Background of the invention: A reduction of carbon based energy supply leads to - the far more efficient - electrification of drive trains (e-mobility) and/or auxiliary systems (e.g. aircrafts) and stationary systems. All have in common the need for an efficient and save, high-voltage and high-energy storage system. Batteries, especially Li-Ion batteries, will play an important role because of their high energy and power density.

High voltage and high energy density is achieved by serial connection of the individual voltage sources (e.g. cells) and needs control of the individual cells in order to avoid serious cell damage leading to sub-optimal performance, reduced lifetime, or safety threats: One way is a system where a common "end-of-charge-voltage" of all cells is reached by a resistive discharging element. It is a commonly used technology today, because of its simplicity and low cost.

Another way to reach a common "end-of-charge- and discharge-voltage" is a system that active- ly transfers charge between the cells. This is either done by transferring charge from one cell to the neighboring cell, i.e. to the next cell along the serial line (US005479083), or from one cell or module to the serial string e.g. by flyback-converter (WO2010088944, US201 10012559).

US 2013/0 062 946 A1 describes a hierarchical balancing system that performs balancing amongst power packs comprising an arrangement of power cells, while the power packs separately perform cell-level balancing. A hierarchical balancing technique may be implemented where balancing is being performed at different levels of the hierarchical balancing system. Any number of levels of balancing may be implemented by creating groupings of power packs that can be balanced.

The object of the present invention is to provide an energy transfer arrangement with an improved effectivity at reasonable costs. Summary of the invention:

An energy transfer arrangement according to the invention comprise multiple serially connected voltage sources, a hierarchical structure of energy transfer units having at least two first level energy transfer units and one second level energy transfer unit. Each two or three neighboring voltage sources are connected to one of the first level energy transfer units. Each two or three neighboring first level energy transfer units are connected to the second level energy transfer unit. A battery management system is adapted to control the energy transfer over each of the energy transfer units based on state-of-health parameters for each voltage source.

In a preferred embodiment, the battery management system is adapted to control the energy transfer based on a state-of-charge target value for each voltage source wherein the target value is determined by the state-of-health parameters for each voltage source. The target value may, by the way of an example, depend on state-of-health parameters like the number load cycles performed, an internal resistance and/or a change of internal resistance over time, an (absolute) capacity and/or a change of capacity over time, a voltage source temperature and/or a temperature change over time, a deterioration of electrolyte, and/or a cathode and/or an anode material and/or their respective change over time, or more general: all deterioration parameters of a voltage source. The energy transfer arrangement comprises preferably means for calculating or measuring the state-of-health parameters for each voltage source.

The voltage sources connected to an energy transfer unit of the first level and the energy transfer unit itself build a "virtual" voltage source of the first level for the superior level, i.e. the second level. These virtual voltage sources connected with an energy transfer unit of the second level build a virtual voltage source of the second level and may be connected to an energy transfer unit of the third level, if applicable. Two or three of these virtual second level virtual voltage sources together with the third level energy transfer unit form a third level virtual voltage sources having eight to 27 voltage sources. With this concept of a hierarchical structure, the energy transfer arrangement may be adapted to the needs regarding capacity and terminal voltage of the energy transfer arrangement. The hierarchical structure of the energy transfer arrangement comprises in a preferred embodiment n levels and each two or three neighboring energy transfer units of levels two to n-1 are connected to an energy transfer unit of a superior level. Accordingly, two or three neighboring virtual voltage sources of levels two to n-1 form a virtual voltage source of the superior level. In a preferred embodiment, switches of the energy transfer unit are controlled in such to limit the current across an energy transfer switch in the energy transfer unit to a maximum value. Generally, the current flowing through a switch is decreasing with an increasing level due to increasing voltage of the connected (virtual) voltage sources. When the same amount of energy is trans- ferred, increasing voltages lead to a decreasing current. This has the positive effect that with increasing levels of the switches the design may be adapted to receive the same or a lower current compared the inferior level. Accordingly, the arrangement can be designed such that each level of energy units has an individual maximum value. Preferably, the switches are controlled to generate a fixed frequency pulse-width modulation sinusoidal waveforms running through the connected voltage sources and to use modified frequencies as analysis for electro-impedance-spectroscopy (EIS) data. The energy transfer units may contain an inductive or capacitive energy transfer element. The voltage sources are preferably chemical battery cells, in particular lithium ion cells, and/or configurations of multiple battery cells, in particular serially or parallel connected cells forming a battery pack. Alternatively, the voltage sources are non-chemical storage elements, in particular Peltier elements. In a further alternative, the voltage sources are capacitors or fuel cells photovoltaic cells.

The invention further contains a method of controlling said energy transfer. The method comprises the following steps: In Step 1 (S1) an individual state-of-charge target value for each voltage source based on state-of-health parameters of said voltage source is determined. In step 2 the amount of energy to be transferred via each energy transfer unit so that each of the voltage sources reaches the individual state-of-charge target value and energy losses in the energy transfer units are calculated. In step 3 the remaining time of transfer (t-rransfer, t' Transfer) until each voltage source reaches the state-of-charge target value is calculated. Each energy transfer unit's power capability during the energy transfer (Ρπ-ransfer, P' .Transfer) is calculated in step 4 by the product of integrated voltage over time and integrated current over time. In step 5 the amount of energy to be transferred from/to each individual cell (EiTransfer, EVransfer) via an energy transfer unit is calculated. The individual state-of-charge point (SOC,A, SOC,E) for each voltage source at which the energy transfer among cells must be activated to reach the individual state-of-charge target values at the same time is calculated in step 6. In step 7 the energy transfer units activate the energy transfer for each cell when the individual state-of-charge point (SOC,A, SOC,E) deter- mined in S6 is reached. The energy transfer via an energy transfer unit of the superior level n+1 is preferably activated when the energy transfer of the level n is terminated.

Brief description of the drawings:

The invention will be explained in more detail with reference to embodiments illustrated in the figures, in which: Fig. 1 shows a schematic diagram of exemplarily three levels of energy transfer units according to the present innovation;

Fig. 2 shows a schematic diagram of three possible implementation of energy transfer units in any of the energy transfer levels;

Fig. 3 shows a schematic diagram of a safety circuit to ensure low voltage across any en- ergy transfer unit;

Fig. 4 shows a schematic diagram of two possible implementations of safety circuits; Fig. 5 shows a schematic diagram of the electro-impedance-spectroscopy waveform generator designed with the energy transfer unit;

Fig. 6 shows the necessary energy transfer and respectively current through the energy transfer units of a normal distributed arrangement of battery cells in an energy storage arrangement with the proposed invention;

Fig. 7 shows an efficiency comparison of three different energy transfer arrangements; Fig. 8 shows the profiles of charge voltage, current and state of charge in typical charge and discharge operations,

Fig. 9 shows the ratio of charge voltage to state of charge in a typical cell in an energy storage arrangement,

Fig. 10 shows the region of the individual charge voltage and the corresponding difference in capacity

Fig. 11 shows the profiles of charge voltage, current and state of charge in typical charge and discharge operations.

Detail description of specific embodiments:

Fig. 1 illustrates three levels of energy transfer units in an energy storage system. In the first level a number of n/2 energy transfer units (2) are connected to each of the battery cells (V0 ... Vn). In the second hierarchy n/4 energy transfer units (2) are connected to each of equal stacks of battery cells. In the third level n/8 energy transfer units (2) are connected to each of equal stacks of battery cells.

The energy transfer arrangement allows to transfer energy between battery cell 1 and 2 or bat- tery cell 3 and 4 in the first level, energy transfer between the battery stacks 1 ,2 and 3,4 or battery stacks 5,6 and 7,8 and so on an so forth.

Fig. 2 shows exemplarily schematic implementation of energy transfer units (2) as inductive or capacitive design. Fig. 2a) illustrates an inductive based transfer with a minimum of compo- nents. In order to transfer energy from battery cell n-1 to battery cell n, switch S1 is turned on and energy is stored in the inductor. In the second phase, switch S1 is turned off and switch S2 is turned on to energy battery cell n with the stored energy of the inductor. Fig. 2b) shows an capacitive based transfer unit. Energy is stored in the capacitor when switch S1 and S3 is turned on. In the second phase, switch S1 and S3 is turned off and switch S2 and S4 is turned on to discharge the capacitor to the battery cell n.

Fig. 2c) shows an inductive DC/DC converter based transfer unit. First switch S1 is turned on to apply current to the DC/DC converter and in the second phase the switch S1 is turned off and switch S2 is turned on to convert the energy to battery cell n.

Fig. 3 shows a possible implementation of a safety circuit, which limits the voltage across the switched of each energy transfer unit to a specific voltage, by enabling both switches if a high voltage is detected.

Fig. 4 shows two exemplarily schematic implementations of the safety circuit. Fig. 4a) shows a voltage detection circuit with zener diode, which overrules both control signals of the switches with two diodes. In Fig 4b) the control signal overrule is performed with electronic switches for higher voltage and powerful overrule. Fig. 5 illustrates the sinusoidal current waveform generation with the energy transfer unit.

Fig. 5a) shows the selected energy transfer unit design, which allows to charge and discharge the battery cells (1). Discharge is performed when switch S1 of the energy transfer unit (2) is turned on. In contrast, when the switch S2 is turned on the current is drawn from the upper battery cell and charged to the lower battery cell after the switch S2 is turned off. The switch signals (PWM 1-1 , PWM1-2) are controlled with fixed frequency pulse-width modulation to generate a sinusoidal current waveform. For example a high fixed frequency of the switches T _SW itcn (e.g. 100kHz), results in a pulse-width period T _s t _e of about a tenth (e.g. 10kHz). This further allows to generate a tenth T _s t _e current frequency for both battery cells at the same time. The pulse-width period is expanded to generate lower current frequencies and to measure other electro- impedance-spectroscopy points. Fig. 6 shows the results of a numerical calculation of 10,000 randomly arranged 128 battery cells. Fig. 6a) shows the necessary maximum energy transfer which is needed for the specific levels according to the invention. Fig. 6b) shows the maximum current flow to the respectively energy transfer units. Fig. 7 shows an efficiency comparison of a cell-to-cell energy transfer arrangement, a cell-to- system energy transfer arrangement and the proposed invention.

Fig. 8 shows the profiles of the charge voltage, current and the state of charge SoC. The current flow through the cells connected in series differs merely in a limited region between times t2 and t3 or between and ts, i.e. in phases in which the cells are already partially discharged.

The invention discloses a mode of charge transfer in an energy storage system enabling the control of the state of charge (SoC) of each individual battery cell with minimum transfer losses and significant reduction of manufacturing costs of the hardware at the same time. The invention comprises an arrangement of energy transfer units in multiple levels, whereas higher level energy transfer units operate with higher voltage levels thus decreasing energy losses and consequently increasing efficiency.

The energy storage comprises a plurality of n battery cells connected in series and a number of (n-1) charge transfer units, which couple in multiple levels to the individual battery cell or stack of multiple battery cells in a higher level of the series string.

In a high voltage system of Fig. 1 , the charge transfer arrangement comprises several levels of charge transfer units. In one embodiment the number of (n/2) energy transfer units are coupled to the battery cells working as voltage sources in the first level. Allowing charge to be transferred in a bi-directional way between the first and the second battery cell, the third and fourth battery cell, the fifth and sixth, the seventh and the eighth respectively and so on and so forth. Within the second level a number of (n/4) energy transfer units (bidirectional) are coupled to a stack of two battery cells allowing charge to be transferred between battery cells one and two to battery cell three and four. The following level comprises a number of (n/8) energy transfer units to allow charge to be passed (bidirectional) between the stack of battery cell one to four and five to eight and so on and so forth. In total this results in n-1 transfer units over all levels. The energy transfer unit between two stacks of battery cells can be designed in already known ways, e.g. inductive or capacitive. In a first design shown in Fig. 2a), an inductor is connected on one side to the common connection of the two battery stacks. The inductor is further connected to the common connection of two switches, connected in parallel to the battery stacks. In this design switch S1 and switch S2 are controlled in an alternating non-overlapping way, allowing charge to be transferred. A second design (Fig. 2b)) can be a capacitive one, whereas each battery cell/stack is coupled to two serial connected switches. Between the two common connections of the two switches in parallel to the battery stacks a capacitor is placed to hold the transferred charge. In a third design (Fig. 2c)) each battery stack is coupled to a DC/DC converter to transfer charge between each individual cell and the battery pack or between the battery modules and the battery pack. In regard of energy losses the first design consists of a minimum of switching components and a conductive energy transfer with low switching losses. Each charge transfer unit may be individually controlled, either centralized or distributed by an integrated circuit that allows implementation of energy transfer algorithms for high performance and high energy applications.

In order to avoid excessive stress of the individual voltage sources, the aging process of the voltage source is typically influenced with a reduced usage of capacity of the specific voltage source.

Essential factors influencing the change in the voltage source properties and therefore the aging of the voltage source are as follows: end-of-charge voltage, storage voltage, operating voltage, charge and discharge current intensity, state of charge, level of defined end-of-charge and end- of-discharge voltage, temperature, calendrical age of the cell, number of previous charge and discharge cycles, speed of charge/discharge changeovers and temperature during all quiescent and operating states, i.e. during storage, in the quiescent state, during charging and during discharging. The temporal duration for which one or more of the influencing factors take effect also substantially influences the voltage source properties.

The voltage source specific data is directly measured with already known techniques or gathered from other measurement devices of the battery management system. These measurement data preferably include the capacity, the internal resistance and other state-of- health (SOH) pa- rameters. The battery management system controls the energy transfer for example on basis of individual maximum values for said influencing factors for each cell of the energy transfer units which are determined by the state-of-health parameters. As a simple example the individual end-of-charge voltage CVL for a storage arrangement comprising 5 cells connected in series are determined as followed.

The determined cell capacities are illustrated in Table 1 :

Table 1

The largest capacity CMAX is defined as reference capacity CREF:

Then, the cells are classified and ordered in accordance with their capacities C, or the difference

AC AC = C - C in the individual cell capacity i with respect to the reference capacity ^{1 REF <'} , as illustrated in table 2:

Table 2

The end-of-charge voltage CVL predetermined by the manufacturer is set as maximum end-of- charge voltage CVLMAX.

Conservation of the weaker cells is achieved according to the invention by a cell-individual reduction in the end-of-charge voltage CVL in accordance with The relationship function f _C v_, shown in Fig. 10, is derived in the exemplary embodiment from the open circuit voltage characteristic OCV predetermined by the manufacturer for the cells, as is illustrated in Fig. 9. Fig. 9 shows the ratio of charge voltage to state of charge in a typical cell in an energy storage arrangement, and the individual end of charge voltage levels defined in the section of the open circuit voltage function. Fig. 10 shows the region of the individual charge voltage and the corresponding difference in capacity.

In this case, it is advantageous if the minimum end-of-charge voltage CVLMIN, i.e. the end-of- charge voltage of the weakest cell, is still fixed in the steep end section of the open circuit volt- age characteristic OCV.

The cell-individual end-of-charge voltages CVU of the remaining cells are then distributed between CVLMAX and CVLMIN in accordance with the ordering of the cells using the differences in capacity. This distribution can in the simplest case take place linearly, but also with any other desired form, for example exponentially or logarithmically.

As a variant, the absolute values of the cell-individual end-of-charge voltage CVU can also be fixed in such a way that only a group of the weaker cells, for example half of them, are given an end-of-charge voltage CVU which is reduced individually in accordance with their ordering and all of the other cells are given the end-of-charge voltage CVL predetermined by the manufacturer. Thus, the reduction in capacity as a result of a reduced end-of-charge voltage CVU is less.

Assuming that the end-of-charge voltage CVL predetermined by the manufacturer is 4.2 V and the gradient of the open circuit voltage characteristic OCV reaches a value of 100% at a cell voltage of 4.0 V and the cell-individual end-of-charge voltage CVU is distributed linearly, the cell- individual end-of-charge voltages CVU shown in table 3 result:

Table 3 Once cell-individual end-of-charge voltages CVL, have been defined on the basis of the determined capacities C, of the cells, all of the cells are charged to their individual end-of-charge volt

The following example describes the end-of-charge voltage CVL in a possible second alternative including the state-of-health parameters for a storage arrangement comprising same 5 cells connected in series as of the previous example:

The cells are shown in descending order of their capacities, as illustrated in table 4. In addition, the state-of-health parameters internal resistance and change in internal resistance ARi as well as the resulting state-of-health are documented. The change in internal resistance is defined as newly measured internal resistance compared to the previously measured internal resistance.

Table 4

In the initialization process and in each of the following calibration processes, the capacity and internal resistance is measured. The state-of-health calculates from the capacity and internal resistance compared to the values of the initialization process. The battery cell is defined to be not suitable for EV application if, for example, only 80% of nominal capacity (C80%) is left or the internal resistance has increased by 200% (R2oo%).

SoHi = (Ci— C8O%)/(CINIT— C8o%)*(R2oo% - Ri)/(R2oo% - RINIT)

For further calculations, the difference XDIFFJ of a parameter compared to other cells parameters is calculated and the variation factor kxj of the difference is defined for the cell i.

XDIFF, i = (X - XMIN)

kx,i = XDIFFJ /(XMAX— XMIN) In table 5 the difference of capacity CDIFF, internal resistance RDIFF, state-of-health SOHDIFF , Increase of internal resistance ARDIFF, as well as the variation factors thereof kc , kp> , ksoHand k _A R are shown.

Table 5

The variation factors define the difference between the cell's parameter compared to all other cells parameters. This factor is used to identify the strength or weakness of a cell. For further rating of the parameters, all parameters must represent a good condition of the cell as low factor and a worse condition as a high factor. A high capacity, a low internal resistance, a high state- of-health and a low change of internal resistance define a good condition of a cell. Therefore the variation factor kx of the internal resistance and the change in internal resistance must be inverted in the further calculation of the rating. The rating is performed as weighted rating with individual weight factors of wc, WR,, WSOH, W _A RI. The selection of the weight factor is dependent on the cell chemistry used. An advantageous base-selection is defined as high value for the change in internal resistance as this indicates a rapid aging, i.e.: wc = 0.2, WR, = 0.1 , WSOH = 0.2, W _A RI = 0.5. Finally the individual end-of-charge voltages CVU are calculated as sum of all variation factors kx,, multiplied by the weight factor, as shown in table 6.

XSUM = (Wc · kc) + (W _RI · (1 - k _Ri )) + (WSOH · ksOH) + (WAR! · (1 - k _A Ri))

CVI_i = CVLMAX— ACVL*(XSUM— XSUM,MIN)/(XSUM,MAX - XSUM.MIN)

Table 6

As advantageous version, the end-of-charge voltage is reduced only for cells under a rated val- ue higher than 50%.

CVLi = CVLMAX— ACVL · (XSUM— XSUM.MIN) / (0.5 · (XSUM.MAX - XSUM.MIN)) - 1

In the operating process, as illustrated in Fig. 1 1 , the individual cells are discharged and recharged on the basis of the now individually predetermined end-of-charge voltages. In the ex- ample, the cells have a matching end-of-discharge voltage DVL, but it may also be expedient for the end-of-discharge voltages DVL to be fixed individually for each cell.

Fig. 11 shows the profiles of the charge voltage, current and the state of charge SoC, i.e. in the specified example shown in table 3 the curve profile of the strongest cell CMAX corresponds to cell number 5 with an end-of-charge voltage CVL ₅ of 4.2 V and the curve profile of the weakest cell CMIN corresponds to cell number 4 with an end-of-charge voltage CVL ₄ of 4.0 V.

By virtue of the conservation according to the invention of the weaker cells, said cells age more slowly, the cell properties harmonize more and more, the cell drift is reduced, and there is auto- matic matching of the cells.

As is apparent from Fig. 1 1 , the current flow through the cells connected in series differs merely in a limited region between times t2 and t3 or between t ₄ and ts, i.e. in phases in which the cells are already partially discharged.

In these phases, stronger cells are loaded with increased current flow to a greater extent as well. This is achieved by suitable driving of the energy transfer units in a battery management system. Limiting the interventions of the battery management system to regions in which the cells are already partially discharged has the effect of limiting the losses owing to the energy transfer units in a battery management system since the operating state of the extensive discharge of the cells occurs comparatively seldom since, in particular in the case of use in electric vehicles, the aim is to keep the energy store as charged as possible in order thus to allow a maximum operating range. During charging and discharging the battery management system performs continuous load current measurements which are processed to a load profile. This profile is used for load current prediction during discharge and charge cycles. In addition to measurements, historic data is included in the calculation to achieve a most accurate approximation of the upcoming load current. Today, state-of-the-art prediction algorithm can be coupled to external prediction data from driver information systems, such as route planner and navigation systems.

Energy storage arrangements may be implemented with different designs of energy transfer units for different energy transfer levels (such as inductive, capacitive, DC/DC converter). In a multiple level charge transfer arrangement, the efficiency of each charge transfer unit may be further increased with a higher level, due to an increased voltage and lower influence of the on-resistance of each switch of the energy transfer unit. This enables energy transfer units with higher voltage to operate nearly at 99% efficiency.

The battery cells of an energy storage system are in a further embodiment produced in automatic manufacturing plants, resulting in a normal distributed output of battery cell characteristics. Therefore characteristics as capacity, internal resistance, state of charge voltage and many other occur in the energy storage arrangement in a normal distributed probability. Extreme fluctuations in value characteristics occur very rarely within the arrangement and with very high probability only one or few per arrangement.

In one configuration, the energy transfer units of higher levels are designed to double the energy throughput only every second energy transfer level. Extreme conditions, whereas for example multiple low capacity battery cells and multiple high capacity battery cells would need a poten- tially higher energy throughput for higher level transfer units (double every level) are ne- glectable. In normal distributed systems, such as battery cells used as storage devices, extreme conditions are unlikely and can be specially treated during production or assembly of the energy storage system. In order to allow control, the necessary energy transfer of the proposed invention are calculated in advance (e.g. for capacity). First, the individual capacity differences CapDiff, between the state of charge and the target state of charge must be determined. The target value for each voltage source is determined based on state-of-health parameters of the voltage source. This requires further a determination of the individual battery cell capacities Cap,, the state of charge SOCi and the targeted state of charge target SOG which should be reached after the charge transfer is completed.

CapDiffi = (targetSOCi * Capi) - (SOCi * Capi)

Each level has a specific number u energy transfer units (e.g.: level 1 : u = n/2; level 2: u = n/4; level 3: u = n/8). The energy transfer units enable exchange of charge between one stack of k/2 battery cells to another stack of k/2 battery cells (e.g.: level 1 : k = 2; level 2: k = 4; level 3: k = 8).

Each energy transfer unit of a specific level has the same efficiency values η (e.g.: level 1 : η = 0.94; level 2: η = 0.96; level 3: η = 0.98). The battery cell with the smallest index of the energy transfer unit is defined as battery cell g.

The amount of energy to be transferred by the energy transfer unit of a specific unit can be cal- culated by the following equations.

g+k/2-1 g+k

∑ (CapDiff _g+x ) - ∑ (CapDijf _g+x )

EnergyShift =— ^ll≡

c cv _, fi < (g+ ^k ) ^: + EnergyShift · η

EnergyShift . = ·<

' [ i≥ (g + k) : - EnergyShift / (1 + η)

The following exemplarily calculation in table 7 illustrates the control of one embodiment of the energy transfer arrangement.

CapShift interim 20.485 Ah 20.485 Ah 19.485 Ah 19.485 Ah

LEVEL 2 (n=4, k=2, η=96%)

CapShift interim 20.485 Ah 20.485 Ah 19.485 Ah 19.485 Ah

ChargeShifti -0.500 Ah -0.500 Ah 0.500 Ah 0.500 Ah

/(1/96%= /(1/96%= ^■ 96%)= ^■ 96%)=

-0.510 Ah -0.510 Ah 0.490 Ah 0.490 Ah

ChargeLoss -0.020 Ah -0.020 Ah

CapShift final 19.975 Ah 19.975 Ah 19.975 Ah 19.975 Ah

Table 7

This results in a 0.485 Ah exchange and 0.060 Ah loss of energy between the battery cells 1 & 2 as well as between battery cell 3 & 4. The charge transfer unit of the second level transfers a total amount of 0.980 Ah between the stack of battery cell 1 & 2 and the stack of battery cells 3 & 4, with 0.040 Ah losses. This results in an energy loss of 0.100 Ah when 1.465 Ah are transferred between the battery cells, which results in an efficiency of 93.6%. The efficiency of the whole energy transfer arrangement is mainly dominated by the efficiency of the first level energy transfer unit efficiency.

In regards to a complete battery arrangement of several hundred batteries (e.g. 128 battery cells), the design of such an arrangement would need higher transferred energy in higher energy transfer levels. In a worst case scenario the theoretical necessary energy transfer for a specific level is exponentially growing with the ratio of the number of charge transfer units between the levels: max_transfer_energy(Level _j ) = max_transfer_energy(Level _i _ _] ) ^{■ U} ^ ^eve ^-^

u(Leveli)

In a high voltage system, this theoretical necessary transfer energy would make such a system cost inefficient and very expensive. A more detailed view the design of the energy transfer units show that in order to meet this requirement, the current profile drawn through the inductor and switches of the energy transfer unit must equal in all energy transfer levels. A higher energy transfer levels the costs for such a system increase due to higher withstand voltage at the same current profile.

The theoretical scenario is very unlikely needed when normal distributed capacity values are present out of battery cell production. The calculation of theoretical probability in a normal dis- tributed battery cell capacity arrangement is quite challenging and can be solved in a numeric calculation. The calculation of necessary energy transfer is performed on 10,000 random generated, normal distributed battery assemblies with 128 battery cells and the maximum energy of each level is calculated for each possibility. The maximum resulting energy of each of the 10,000 possibilities is then added to a diagram to show the necessary transfer energy at the different levels. The diagram in Fig. 6a) shows the exponentially increasing need of energy transfer in higher energy transfer level. However, there is no need to double the transfer energy for each energy transfer level. It may be sufficient to double the transfer energy only every fourth level.

Hence, energy transfer units with decreased current rating may be used for higher energy transfer levels. The anticipated current, which decreases with higher energy transfer levels, may be seen in Fig. 6b). The reduced current rating for higher energy transfer units may improve the efficiency of these units and/or reduce the electromagnetic influence (EMI) of these energy transfer units.

In a numerical calculation of the 10,000 random generated, normal distributed battery assemblies, the transfer losses are compared with two state of the art battery balancing arrangements. The first arrangement is a battery cell to battery cell energy transfer. In this design, the energy of one battery cell can be transferred a neighboring battery cell. In a worst case scenario, the energy of the first battery cell has to be transferred over 126 battery cells to the last battery cell. In a normal distributed battery cells assembly this worst case is rare, however is also unlikely that the transfer must only be exchanged between its neighboring battery cells. The second ar- rangement is a battery cell to battery system design, where energy is transferred by a DC/DC converter to the battery system and then via a DC/DC converter to the specific battery cell. These designs are dominated by the efficiency of the DC/DC converter.

In comparison of these energy transfer arrangements (Fig. 7), the efficiency of the cell-cell de- sign is strongly dependent on the battery cell configuration and varies strongly. The efficiency of the energy transfer is assumed to be 94%. The efficiency of the cell-system design stays constant due to the direct energy transfer of the DC/DC converter. The efficiency of the DC/DC converter is assumed to be 90%, where energy must be passed two times (from low cell voltage to medium voltage of the module to high voltage of the system) which results in a efficiency of 90% of cell to system and another 90% efficiency from the system back to cell - resulting in an overall efficiency of the system of 81 %. The underlying efficiency values for the proposed design are 94% for the first level (low voltage), 96% for the second level(low/medium voltage), 97% of the third level(medium voltage), 98% for the all other levels(high voltage). The energy for all battery assembly possibility with this design is more than 93%, which makes the system more energy efficient and cost efficient compared to all other designs. The specific arrangement of the energy transfer units can further be used for a safety relevant issue, which arises during assembly of an energy storage system. The connection between the battery cells and the energy transfer units is implemented either as high pin-count connector or single wires screwed to the battery cell. In both types it is not guaranteed that neighboring battery cells connect simultaneously. In a worst case scenario, this can result in a high voltage on any of the energy transfer switches. In order to implement electronic switches with less breakdown voltage, it is needed to limit the voltage if necessary. In an advantageous energy transfer design (as of Fig. 2a) or Fig. 2b)), the switches are connected in parallel to the stack of battery cells. If the voltage of the energy transfer reaches a high voltage, a safety circuit can switch on both (Fig. 2a)) or all four (Fig. 2b)) switches to reduce the voltage. The safety circuit measures the maximum voltage across the energy transfer unit and overrules the control signals to turn on both switches at the same time. Fig 4a) shows a safety circuit with a Zener diode to measure the maximum voltage of the energy transfer unit and two diodes that overrule both control signals. In Fig 4b) the diodes are replaced by switches for higher voltage and more powerful control signal overrule.

Besides the proposed energy transfer between battery cells and stacks of battery cells, the energy transfer units can be used as waveform generator to implement an electro-impedance- spectroscopy (EIS) for a specific low range of frequencies. The supported maximum frequency is dependent on the switching frequency of the charge transfer unit switches.

In order to support an electro-impedance spectroscopy, the switches of the energy transfer units are controlled in such a way that the resulting current through a battery cell forms a sinusoidal waveform. The resulting phase shift in voltage can be measured and plotted in a Bode or Nyquist diagram.

For example at a switching frequency of 100kHz, the resulting step frequency of the sinus control can be 10kHz, which then results in a sinusoidal current frequency of about 1 kHz. (Fig. 5)

In order to avoid inefficient continuous voltage balancing it is necessary to calculate the end of charge voltage, or any target SoC for each individual battery cell on a regular basis. The battery management system performs continuous load current measurements which are processed to a load profile. This profile is used for load current prediction during discharge and charge cycles. In addition to measurements, historic data is included in the calculation to achieve a most accurate approximation of the upcoming load current. Today, state-of-the-art prediction algorithm can be coupled to external prediction data from driver information systems, such as route planner and navigation systems.

In model regions for e-Mobility it has been demonstrated that usually less than half of the battery capacity of an EVs is being used. Consequently energy losses of active balancing systems could be reduced to a minimum, if energy transfer is started in the last moment necessary. In one configuration with limited resources a mean load current can be predicted with the principle of averaging the measured current. An adapted recursive average can be achieved with the fo influence of the new measured value.

In order to achieve accurate prediction during fast current changes and a long measurement time, the factor x is made dependent on the current state. Therefore for periods of time with no current the factor x is held to a low value. This allows quick adaptation to the new value. In all other cases the factor x is incremented to form over time a more accurate mean value.

The prediction of the load current allows the calculation of the remaining runtime to end-of- discharge voltage of the energy store. This remaining time is dependent on the remaining ca- pacity of each battery cell and their state-of-charge. The battery cell with the least amount of energy is defining the remaining runtime of the energy store.

training = ^ί ^η ^, ^■ SoC,, C _∑ ^■ SoC ₂ , C, ^■ SoC ₃ , . . .) / I _MEAN

During discharging the start point t for the beginning of the additional energy transfer is firstly dependent on the performance of the energy transfer units in the battery management system, i.e. on the technical implementation thereof, and secondly also on the sum of the differences in capacity of the cells in the energy store. In any case, the start point should be selected such that the maximum capacity of all of the cells is exhausted when said cells have reached their end-of- discharge voltage DVL. The start point k for the beginning of the additional energy transfer is advantageously defined as a specific state of charge of the highest capacity cell. The individual energy transfer of each cell is activated in the latest possible state-of-charge point, in order to reach their end-of-discharge voltage at the same time.

The start point k for the beginning of the additional energy transfer is advantageously defined as a specific state of charge of the highest capacity cell. The individual energy transfer of each cell is activated in the latest possible state-of-charge point, in order to reach their end-of-discharge voltage at the same time.

The individual start points are calculated from the energy to be transferred Επ-ransfer, the power of transfer unit r iTransfer and the transfer time transfer and is an iterative process.

First, an individual state-of-charge target value for each voltage source based on state-of-health parameters of said voltage source is determined in Step 1. The capacity loss due to transfer losses results from the amount of energy to be transferred and the efficiency of each individual transfer unit per cell.

n

Step 2: C _LossTransfer = ∑ [(c, - C _MIN ) ^■ (1 - ^)] Fehler! Es ist nicht moglich, durch die Bear- beitung von Feldfunktionen Objekte zu erstellen.

The time of the power transfer is relative to the mean energy of the energy store and the dis- charging current I MEAN. However, as for real systems, the transfer losses CLossTransfer must be considered for accurate start point calculation.

Step 3. t _Transfer — (C _MEAN ^LossTransfer) ^ ^MEAN

The power capability of the transfer unit can be calculated by the product of voltage and current.

Step 4: P _lTransfer = j A _iTransfer dt ^■ j V _∑ dt

SoCiA SoCiA

If the current of the energy transfer unit is controlled to be constant, then the integral of the current can be simplified to a constant value.

The necessary energy is dependent on the capacity of the selected cell C,, the minimum capacity cell CMIN of the system and the already transferred energy of the selected transfer unit C,Fin- ished-

Step 5: E _iTiansfer = (c, - C _MIN - C _iFinished ) ^■ \ v.dt

SoCiA

In a hardware topology other than in WO 2010/088944, i.e. a bidirectional energy transfer unit for individual cells, the calculation step 5 must be adapted.

Step 6. E _1Transfer — (c _i — C _MEAN — C _LossTransfer — C _1Flnlshed ) ^■ J ^" ^dt

SoCiA

The individual start points for the energy transfer are then the quotient of transfer energy and transfer power and time. Step 7: SoCiA = abs(EiTransfer PiTransfer) / transfer

The calculation steps can be repeated in defined time periods or continuously throughout the discharge of the energy store. During discharge of the energy store, the amount of transferred capacity iFinished I s counted.

In addition, for specific configurations could consider under a specific state-of-charge or voltage point close to the end-of-discharge a voltage balancing algorithm. This would allow balancing corrections caused by the imprecision of measurement or calculation. The lack of measurement and calculation accuracy can be induced by temperature change, chemical changes in the en- ergy store or integrative errors over time. The regulation threshold of the voltage balancing algorithm is advantageously selected as half voltage between the highest voltage and the lowest voltage, whereas the energy transfer of high voltage cells is activated.

^ ≥ (v _mx + V _mN ) / 2

Moreover the power of the energy transfer unit near the end-of-discharge voltage might be out of specification and the energy transfer might be terminated before reaching the individual or equal end-of-discharge voltage DVU

During charging the start point of the additional energy transfer is dependent on the application and the amount of energy to be transferred. Due to the fact that the power of the energy transfer unit might be out of specification, the start point t2 can be selected at a specific state-of-charge or voltage point.

The energy transfer is controlled to start immediately for all cells and is terminated when the amount of energy E ransfer has been transferred. The cell individual point of energy transfer completion, i.e. the target value, is defined as SOC,E. An individual state-of-charge target value for each voltage source based on state-of-health parameters of said voltage source is determined in Step 1. In order to calculate this state-of-charge point the transfer time t' Transfer and the transfer losses C'i_ossTransfer are calculated in step 2 and 3. Moreover the power capability of the energy transfer is calculated in step 4. ^{ste 2:} C ' _LossTransfer = ∑ [(c _MX - cj · (l - η _∑ )]

Step 3. t p

Step 4: P ' _1Transfer = j A _1Transfer dt ^■ jV. dt

This amount is related to the transferred energy during discharging, the charging efficiency k of the cell and further defines the point SOC,E of the individual transfer end. SoCiE

Step 5: E' _iTransfer = 1 / k ^■ (c _mx - C _∑ + C' _iFinished ) ^■ j v. dt

0

The individual end points SOC,E are defined by the quotient of transfer energy and transfer power and time.

Step 6: SoC _iE = abs(E' iTransfer / P iTransfer) / t Transfer

Alternatively, the amount of energy to be transferred can be determined by accumulating the energy amount for each battery cell during discharge. The accumulated value must be reversed as the charging current from a low capacity cell must be reduced.

In one implementation of the invention, the calculation and measurement error are considered and a voltage regulation algorithm is applied to the near end-of-charge voltage. This voltage regulation performs charge corrections of the individual cells in order to reach the selected end- of-charge voltage CVLi.

Cell aging brings about a change in the cell properties. This relates in particular to the cell ca- pacity, which does not reach a relatively stable value until approximately 100 charge/discharge cycles after the cell has first been brought into operation and then constantly decreases as the age increases.

It is therefore necessary to recalibrate the control system at regular time intervals as described in WO2012139604.

List of reference symbols

V Voltage

I Current

CVL End-of-charge voltage

CVLMAX End-of-charge voltage of strongest cell

CVLMIN End-of-charge voltage of weakest cell

CVU End-of-charge voltage of cell No. i

SoC State of charge

DVL End-of-discharge voltage

t1 , t2, ... , t7 Notable times in the operating process

CMAX Capacity of the strongest cell in a storage arrangement

CMIN Capacity of the weakest cell in a storage arrangement Ci , C ₂ , C ₃ , ... , C _N Storage cells