven irradiance, ground-based irradiance, or proxy plants. An irradiance point sensor time series can be used as input data and a transfer function-based model (low pass filter) can model the PV output variability. A smoothing performance of the RMCC and S-RMCC can be studied using a simple PV model, in which a PV system is modeled as a transfer function that varies peak output of a photovoltaic system in response to solar irradiance. The PV model can be assumed to smooth out high-frequency fluctuations, and a produced PV power output can be proportional to land area of the installed PV panels. (1)where and are the instantaneous and rated power of the PV system, respectively, is solar irradiation, is land area of installed PV arrays in hectares, and is a curve fitting constant that can depend on a PV configuration, such as a tilt angle for a PV plant and a ground coverage ratio.

[0043] Variability of solar irradiance can differ from place to place and can vary by geographic location, climate, and seasons. FIGs. 2A and 2B are exemplary graphs of solar irradiance and variability index throughout a year according to some aspects of the present application. A contour in FIG. 2A can highlight an irradiance pattern of a selected site throughout the year. One interpretation of data in FIG. 2A is that the selected site can be characterized by higher solar irradiance values during a first and last four months of the year. FIG. 2B shows the variability index of the selected site. Variability can be high during the first and last months. A variability index can be a metric to quantify variability of solar irradiance for a specific location based on geometrical inputs (e.g., latitude and longitude). Maintaining higher irradiance values and variability index can characterize the selected site by cloudy days. Such a location can be ideal for studying an effect of smoothing and an effect on battery aging.

[0044] To mitigate the effects of RE intermittency, RR limits can be incorporated into power grid codes of utilities around the world. Some RR limits can defined (e.g., by a Puerto Rico Electric Power Authority (PREPA) code as a maximum of 10% percent per minute). As an example, if a weekly RRC percentage is less than 98.5%, a PV power plant can be considered non-compliant, and a penalty in a form of a power curtailment can be imposed on the power plant. Non-compliant instances can accumulate to determine a weekly RRC. A non-compliance index , can equal a number of instances that the PV plant violates a ramp rate limit (), relative to a total number of instances ().  (2) (3)

[0045] A BESS system can be modeled as a generic battery model that accurately simulates dynamics of a battery in terms of battery current and SOC. A terminal voltage of the battery ( can depend on a BESS discharge current (), and an open-circuit voltage which in turn can depend on a SOC level during charging and discharging processes, as shown by equations (4) and (5), respectively.  (4) (5)where is a rated battery capacity, is an actual battery charge, is a polarization resistance, is an exponential zone amplitude, is an exponential zone time constant, and is a low-pass filtered battery current. A main advantage of this model can be an ease with which battery parameters can be obtained, as only three points from battery discharge characteristics plot can be important: 1) a fully charged voltage 2) an end of an exponential zone3) a nominal zone ). Then, parameter can be approximated as (), and the parameter B can be approximated as . A battery block from SimPowerSystems / MATLAB can be used to facilitate calculations of the battery parameters. The state of charge of the battery system can depends on , and an initial value at , which can be expressed by: (6) 

[0046] Section II. Power allocation model for a monotonic charging scheme

[0047] Smoothing filters and controllers can be coupled with a PV-BESS system to generate a reference signal that drives BESS operation to smooth out PV output power. The reference signal can be created by subtracting a filter output from generated PV power . A smoothing performance of an RMCC versus an independent monotonic charging controller can be demonstrated by using a first-order low-pass filter (LPF). Using the independent charging controller with LPF smoothing may magnify an issue of the LPF (e.g., power lagging) if both units operate in a same mode. Output power of the LPF filter can be expressed with eqn. (7), and a reference power can be expressed as shown in eqn. (8): (7) (8)where is a smoothing time constant. During a ramp-up interval, can be negative which means the battery will be charged, and positive during a ramp-down period which means the BESS can discharge power stored to compensate for the fluctuation. A negative power reference can charge the battery and a positive power reference can discharge the battery. Usually, the BESS system can be in charging or discharging mode based on a following condition: (9) (10)Two saturation limits can be used to separate into and . A first saturation limit can be adjusted from 0 to +∞, and a second saturation limit can be adjusted from -∞ to 0, as shown in FIG. 3. FIG. 3 is a schematic of a power smoothing and a RMCC-based power allocation model according to some aspects of the present application. A charge level, known as state of charge, can be a ratio of the battery charge to a full battery capacity , which can be defined by equation (11). Assuming that there is a BESS composed of a unit A and a unit B, charge levels of these units can be denoted and . To prolong a life of the BESS, an operation of each unit can be we bounded to operate between an allowable maximum and a minimum state of charge, as described in equation (12).  (11) (12)Battery charge or consumed battery energy can be an integral of and by assuming a lossless conversion stage, , the battery charge can be expressed as in equation (13): (13)An amount of energy consumed or stored in the BESS can be proportional to , implying that battery capacity can be a function of a filtering time constant as well as PV system rating. However, charging and discharging operations of the BESS can be bound between and which means a marginal energy capacity exists that may not be utilized from the battery system and can be subtracted. A marginal energy capacity can be equivalent to battery charge subtracted from full battery capacity and can be expressed as in equation (14): (14) can be defined as marginal battery power and can be described by equation (15) below. Actual realized battery power can be defined by equation (16). (15) (16)Smoothed power at a power grid can be a sum of photovoltaic power with the battery power , as expressed in (17). Where, and are outputs of unit A and unit B, respectively. (17)A power-sharing model can be developed to control states of the units (e.g., charging / discharging modes), and can cause switching from one state to another if a level of charge of a specific unit reaches minimum or maximum limits. A state machine toolbox from MATLAB can be used to model an event-driven controller and to model an RMCC-based power allocation model. For example, if unit (A) is in charging mode () and hits a maximum capacity (), the controller can change a state of unit (A) to discharging (), as well as changing a state of unit B. FIG. 4 is a flowchart of a charging / discharging operation performed by a RMCC according to some aspects of the present application.

[0048] Section III. Structure of S-RMCC

[0049] The S-RMCC controller can lengthen a battery cycle and, as a result, can reduce a number of battery cycles of a BESS made up of more than two units. The S-RMCC controller can use a multivariate forecaster model for short-term irradiance prediction, for example, forecasting an irradiance of an upcoming 30-minute interval based on historical irradiance, weather data, or both. Solar PV power can be obtained based on the forecasted irradiance and passed through a ramp event detector and smoothing filter. A role of the ramp event detector can be to measure a power derivative and determine if there is a ramp event that exceeds an allowable ramp rate limit (e.g., the ramp event detector can include a dead zone feature). The dead zone feature can be important for preventing a battery system from tracking small perturbations. Also, the ramp event detector can determine a type of ramp event (e.g., down / up) by decomposing a power difference into two components (e.g., charging and discharging reference power profiles). Both the forecasted power during a ramp-up event and the forecasted power during a ramp-down event can be mathematically integrated to compute an energy to be compensated by the battery units in charging mode and discharging mode, respectively.

[0050] In such an analysis, the BESS can be assumed to be made up of four units divided into two groups, a charging level of battery units can be, for example, (𝑆𝑂𝐶𝐴 = 20%, 𝑆𝑂𝐶𝐶 = 40%) in charging mode, and (𝑆𝑂𝐶𝐵 = 60%, 𝑆𝑂𝐶𝐷 = 50%) in discharging mode. The controller can include an operation mode selector that compares forecasted energy to a remaining energy of each unit, and accordingly, can determine a number of units to activate in each group, for instance, determining to activate unit A, unit C, or both units. The model can first identify a unit with a highest amount of remaining energy (REM). If a forecasted charging energy (FCE) is higher than the unit with the highest REM value, both units can operate. Otherwise, the unit with the higher REM value can start charging and a second unit can be in a standby mode. An expression for determining the FCE is given by: (18)and a forecasted discharging energy (FDE) is given by: (19)A remaining energy for battery units in charging mode () can be calculated with: (20)and a remaining energy for battery units in discharging mode () can be calculated using: (21)

[0051] The S-RMCC model can include a power allocation model that allocates power to battery units based on SOC information of each group and an activation signal (U) that can be received from an operation mode selector as shown in FIG. 5. FIG. 5 is a schematic of a framework of a smart monotonic charging controller according to some aspects of the present application.

[0052] FIG. 6 is a flow chart for a power allocation model of a S-RMCC according to some aspects of the present application. The controller can activate one or two units based on a forecasted ramp event. Every 30 minutes, the controller can update a number of battery units involved in charging and discharging based on forecasted energy and a REM value of each unit to reach 𝑆𝑂𝐶𝑚𝑖𝑛 and 𝑆𝑂𝐶𝑚𝑎𝑥. A 30-minute time scale can be selected to avoid forecasting errors when a longer forecasting horizon is used. Also, in a case of MA or LPF filters, the selected time scale can be greater than a smoothing time constant used to ensure that all power ramps are detected. If a single unit can handle a next ramp event, then the controller can select the unit with higher remaining energy, and the other unit can be in idle mode. The idle unit can be activated if the remaining energy of the active unit drops below that of the idle unit or reaches operation boundaries. Otherwise, the controller can activate both units to operate during the next ramp event, as shown in FIG. 7. FIG. 7 is a flow chart of an operation mode selector of a S-RMCC according to some aspects of the present application. The controller can switch to an alpha mode if charge levels of the battery units differ. The alpha mode can assist the units in the group to participate in PV power smoothing based on remaining energy. For example, the controller can activate two units in charging mode. A first unit in charging mode can be unit A with a state of charge of 30% and a second unit in charging mode can be unit C with a state of charge of 60%. A participation ratio (𝛼) can be calculated based on a remaining energy level (REL) of BESS units and can be defined as a ratio of a battery unit with a lower remaining energy to the unit with a higher remaining energy and can take on values between 0 and 1. The participation ratio for the example can be calculated as: (22)In the example, based on the participation ration, a power allocation model can share power between unit A and unit C as follows: (23) (24)

[0053] Alpha mode can be achieved by enabling fast charging for a unit with a higher remaining energy to reach a (e.g., unit C from the example) and low charging for a unit with a lower remaining energy to reach 𝑆𝑂𝐶𝑚𝑎𝑥 (e.g., unit A of the example). However, when an REL value of both units are equal, a different mode, referred to as half mode, can be activated to prevent oscillating between the two alpha values when is lower than . FIGs. 8A, 8B, 8C, and 8D are exemplary graphs of state of charge and power for battery units in an alpha mode of a S-RMCC according to some aspects of the present application. FIGs. 8A-8D show an example where two units can operate in alpha mode based on REL values.

[0054] Long Short-Term Memory (LSTM) networks can be used in time series forecasting due to an ability to learn long-term relationships between time steps of data. A single-layer LSTM network, known as vanilla LSTM, can exhibit improvements compared to a standard Recurrent Neural Network (RNN). An LSTM model can include three control gates: a forget gate , an input gate , and an output gate . The output gate can control an operation of a memory cell and can initiate addition of data to or deletion of data from the memory cell. For each memory block, a cell state can be created that contains information learned from previous steps. A memory block can be made up of several memory cells. At each time step , a feature vector can be specified as , and the three gates can be updated using a sigmoid activation function ) and (Hadamard product) for element-wise multiplication. A hidden or output state , can be calculated using and ). A hyperbolic tangent activation function can be used to calculate a cell candidate . LSTM learning objects can include an input weight matrix concatenated as , a corresponding recurrent weight matrix concatenated as , and a bias matrix concatenated as . At each time step , the memory block can be updated as follows: (25) (26) (27) (28) (29) (30)

[0055] FIGs. 9A and 9B are exemplary schematics of a long short-term memory (LSTM) memory cell and stacked LSTM memory architecture according to some aspects of the present application. A stacked LSTM network can enable a reliable method for solving complicated sequence prediction problems. Stacked LSTM networks can include a deep neural network structure with multiple hidden layers. Stacked LSTM networks can efficiently extract features from time series and fully exploit a relationship between a previous state and a current state among sequences. Stacked LSTM networks can include multiple LSTM layers. Each LSTM layer can include multiple connected LSTM cells, as illustrated in FIG. 9B. Raw data can be an input of LSTM layer-1, and a hidden state of LSTM layer-1 can represent an input of subsequent LSTM layers. Section IV. Battery Degradation Assessment

[0056] A multivariate forecasting model can be constructed using historical data collected from a SONDA network database. The historical data can include time (e.g., day, hour, and minute) as well as weather data (e.g., irradiance, pressure, wind speed, rain, humidity, and lux). Constructing the multivariate forecasting model from historical data can involve multiple steps. In a first step, data processing can eliminate missing data. A second step can involve data segmentation, with training data accounting for 80% of a total and testing data accounting for 20%. Then, data standardization can be performed to prevent a learning process from divergence and to obtain improved fitting performance. The multivariate forecasting model can be trained using data from a preceding four months, and the forecasting model can be trained to forecast a future 30 minutes of irradiance based on a previous 30 minutes of irradiance measurement and meteorological conditions. A stacked LSTM network can include two layers each with one hundred units. After data training, the data can be unstandardized and transformed back into original dimensions.

[0057] Stress factors for aging of battery units can include a number of battery cycles, SOC, depth of discharge (DOD), time, and temperature. The number of battery cycles and the DOD can have a significant impact on cycle aging of batteries. Thus, to extend battery life expectancy, a shallow depth cycle can be maintained. A battery degradation model due to cycling can help evaluate an impact of long-term power smoothing on BESS. A methodology used to calculate battery degradation loss due to cycling can be summarized as follows. A first step can involve calculating a SOC for the battery due to smoothing out power fluctuations. A second step of a battery aging model can involve providing the SOC for the battery to a rain-flow counting algorithm (RFC), which can calculate a number of cycles related to each DOD. The RFC algorithm can split complicated charge / discharge cycles into sub-cycles, which can be important for addressing a cycling impact caused by small changes in a battery discharge level. A cumulative ratio of number of cycles at each DOD level to a maximum number of cycles can be used to calculate a battery degradation using a Palmgren-Miner (PM) rule. The PM rule can be expressed as: (31)where denotes battery degradation expressed as a percentage, and is a cycle number related to discharge level. A maximum number of charging and discharging cycles that a battery can withstand before reaching an end-of-life (EOL) is denoted by , and a relationship with for a battery can be found from versus curves provided by a manufacturer, as shown in FIG. 10. FIG. 10 is an exemplary graph of cycle number versus depth of discharge for a battery according to some aspects of the present application. Batteries can have a limited amount of energy that can be used before reaching an EOL limit, at which point the batteries are considered retired or dead. The manufacturer can define the EOL of a battery as 70% of an initial capacity. Additionally, a SAFT-approximated formula for Intensium Max High Power VL30P batteries can be expressed as: (32)

[0058] Section IV. Examples

[0059] This section shows simulated results that highlight a capability of RMCC and S-RMCC controllers to smooth PV variability while maintaining a BESS lifetime and compares simulated results to data for conventional charging controllers.

[0060] FIGs. 11A, 11B, and 11C are exemplary graphs illustrating smoothing characteristics of a RMCC according to some aspects of the present application. Results presented in FIGs. 11A-11C display smoothing performance on an average cloudy day and compare grid power (e.g., smoothed power profile) to original unsmoothed PV power. In this study, the smoothing performance is evaluated using a 1st-order LPF with a time constant of ten minutes. The ten-minute time constant may be a cause for lagging behavior of a smoothed power profile as seen in FIG. 11A. A state of charge of battery units can be seen in FIG. 11B. At a start of the average cloudy day, charge levels of unit A operating on charging mode and unit B operating on discharging mode can be 50% and 60%, respectively. Whereas maximum and minimum allowable charge levels can be set at 70% and 30%, respectively. When unit A reaches its maximum charging capacity at a 600th min, unit A can switch to a discharging state, while unit B can switch to a charging mode. A summary of power-sharing between the battery units is highlighted in FIG. 11C. Shared power by unit A is negative and shared power by unit B is positive since unit A is in charging mode and unit B is working in discharging mode.

[0061] A power-sharing model can change a status of shared power between connected units when unit A reaches the maximum limit and can provide instantaneous switching between the battery units without affecting the smoothing performance. FIG. 12 is an exemplary graph depicting a time dependence of variability reduction using a RMCC according to some aspects of the present application. In FIG. 12, results from the smoothing performance are compared to results from a system without smoothing.

[0062] FIGs. 13A, 13B, and 13C are exemplary graphs illustrating a performance of an independent monotonic charging controller according to some aspects of the present application. Monotonic charging and discharging using conventional independent control when a connected BESS includes battery units with different energy capacities can include limitations. In an illustrative example, a capacity of unit B can be twice the capacity of unit A. At a beginning of a day, charge levels of unit A and unit B can be 50% and 60% respectively. As depicted in FIG. 13A, smoothing performance can be negatively affected around a middle of the day when unit A reaches a maximum state of charge and changes to a discharge state while unit B is also discharging. Unit A can exhibit fast dynamics (e.g., fast charging) and can charge from 50% to 70%, whereas unit B can discharge from 60% to >55%, owing to a small energy capacity, as shown in FIG. 13B. Both units can be in discharge mode, which means the two units may only provide compensation during ramp-down events and not during ramp-up events. In addition, because the two units discharge at an end of the day, smoothed power can experience additional lagging behind an original PV power. Power shared by BESS units is shown in FIG. 13C, when unit A reaches the maximum limit, unit A starts to discharge with unit B. But the shared power can be lower. Reduced shared power can be due to a size of unit A. FIG. 14 is an exemplary graph depicting a time dependence of variability reduction using an independent monotonic charging controller according to some aspects of the present application. A change of power can increase when unit A gets bounded by a maximum limit as illustrated in FIG. 14. This increased change in power can illustrate a drawback of using independent control with a BESS that includes units of different capacities.

[0063] FIGs. 15A, 15B, and 15C are exemplary graphs illustrating smoothing characteristics of a RMCC with battery units of different capacities according to some aspects of the present application. An RMCC controller can be effective in providing monotonic charging and discharging for a battery system with different capacities while maintaining smoothing efficiency. In an illustrative example, reference charge levels of units A and unit B can be 50% and 60%, respectively. A capacity of unit B can be equivalent to double a capacity of unit A. Compared with the independent control, A state of unit B can be changed to charging when unit A reaches the maximum limit, resulting in a good smoothing performance compared to data from a system using conventional independent control, as shown in Fig. 15A. Charge levels of the two units are depicted in FIG. 15B. The charge levels of units B and A can be dependent on one another. A shared power profile by BESS units is shown in FIG. 15C. The BESS units can provide compensation power during both ramp-up and ramp-down events as opposed to a system using conventional independent control that may only provide compensation power during ramp-down events. A state machines-based power allocation model can provide mutually exclusive power-sharing, and the battery units may not operate in a same state at the same time. FIG. 16 is an exemplary graph depicting a time dependence of variability reduction using a RMCC with battery units of different capacities according to some aspects of the present application. A change of photovoltaic power can be greatly reduced with the RMCC controller compared to a conventional independent control approach as shown in FIG. 16.

[0064] An S-RMCC may effectively balance power smoothing and battery aging. The S-RMCC can control units to eliminate PV intermittency and can smooth with a subset of units to extend a BESS lifetime. The S-RMCC can forecast irradiance for a predetermined time interval (e.g., an upcoming thirty minutes) and passes the forecast to a PV model. A PV output power can be fed into a filtering controller to generate a reference signal to control a BESS operation.

[0065] A forecasting horizon can be set to thirty minutes to strike a balance between avoiding frequent switching between battery units that can be associated with shorter forecasting horizons and avoiding poor forecasting performance that can be associated with longer forecasting horizons. Furthermore, as the forecasting horizon is reduced, forecasting error can decrease. But the forecasting horizon may be pre-selected as longer than a filtering time constant because a filter can impose a delay. If the forecasting horizon is less than the filtering time constant, performance of a ramp event detector can be affected. An RR controller can reduce variability of a PV power plant, demonstrating an ability of a framework to adapt to different smoothing controllers. FIGs. 17A, 17B, 17C, and 17D are exemplary graphs of energy profiles and operation mode selections for battery units operating under a S-RMCC scheme according to some aspects of the present application. Forecasted irradiance shown in FIG. 17A can demonstrate a capability of a stacked LSTM to track rapid dynamics of irradiance intermittency. Forecasting accuracy can affect smoothing performance and battery operation. In an event of an LSTM forecasting error, predicted irradiance can be overestimated or underestimated. Overestimation of irradiance, for example, can indicate that a predicted charging energy may be greater than an actual charging energy. As a result, the controller may activate two units instead of one (e.g., more battery operation). In contrast, forecasting underestimation can imply that the predicted charging energy is less than the actual charging energy. As a result, the controller may only activate one unit instead of two, affecting smoothing performance.

[0066] A created reference signal can be partitioned into two components, a forecasted and a forecasted as illustrated in FIG. 17B. These two components can be mathematically integrated to yield FCE and FDE profiles, respectively. In an illustrative example, unit A and unit C can be in charging mode, and unit B and unit D can be in discharging mode. A remaining charging energy of units A and C can be compared to an FCE value every thirty minutes. Because can be greater than , (e.g., 8.4 kWh compared to 4. 2 kWh), unit A can be activated in a 200th minute and can start charging to compensate for a ramp-up event. Due to the FCE value being larger than and , both units can begin operating in an alpha mode at about a 400th minute of the day. In contrast, discharging battery units can be compared to a FDE value. Because unit B can have a greater capacity than unit D, (e.g., 12.6 kWh vs. 6.3 kWh), unit B can be enabled at a 300th minute and can begin discharging to compensate for a ramp-down event. FIGs. 18A, 18B, 18C, 18D, 18E, 18F, 18G, and 18H are exemplary graphs of charge level and power sharing for battery units operating under a S-RMCC scheme according to some aspects of the present application. Power-sharing of battery units and SOC values shown in Fig. 18 can emphasize that the S-RMCC can strikes a balance between PV power smoothing and reducing overwork for of batteries. Unit C and unit D can be deactivated at a start of the day because capacity of other units can be sufficient to handle an upcoming ramp event, thereby slowing battery aging. Unit A can start charging individually as shown in FIG. 18A and FIG. 18B and can change from 50% capacity to 60% at the 400th minute. Whereas, unit C can be idle for the first four hundred minutes as shown in FIG. 18C, and power-sharing for unit C can be zero during this interval, as shown in FIG. 18D. Capacity for unit B can decrease from 60% to 35% during discharging, as shown in FIG. 18E, whereas unit D can be idle for a first six hundred minutes of the day, as shown in FIG. 18G. FIG. 18F and FIG. 18H depict a power-sharing of units B and D, demonstrating that unit B can provide more discharging power to compensate for smoothing. FIG. 19 is an exemplary graph depicting a time dependence of variability reduction using a S-RMCC according to some aspects of the present application. The S-RMCC coupled with the RR controller can exhibit a good smoothing profile as compared to a data from a system with unsmoothed PV power in FIG. 19.

[0067] Battery degradation due to cycling can be assessed when a single BESS is used for smoothing and when a set of units can operate in a monotonic charging scheme. Battery degradation caused by a BESS composed of two units operating in an RMCC mode can be compared to battery degradation caused by a single BESS. LPF-based smoothing with a filtering time constant of ten minutes can be used for both structures. An annual effective charge level of battery units derived from a RMCC framework and that resulted from smoothing with a single battery unit can be imported into RFC to determine a number of cycles associated with each DOD and to compute a battery degradation rate. In an illustrative example, minimum and maximum allowable charging limits are can be 30% and 70%, respectively, and capacity of a single battery unit can be twice that of a single RMCC unit. FIGs. 20A, 20B, and 20C are exemplary histograms illustrating frequency distributions of cycle numbers at various discharge levels for battery units according to some aspects of the present application. FIGs. 20A-20C show histogram plots using a RFC algorithm which illustrates NC vs DOD for unit A (FIG. 20A), for unit B (FIG. 20B), as well as for a conventional smoothing approach (e.g., a single BESS), which can start with a 30% DOD (FIG. 20C).

[0068] As can be seen in FIG. 20A and FIG. 20B, units A and B can have a large stress range of DOD, for example, between 35% and 65%. The large stress range can lead to a significant reduction in a number of cycles compared to a conventional smoothing approach. The conventional smoothing approach can maintain a narrow stress range of DOD that is bounded from 23% to 30%. Accordingly, a number of cycles at 30% DOD can be 2500. The 2500 cycles can occur due to a nature of a low pass filter returning to a reference charge level, as presented in FIG. 20C. A maximum number of cycles for both unit A and unit B can be sixteen. The battery degradation rate can be calculated from a PM rule using equations (31) and (32). A percentage of battery degradation cane be extremely low in unit A and unit B, such as 0.1877% and 0.1918%, respectively. Battery degradation can be 9.1514% in a case of the conventional smoothing approach. Obtained results can clarify that using a reliant monotonic charging scheme can extend battery life with associated lower cycle losses.

[0069] FIGs. 21A, 21B, 21C, and 21D are exemplary graphs of cycle counts for battery units operating under a S-RMCC scheme according to some aspects of the present application. FIGs. 22A and 22B are exemplary graphs of cycle counts for battery units operating under a RMCC scheme according to some aspects of the present application. As depicted in FIG. 21 and FIG. 22, battery degradation for a system with a S-RMCC can be compared to battery degradation for a system with a RMCC using a LPF with ten minutes. The RMCC can control an operation of two units, whereas the S-RMCC can control an operation of four units. A capacity of a single RMCC unit can be double that of a single S-RMCC unit. Because the S-RMCC can involve irradiance data from multiple years to have a complete year of predicted irradiance, only two months were used to examine the comparison. Simulated battery degradation rates for the S-RMCC units were found to be 0.0495%, 0.05088%, 0.0353%, and 0.0370%, whereas the simulated battery degradation rates were found to be 0.0622% and 0.0413% for battery units operating in the RMCC.

[0070] A comparative analysis can investigate an effect of charging limits (e.g., and ) on battery degradation of battery units operating in an RMCC mode. The comparative analysis can be carried out over a year, and different charging limits with different filtering times can be evaluated. Operational limits for the comparative analysis were selected as 90%-10%, 85%-15%, 80%-20%, 75%-25%, and 70%-30%. In the comparative analysis, three different values for filter time constant were used: fifteen, ten, and five minutes. The comparative analysis focused on relating controller performance with different charging limits to improvements in battery life expectancy, rather than effective operation boundaries for the battery. Manufacturers can provide information on maximum and minimum permissible operating limits, which may vary depending on battery size and chemistry (e.g., battery type).

[0071] As shown in Table I, a larger operation range (e.g., a difference between maximum and minimum charging limits) can reduce battery degradation. The reduction can be due to an increase in cycle length, reducing a number of cycles and thus battery degradation. Also, increasing time constant can improve a power smoothing performance, resulting in longer battery operation, which can translate to an increase in battery degradation rate. A slight difference can occur in degradation rate among batteries that operate in RMCC mode. A charging energy may not equal a discharging energy that is used to compensate for power fluctuation each day, and a difference between the two energies can be determined by a smoothing filter used. Setting maximum and minimum allowable discharge rates to 90% and 10%, respectively, can reduces the battery degradation rate. Many cell chemistries, however, may not withstand deep discharge and may be permanently damaged if fully discharged. RMCC Battery UnitsCharging LimitsBattery Degradation () (%) (%) min. min. min.Unit A90100.12500.08840.0766Unit B0.13950.09270.1042Unit A85150.09520.12940.0809Unit B0.14120.11750.1283Unit A80200.16980.15090.1463Unit B0.19540.14970.1285Unit A75250.18410.16390.1693Unit B0.19280.15300.1442Unit A70300.20580.18770.1604Unit B0.20480.19180.1513Table 1. Effect of charging limits on battery degradation for a system with a RMCC.

[0072] An efficiency of a monotonic charging scheme in complying with the ramp rate complaint test can be analyzed. A weekly compliance rate (RRC) should aim to be greater than 98.5%. Otherwise, a PV plant can be penalized a following week through power curtailment (e.g., the PV plant may not work on maximum operation). A performance of the RMCC controller can be compared to that of a PV plant without smoothing and to that of an independent controller. The RMCC controller and the independent controller can manage an operation of a BESS composed of two battery units. A performance of both control schemes can be tested over a year of continuous operation. Minimum and maximum charging limits for the analysis are set at 30% and 70%, respectively. Power smoothing can be performed using a low pass filter with a time constant of 15 minutes, and a daily ramp rate can be calculated to determine events in which a ramp rate of the PV plants exceeds a 10% / min limit.

[0073] Noncompliant events can be aggregated weekly to produce RCC values. FIG. 23 is an exemplary graph of ramp rate compliance evaluation of a RMCC compared to an independent monotonic charging controller according to some aspects of the present application. A plot in FIG. 23 depicts weekly RRC percentages, highlighting that a RMCC scheme can yield a good smoothing performance with RCC values close to 100%. Independent control-based smoothing can violate a weekly RCC threshold in different weeks. Results suggest that a PV plant can be penalized for each of the different weeks if the PV plant is operated using independent control. Based on the results, a performance of the RMCC scheme can surpass the independent control. Independent controlled-battery units may operate concurrently in a single operation mode (e.g., charging or discharging simultaneously), which can affect smoothing performance. A performance of the PV plant without smoothing can frequently exceeds the RCC threshold, owing to a high variability index of data used during first and last four months of the year.

[0074] FIG. 24 is a flow chart of a process 2400 for controlling battery units of a BESS according to certain aspects of the present disclosure. Operations of processes may be performed by software, firmware, hardware, or a combination thereof. Other examples can involve more operations, fewer operations, different operations, or a different order of operations than shown in FIG. 24. The operations of the process 2400 can begin at block 2410.

[0075] At block 2410, the process 2400 involves receiving state of charge (SOC) estimates associated with first battery units of first battery unit group and with a second battery units of a second battery unit group in the BESS. A group of battery units can include battery units combined in a single battery package or battery units from different battery packages in the BESS. The first group can include any number of battery units including a single battery unit. The second group can include any number of battery units including a single battery unit. Each of the battery units can have an equivalent capacity or the capacities can be different. The SOC estimates for both the first and second groups can be determined using algorithms that can rely on a current sensor of a battery management system (BMS). The SOC estimates can be received from the BMS by a controller. The controller can be a RMCC or an S-RMCC.

[0076] At block 2420, the process 2400 involves comparing at least one of the SOC estimates with a charging limit. For example, the comparison can involve comparing an SOC estimate for a single battery unit to a single unit charging limit. Conversely, the comparison can involve comparing a SOC estimate associated with multiple batteries to a combined charging limit. As an example, the at least one SOC estimate can be an estimated SOC for a single battery unit in the first group and the estimated SOC can be compared to a maximum state of charge or a minimum state of charge for a single battery unit. Alternatively, or additionally, the at least one SOC estimate can be a combined SOC estimate for all battery units in a group and the combined SOC estimate can be compared to a maximum total state of charge or a minimum total state of charge. The charging limit can also be based on remaining energy levels.

[0077] At block 2430, the process 2400 involves controlling, based on the comparison between the at least one SOC estimate with the charging limit, a first operation of the first group. The first operation can be charging the battery units in the first group. For example, the comparison can show that each battery unit in the first group has an estimated SOC that is less than a maximum state of charge or a first predetermined threshold value for state of charge. In some examples, for as long as the estimated state of charge of all battery units in the first group is less than the maximum state of charge, the first operation can be a charging operation. In some examples, a status of the first operation can change when a predetermined condition is met. For example, if an estimated state of charge of all of the battery units in the first group is found to meet or exceed the first predetermined threshold value, the first operation can change from a charging operation to a discharging operation, where each of the battery units in the first group is discharged instead of charged.

[0078] While the first operation is a charging operation, each battery unit in the first group can have a charging rate. In some examples, the charging rates of each battery can be the same. Alternatively, or additionally, the charging rates can be based on the state of charge for each battery. As an example, the first group can include two battery units: unit A and unit B. A state of charge for unit A can be 60% and a state of charge for unit B can be 30%. During the first operation, the charging rate for unit B can be double the charging rate of unit A. The charging rates of the battery units can be time dependent as well as state dependent.

[0079] At block 2440, the process 2400 involves controlling, based on the comparison between the at least one SOC estimate with the charging limit, a second operation of the second group. The second operation can be concurrent with the first operation. The second operation can be discharging the battery units in the second group. For example, the comparison can show that each battery unit in the first group has an estimated state of charge that is equivalent to or greater than a minimum state of charge or a second predetermined threshold value for state of charge. In some examples, for as long as the estimated state of charge of all battery units in the second group is greater than the second predetermined threshold value for state of charge, the second operation can be a discharging operation. In some examples, a status of the second operation can change when a predetermined condition is met. The status of the second operation and the status of the first operation can change simultaneously. For example, if an estimated state of charge of any of the battery units in the second group is found to meet or subceed (fall short of) the second predetermined threshold value, the second operation can change from a discharging operation to a charging operation, where each of the battery units in the second group is charged instead of discharged.

[0080] While the second operation is a discharging operation, each battery unit in the second group can have a discharging rate. In some examples, the discharging rates of each battery can be the same. Alternatively, or additionally, the discharging rates can be based on the state of charge for each battery. As an example, the second group can include two battery units: unit A and unit B. A state of charge for unit A can be 60% and a state of charge for unit B can be 30%. During the second operation, the discharging rate for unit A can be double the discharging rate of unit B. The discharging rates of the battery units can be time dependent as well as state dependent.

[0081] The first and second operation can be controlled with a controller associated with the BESS. In some examples, such as when the first group has a single first battery unit and the second group has a single second battery unit, the controller can be an RMCC. The controller can be an S-RMCC, such as when the BESS includes more than two battery units. The S-RMCC can additionally implement an alpha mode. During a charging or discharging process, multiple battery units can reach a first or second threshold by utilizing the alpha mode. Alpha mode can implement an efficient and balanced charging operation across all battery units, ensuring that each battery unit can achieve a full charging potential in a coordinated manner. Through alpha mode, the S-RMCC can dynamically adjust charging rates and thresholds for each unit based on real-time data and predictive analytics, allowing for optimal energy storage and prolonged battery health. Such capabilities can ensure that all battery units simultaneously attain maximum charging capacity, enhancing overall efficiency and effectiveness of a charging process.

[0082] The process 2400 can further involve modeling an irradiance profile associated with the BESS. The irradiance profile can be modeled using an algorithm such as a neural network. Controlling the first operation, the second operation, or both can be based on the modeled irradiance profile. For example, the first predetermined threshold value or the second predetermined threshold value can be based on the irradiance profile.

[0083] The process 2400 can further involve controlling a third operation of a third group of third battery units. The third operation can involve idling the third group of battery units. In some examples, the process 2400 can involve selecting battery units from the first group, the second group, or both to include in the third group. A number of battery units included in the third group can be based on the irradiance profile. The third operation can be concurrent with the first operation, the second operation, or both. The process can further involve estimating a photovoltaic (PV) output based on the irradiance profile and determining a ramp-up event (or a ramp-down event) for the first group or the second group.

[0084] FIG. 25 is a block diagram of a controller 2500 for a BESS according to certain aspects of the present disclosure. The controller 2500 can be a RMCC or a S-RMCC. As shown, the controller 2500 includes a processor 2502 communicatively coupled to memory 2504. The processor 2502 can include one processing device or multiple processing devices. Non-limiting examples of the processor 2502 include a Field-Programmable Gate Array (FPGA), an application specific integrated circuit (ASIC), a microprocessor, or any combination of these. The processor 2502 can execute instructions 2510 stored in the memory 2504 to perform operations, such as the operations of process 2400 from FIG. 24. In some examples, the instructions 2510 can include processor-specific instructions generated by a compiler or an interpreter from code written in any suitable computer-programming language, such as C, C++, C#, Python, or Java.

[0085] The memory 2504 can include one memory device or multiple memory devices. The memory 2504 can be non-volatile and may include any type of memory device that retains stored information when powered off. Non-limiting examples of the memory 2504 include electrically erasable and programmable read-only memory (EEPROM), flash memory, or any other type of non-volatile memory. At least some of the memory 2504 can include a non-transitory computer-readable medium from which the processor 2502 can read instructions 2510. The non-transitory computer-readable medium can include electronic, optical, magnetic, or other storage devices capable of providing the processor 2502 with the instructions 2510 or other program code. Non-limiting examples of the non-transitory computer-readable medium include magnetic disk(s), memory chip(s), RAM, an ASIC, or any other medium from which a computer processor can read instructions 2510.

[0086] The memory 2504 can further include a charging limit 2512, a SOC estimate 2514, an irradiance profile 2516, and an algorithm 2518. The processor 2502 can compare the SOC estimate 2514 to the charging limit 2512 and control operations of the BESS based on the comparison. Additionally, the processor 2502 can model the irradiance profile 2516 using the algorithm 2518. In some examples, the algorithm 2518 can be a neural network. The processor 2502 can control or adjust operations of the BESS based on the irradiance profile 2516. For example, the processor 2502 can select a number of battery unit of the BESS to idle based on the irradiance profile 2516.

[0087] While the present subject matter has been described in detail with respect to specific embodiments thereof, it will be appreciated that those skilled in the art, upon attaining an understanding of the foregoing may readily produce alterations to, variations of, and equivalents to such embodiments. Accordingly, it should be understood that the present disclosure has been presented for purposes of example rather than limitation, and does not preclude inclusion of such modifications, variations, and / or additions to the present subject matter as would be readily apparent to one of ordinary skill in the art. Indeed, the methods and systems described herein may be embodied in a variety of other forms; furthermore, various omissions, substitutions, and changes in the form of the methods and systems described herein may be made without departing from the spirit of the present disclosure. The accompanying claims and their equivalents are intended to cover such forms or modifications as would fall within the scope and spirit of the present disclosure.

[0088] Conditional language used herein, such as, among others, “can,” “could,” “might,” “may,” “e.g.,” and the like, unless specifically stated otherwise, or otherwise understood within the context as used, is generally intended to convey that certain examples include, while other examples do not include, certain features, elements, and / or steps. Thus, such conditional language is not generally intended to imply that features, elements and / or steps are in any way required for one or more examples or that one or more examples necessarily include logic for deciding, with or without author input or prompting, whether these features, elements and / or steps are included or are to be performed in any particular example.

[0089] Disjunctive language such as the phrase “at least one of X, Y, or Z,” unless specifically stated otherwise, is otherwise understood within the context as used in general to present that an item, term, etc., may be either X, Y, or Z, or any combination thereof (e.g., X, Y, and / or Z). Thus, such disjunctive language is not generally intended to, and should not, imply that certain examples require at least one of X, at least one of Y, or at least one of Z to each be present.

[0090] Use herein of the word “or” is intended to cover inclusive and exclusive OR conditions. In other words, A or B or C includes any or all of the following alternative combinations as appropriate for a particular usage: A alone; B alone; C alone; A and B only; A and C only; B and C only; and all three of A and B and C.

[0091] The use of the terms “a” and “an” and “the” and similar referents in the context of describing the disclosed examples (especially in the context of the following claims) are to be construed to cover both the singular and the plural, unless otherwise indicated herein or clearly contradicted by context. The terms “comprising,” “including,” “having,” and the like are synonymous and are used inclusively, in an open-ended fashion, and do not exclude additional elements, features, acts, operations, and so forth. Also, the term “or” is used in its inclusive sense (and not in its exclusive sense) so that when used, for example, to connect a list of elements, the term “or” means one, some, or all of the elements in the list. The use of “adapted to” or “configured to” herein is meant as open and inclusive language that does not foreclose devices adapted to or configured to perform additional tasks or steps. The term “connected” is to be construed as partly or wholly contained within, attached to, or joined together, even if there is something intervening. Recitation of ranges of values herein are merely intended to serve as a shorthand method of referring individually to each separate value falling within the range, unless otherwise indicated herein, and each separate value is incorporated into the specification as if it were individually recited herein. Additionally, the use of “based on” is meant to be open and inclusive, in that a process, step, calculation, or other action “based on” one or more recited conditions or values may, in practice, be based on additional conditions or values beyond those recited. Similarly, the use of “based at least in part on” is meant to be open and inclusive, in that a process, step, calculation, or other action “based at least in part on” one or more recited conditions or values may, in practice, be based on additional conditions or values beyond those recited. Headings, lists, and numbering included herein are for ease of explanation only and are not meant to be limiting.

[0092] The various features and processes described above may be used independently of one another or may be combined in various ways. All possible combinations and sub-combinations are intended to fall within the scope of the present disclosure. In addition, certain method or process blocks may be omitted in some implementations. The methods and processes described herein are also not limited to any particular sequence, and the blocks or states relating thereto can be performed in other sequences that are appropriate. For example, described blocks or states may be performed in an order other than that specifically disclosed, or multiple blocks or states may be combined in a single block or state. The example blocks or states may be performed in serial, in parallel, or in some other manner. Blocks or states may be added to or removed from the disclosed examples. Similarly, the example systems and components described herein may be configured differently than described. For example, elements may be added to, removed from, or rearranged compared to the disclosed examples.

[0093] All references, including publications, patent applications, and patents, cited herein are hereby incorporated by reference to the same extent as if each reference were individually and specifically indicated to be incorporated by reference and were set forth in its entirety herein

Claims

1. A battery energy storage system (BESS) comprising: a first group of first battery units; a second group of second battery units; and a controller configured to:receive state of charge estimates associated with the first group and the second group; andcontrol, based on a comparison of at least one of the state of charge estimates with a charging limit and absent additional sensor data, a first operation of the first group and a second operation of the second group, wherein the first operation comprises charging the first group, and wherein the second operation comprises discharging the second group and is concurrent with the first operation. 2. The BESS of claim 1, wherein the controller is a reliant monotonic charging / discharging controller (RMCC) or a smart-reliant monotonic charging / discharging controller (S-RMCC). 3. The BESS of claim 2, wherein the comparison indicates that the at least one of the state of charge estimates is above the charging limit and wherein the first operation further comprises discharging the first group, and wherein the second operation further comprises charging the second group. 4. The BESS of claim 1, wherein the controller is further configured to model an irradiance profile to associate with the BESS, and wherein the first operation and the second operation are controlled based on the irradiance profile. 5. The BESS of claim 4, further comprising a third group of third battery units, wherein the controller is further configured to control a third operation of the third group concurrent with the first operation and the second operation. 6. The BESS of claim 5, wherein the third operation comprises idling the third group of third battery units. 7. The BESS of claim 6, wherein the controller is further configured to select a number of battery units to include in the third group based on the irradiance profile. 8. A method comprising:receiving state of charge estimates associated with first battery units of a first battery unit group and with second battery units of a second battery unit group units for a battery energy storage system (BESS);comparing at least one of the state of charge estimates with a charging limit;controlling, based on the comparison, a first operation of the first battery unit group, wherein the first operation comprises charging the first battery unit group; andcontrolling, based on the comparison, a second operation of the second battery unit group, wherein the second operation comprises discharging the second battery unit group and is concurrent with the first operation. 9. The method of claim 8, further comprising generating a forecast irradiance of at least the first battery unit group or the second battery unit group over a predefined time interval. 10. The method of claim 9, further comprising:estimating, based on the forecast irradiance, a photovoltaic (PV) output; anddetermining a ramp-up event that should be compensated by at least the first battery unit group or the second battery unit group.  11. The method of claim 10, further comprising calculating a compensation for charging based on the determined ramp-up event. 12. The method of claim 10, further comprising:estimating an amount of energy needed from the BESS during charging and discharging;comparing the amount of energy with a remaining charge of battery units; anddetermining a set of battery units to activate. 13. The method of claim 8, wherein the comparison indicates that the at least one state of charge estimates is above the charging limit and wherein the first operation further comprises discharging the first battery unit group and wherein the second operation further comprises charging the second battery unit group. 14. The method of claim 9, further comprising selecting a group of battery units to idle based on the forecast irradiance. 15. A non-transitory computer-readable medium comprising instructions that, upon execution on a system, cause the system to:receive state of charge estimates associated with a first group of first battery units and a second group of second battery units for a battery energy storage system (BESS);compare at least one of the state of charge estimates with a charging limit;control, based on the comparison, a first operation of the first group, wherein the first operation comprises charging the first group; andcontrol, based on the comparison, a second operation of the second group, wherein the second operation comprises discharging the second group and is concurrent with the first operation. 16. The non-transitory computer-readable medium of claim 15, wherein the comparison indicates that the at least one of the state of charge estimates is above the charging limit, wherein the first operation further comprises discharging the first group, and wherein the second operation further comprises charging the second group. 17. The non-transitory computer-readable medium of claim 15, wherein the execution of the instructions further causes the system to model an irradiance profile to associate with the BESS, and wherein controlling the first operation and controlling the second operation is based on the irradiance profile. 18. The non-transitory computer-readable medium of claim 17, wherein the execution of the instructions further causse the system to control a third operation of a third group of battery units, wherein the third operation is concurrent with the first operation and the second operation. 19. The non-transitory computer-readable medium of claim 18, wherein the third operation comprises idling the third group of battery units. 20. The non-transitory computer-readable medium of claim 19, wherein the execution of the instructions further causes the system to select a number of battery units to include in the third group based on the irradiance profile. 

Images