System for monitoring charging/discharging cycles of a rechargeable battery, and host device including a smart battery

• US 5,825,156 A
• Filed: 10/25/1996
• Issued: 10/20/1998
• Est. Priority Date: 10/31/1995
• Status: Expired due to Term

• Current Assignee: US Philips Corporation (Koninklijke Philips N.V.)
• Original Assignee: US Philips Corporation (Koninklijke Philips N.V.)
• Inventors: Gerard, Olivier, Patillon, Jean-Noel
• Primary Examiner(s): TSO, EDWARD H
• Application Number:
  US08/736,758
• Time in Patent Office: 725 Days
• Field of Search: 320/5, 320/12, 320/13, 320/9, 320/21, 320/27, 320/28, 320/29, 320/30, 320/39, 320/48
• US Class Current: 702/63
• CPC Class Codes: G01R 31/367 Software therefor, e.g. for battery testing using modelling or look-up tables
• CPC Class Codes: H02J 7/00712 the cycle being controlled or terminated in response to electric parameters

Abstract

System for monitoring the discharging/charging cycles of a rechargeable battery which includes adaptive calculation means for providing a predictive indication of when the battery will reach a critical discharge voltage. The adaptive calculation means includes parameters which can be modified by other adaptive calculation means so as to optimize the monitor'"'"'s performance depending on the battery'"'"'s actual use. The adaptive calculation means may be neural networks formed by a microprocessor and memory, and the monitor system may be coupled to a host system.

Claims

1. A monitoring system for monitoring discharging/charging cycles of a battery coupled to a rechargeable battery having charging periods which alternate with discharging periods depending on discharging/charging cycles, this system comprising:

first adaptive calculation means (NN1) having three ports, said means are arranged for collecting, at the beginning of a discharging period of a discharging/charging cycle of the battery, on the first port, a batch of operating parameters called first parameters (WjA), and for receiving on the second port a predetermined value of a critical discharge voltage threshold (V_TH), and for producing on the third port a calculated predictive indication of the instant (t_TH) at which the battery will reach this critical threshold (V_TH) that corresponds to the end of this discharging period,

said monitoring system also comprising;

second and third adaptive calculation means (NN2, NN3) coupled to the first calculation means, each having four ports and each are arranged for receiving, at an initial instant at the beginning of said discharging period of the battery, a value of the battery voltage called initial voltage (Vo) on the first port, a value of a variation (Δ

Vo) of this initial voltage after a short lapse of time starting from this initial instant on the second port, and a value of the initial number (No) of discharging/charging cycles of this battery effected before said discharging period on the third port, and which are arranged for producing, from the instant of said discharging period at which the initial values are available, a batch of approximate parameters (WjB) on the fourth port of said second calculator means (NN2), and a batch of corresponding correction parameters (WjC) on the fourth port of said third calculation means (NN3), which are added together to form said first operating parameters (WjA) imposed on the first port of said first calculation means (NN1).

2. A monitoring system as claimed in claim 1, also comprising:

a read/write memory area for recording, during said discharging period, batches of instantaneous real values,

said real values being formed by a discharge voltage measurement (Vt) of the battery and by the current instant (t) corresponding to this measurement, and

a calculator,

in which system, during the battery charging period (PC1) following said discharging period (PD1) of the respective discharging/charging cycle;

the first calculation means (NN1) are furthermore arranged for calculating and producing, a posteriori in an autonomous manner, parameters called real parameters (WjA*) on said first port of said first calculation means which correspond to the operation of said first calculation means (NN1) in the situation when the batches of instantaneous real values are imposed, while the discharge voltage measurement (Vt) is imposed on its second port and the corresponding current instant (t) is imposed on its third port,

the calculator (160) is arranged for receiving said approximate parameters (WjB) calculated by the second calculation means (NN2) during the discharging period (PD1), and said real parameters (WjA*) calculated by the first calculation means (NN1) during the charging period (PC1), and for producing respective differences between these parameters called error parameters (WjC*),

and the third calculation means (NN3) are arranged for calculating autonomously parameters called adaptive parameters (WkC) which correspond to the operation of these third calculation means in the situation when the error parameters (WjC*) are imposed on its fourth port, whereas the initial values (Vo, Δ

Vo, No) of the previous discharging period (PD1) are imposed on its first through third ports, and in which system the third calculation means (NN3) save, as operating parameters in the later discharging period of the next discharging/charging cycle, the adaptive parameters (WkC) calculated in said charging period (PC1).

3. A monitoring system as claimed in claim 1, in which:

the first, second and third calculation means (NN1, NN2, NN3) are formed by a first, second and third neural network respectively, the first operating parameters are the synaptic coefficients (WjA) of the first neural network, the first neural network (NN1) having an input cell (EC1A) for a voltage value at its second node and an output cell (NSA) for a time value at its third node,

the second neural network (NN2) having three input cells (NE1B, NE2B, NE3B) for said initial values (Vo, Δ

Vo, No) at its first through third ports, and a number of output cells (NE1B-NE13B) at its fourth port for the approximate parameters (WjB) having a like quantity to the synaptic coefficients (WjA) of the first neural network (NN1), and the third neural network (NN3) having three input cells (NE1C, NE2C, NE3C) for said initial values (Vo, Δ

Vo, No) at its first through third port and a number of output cells at its fourth port (NS1C-NS13C) for the correction parameters (WjC) having a like quantity to the number of synaptic coefficients (WjA) of the first neural network (NN1), and in which monitoring system;

the calculator (160) is arranged for receiving and adding together the approximate parameters (WjB) and the correction parameters (Wjc) and for producing said synaptic coefficients (WjA) imposed on the first neural network (NN1).

4. A monitoring system as claimed in claim 3, in which:

the first neural network (NN1), which forms the first calculation means, is arranged for calculating during the charging period (PC1) that follows the discharging period (PD1) of the respective discharging/charging cycle, by a reverse propagation method, real parameters (WjA*) which are its own real synaptic coefficients, in the situation when for each batch of instantaneous real values the measurement of the discharge voltage (Vt) is imposed on its second port, and the corresponding current instant (t) is imposed on its third port,

the calculator (160) is arranged for producing the error parameters (WjC*) formed by the respective differences between said real synaptic coefficients (WjA*) calculated by the first neural network (NN1) during said charging period (PC1), and said approximate parameters (WjB) calculated by the second neural network (NN2) for said previous discharging period (PD1),

the third neural network (NN3) forming the third calculation means is arranged for calculating, via a reverse propagation method, adaptive parameters (WkC) which are its own adaptive synaptic coefficients, in the situation when the error parameters (WkC*) are imposed on its fourth port and the initial values (Vo, Δ

Vo, No) of the previous discharging period (PD1) are imposed on its first through third ports,

and this third neural network (NN3) in the subsequent discharging period (PD2) of the next discharging/charging cycle maintains these adaptive synaptic coefficients (WkC) calculated in said charging period.

5. A monitoring system as claimed in claim 1, in which:

the calculator is also arranged for calculating and producing at each current instant (t), on the basis of the measurement of this current instant (t) and on the basis of the predictive indication of the critical instant (t_TH) produced by the first adaptive calculation means, a predictive indication of the lapse of time (Δ

t_TH) that is left from this current instant (t) onwards until the critical instant (t_TH) at which the battery will reach the predetermined critical discharge voltage threshold (V_TH).

6. A monitoring system as claimed in claim 3, in which:

the calculator is also arranged for calculating and producing at each current instant (t), on the basis of the measurement of this current instant (t) and on the basis of the predictive indication of the critical instant (t_TH) produced by the first neural network, a predictive indication of the lapse of time (Δ

t_TH) that is left from this current instant (t) onwards until the critical instant (t_TH) at which the battery will reach its predetermined critical discharge voltage threshold (V_TH).

7. A monitoring system as claimed in claim 3, in which the second neural network (NN2) is combined in series with the first neural network (NN1), and in which the third neural network (NN3) is combined in parallel with the second neural network.

8. A monitoring system as claimed in claim 7, in which the first neural network (NN1) has three layers, of which one input layer of a neural cell (NE1A) for a voltage value has a hidden layer of neural cells and an output layer having a single neural cell (NSA), the cells of the hidden layer having a sigmoidal activating function with slopes varying from one cell to the next, and the cell of the output layer having a linear activating function.

9. A monitoring system as claimed in claim 8, in which the second neural network (NN2) has three neural cell layers, of which one input layer of three neural cells (NE1B, NE2B, NE3B) for each of the initial values (Vo, Δ

Vo, No), one layer of hidden cells, and one output layer of neural cells (NS1B-NS13B), the cells of the hidden layer (NC1B-NC8B) having a sigmoidal activating function with different slopes from one cell to the next, and the cells of the output layer having a like quantity to the number of synaptic coefficients which are necessary for the operation of the first neural network and having a sigmoidal activating function.

10. A monitoring system as claimed in claim 9 in which the third neural network (NN3) has three neural cell layers, of which one input layer of three neural cells (NE1C, NE2C, NE3C) for each of the initial values (Vo, Δ

Vo, No), one layer of a single hidden cell (AU), and one output layer of neural cells (NS1C-NS13C), the cells of the output layer having a like quantity to the number of synaptic coefficients necessary for the operation of the first neural network (NN1), and having a sigmoidal activating function.

11. A monitoring system as claimed in claim 4, comprising a microprocessor for performing the calculations to form the first, second and third neural networks, and the calculator, and memory areas for storing the data, these memory areas being accessible via the microprocessor and including a memory area for storing the structure data of the neural networks, the fixed parameters and the synaptic coefficients of the second neural network, and a random access memory area for storing or producing the variable measurements, and the synaptic coefficients of the first and of the third neural networks.

12. A monitoring system as claimed in claim 1, this system being coupled to a rechargeable battery, to time measuring means, voltage measuring means, and display means arranged for producing either an indication of the critical instant (t_TH) at which the battery will reach the critical voltage threshold (vt_TH), or an indication of the lapse of time (Δ

t_TH) that is left from a current instant of use (t) onwards until the instant (t_TH) at which the battery will reach a predetermined critical voltage threshold (Vt_TH), or the two indications, and possibly an indication of the end of the recharging period of the battery.

13. A host device fed by a rechargeable battery and comprising a monitoring system as claimed in claim 1, coupled to this battery.

14. Method of producing neural networks of the monitoring system as claimed in claim 7, comprising in a learning period:

the learning by the first neural network, of discharging time curves (t) as a function of the discharge voltage (Vt) during which discharge voltages (Vt) are imposed on the second port of the first neural network normally intended for the voltage value, and the corresponding instants (t) are imposed on the third port for forming a data base formed by vectors of the synaptic coefficients (WjA) of this first neural network,

the learning by the second neural network of relations between the initial values (Vo, Δ

Vo, No) and the synaptic coefficients (WjA) of the first neural network, determined in its learning procedure,

the learning of the third neural network to make it determine its own adaptive synaptic coefficients (WkC), this learning comprising the steps of;

1) in the discharging period (PD1) of a discharging/charging cycle of the battery, storing the initial values (Vo, Δ

Vo, No) and instantaneous real values of the battery voltage (Vt) as a function of the corresponding current instant (t), and approximate parameters (WjB) produced by the second neural network (NN2),

2) in the subsequent charging period (PC1), this learning comprising the sub-steps of;

2a) calculating by the first neural network its own real synaptic coefficients (WjA*) when the instantaneous voltage (Vt) and time (t) values are imposed on its second and third ports respectively,

2b) calculating error parameters (WjC*) by calculating the differences between the real parameters (WjA*) and the approximate parameters (WjB),

2c) calculating by the third neural network (NN3) its own synaptic coefficients (WkC) called adaptive parameters when the error parameters (WjC*) are imposed on its fourth port and the initial values (Vo, Δ

Vo, No) are imposed on its first through third ports, and

method of use of the monitoring system comprising the use of said adaptive parameters (WkC) as synaptic coefficients of the third neural network (NN3) in the later discharging period of the next discharging/charging cycle.

15. A monitoring system as claimed in claim 2, in which:

the first, second and third calculation means (NN1, NN2, NN3) are formed by a first, second and third neural network respectively, the first operating parameters are the synaptic coefficients (WjA) of the first neural network, the first neural network (NN1) having an input cell (EC1A) for a voltage value at its second node and an output cell (NSA) for a time value at its third node,

the second neural network (NN2) having three input cells (NE1B, NE2B, NE3B) for said initial values (Vo, Δ

Vo, No) at its first through third ports, and a number of output cells (NE1B-NE13B) at its fourth port for the approximate parameters (WjB) having a like quantity to the synaptic coefficients (WjA) of the first neural network (NN1), and the third neural network (NN3) having three input cells (NE1C, NE2C, NE3C) for said initial values (Vo, Δ

Vo, No) at its first through third port and a number of output cells at its fourth port (NS1C-NS13C) for the correction parameters (WjC) having a like quantity to the number of synaptic coefficients (WjA) of the first neural network (NN1), and in which monitoring system;

the calculator (160) is arranged for receiving and adding together the approximate parameters (WjB) and the correction parameters (WjC) and for producing said synaptic coefficients (WjA) imposed on the first neural network (NN1).

16. A monitoring system as claimed in claim 15, in which:

the first neural network (NN1), which forms the first calculation means, is arranged for calculating during the charging period (PC1) that follows the discharging period (PD1) of the respective discharging/charging cycle, by a reverse propagation method, real parameters (WjA*) which are its own real synaptic coefficients, in the situation when for each batch of instantaneous real values the measurement of the discharge voltage (Vt) is imposed on its second port, and the corresponding current instant (t) is imposed on its third port,

the calculator (160) is arranged for producing the error parameters (WjC*) formed by the respective differences between said real synaptic coefficients (WjA*) calculated by the first neural network (NN1) during said charging period (PC1), and said approximate parameters (WjB) calculated by the second neural network (NN2) for said previous discharging period (PD1),

the third neural network (NN3) forming the third calculation means is arranged for calculating, via a reverse propagation method, adaptive parameters (WkC) which are its own adaptive synaptic coefficients, in the situation when the error parameters (WkC*) are imposed on its fourth port and the initial values (Vo, Δ

Vo, No) of the previous discharging period (PD1) are imposed on its first through third ports,

and this third neural network (NN3) in the subsequent discharging period (PD2) of the next discharging/charging cycle maintains these adaptive synaptic coefficients (WkC) calculated in said charging period.

17. A monitoring system as claimed in claim 2, in which:

the calculator is also arranged for calculating and producing at each current instant (t), on the basis of the measurement of this current instant (t) and on the basis of the predictive indication of the critical instant (t_TH) produced by the first adaptive calculation means, a predictive indication of the lapse of time (Δ

t_TH) that is left from this current instant (t) onwards until the critical instant (t_TH) at which the battery will reach the predetermined critical discharge voltage threshold (V_TH).

18. A monitoring system as claimed in claim 4, in which:

the calculator is also arranged for calculating and producing at each current instant (t), on the basis of the measurement of this current instant (t) and on the basis of the predictive indication of the critical instant (t_TH) produced by the first neural network, a predictive indication of the lapse of time (Δ

t_TH) that is left from this current instant (t) onwards until the critical instant (t_TH) at which the battery will reach its predetermined critical discharge voltage threshold (V_TH).

19. A monitoring system as claimed in claim 15, in which:

the calculator is also arranged for calculating and producing at each current instant (t), on the basis of the measurement of this current instant (t) and on the basis of the predictive indication of the critical instant (t_TH) produced by the first neural network, a predictive indication of the lapse of time (Δ

t_TH) that is left from this current instant (t) onwards until the critical instant (t_TH) at which the battery will reach its predetermined critical discharge voltage threshold (V_TH).

20. A monitoring system as claimed in claim 16, in which:

the calculator is also arranged for calculating and producing at each current instant (t), on the basis of the measurement of this current instant (t) and on the basis of the predictive indication of the critical instant (t_TH) produced by the first neural network, a predictive indication of the lapse of time (Δ

t_TH) that is left from this current instant (t) onwards until the critical instant (t_TH) at which the battery will reach its predetermined critical discharge voltage threshold (V_TH).

21. A monitoring system as claimed in claim 4, in which the second neural network (NN2) is combined in series with the first neural network (NN1), and in which the third neural network (NN3) is combined in parallel with the second neural network.

22. A monitoring system as claimed in claim 6, in which the second neural network (NN2) is combined in series with the first neural network (NN1), and in which the third neural network (NN3) is combined in parallel with the second neural network.

23. A monitoring system as claimed in claim 15, in which the second neural network (NN2) is combined in series with the first neural network (NN1), and in which the third neural network (NN3) is combined in parallel with the second neural network.

24. A monitoring system as claimed in claim 10, comprising a microprocessor for performing the calculations to form the first, second and third neural networks, and the calculator, and memory areas for storing the data, these memory areas being accessible via the microprocessor and including a memory area for storing the structure data of the neural networks, the fixed parameters and the synaptic coefficients of the second neural network, and a random access memory area for storing or producing the variable measurements, and the synaptic coefficients of the first and of the third neural networks.

25. A monitoring system as claimed in claim 16, comprising a microprocessor for performing the calculations to form the first, second and third neural networks, and the calculator, and memory areas for storing the data, these memory areas being accessible via the microprocessor and including a memory area for storing the structure data of the neural networks, the fixed parameters and the synaptic coefficients of the second neural network, and a random access memory area for storing or producing the variable measurements, and the synaptic coefficients of the first and of the third neural networks.

26. A monitoring system as claimed in claim 3, this system being coupled to a rechargeable battery, to time measuring means, voltage measuring means, and display means arranged for producing either an indication of the critical instant (t_TH) at which the battery will reach the critical voltage threshold (vt_TH), or an indication of the lapse of time (Δ

t_TH) that is left from a current instant of use (t) onwards until the instant (t_TH) at which the battery will reach a predetermined critical voltage threshold (Vt_TH), or the two indications, and possibly an indication of the end of the recharging period of the battery.

27. A monitoring system as claimed in claim 4, this system being coupled to a rechargeable battery, to time measuring means, voltage measuring means, and display means arranged for producing either an indication of the critical instant (t_TH) at which the battery will reach the critical voltage threshold (vt_TH), or an indication of the lapse of time (Δ

t_TH) that is left from a current instant of use (t) onwards until the instant (t_TH) at which the battery will reach a predetermined critical voltage threshold (Vt_TH), or the two indications, and possibly an indication of the end of the recharging period of the battery.

28. A monitoring system as claimed in claim 11, this system being coupled to a rechargeable battery, to time measuring means, voltage measuring means, and display means arranged for producing either an indication of the critical instant (t_TH) at which the battery will reach the critical voltage threshold (vt_TH), or an indication of the lapse of time (Δ

t_TH) that is left from a current instant of use (t) onwards until the instant (t_TH) at which the battery will reach a predetermined critical voltage threshold (Vt_TH), or the two indications, and possibly an indication of the end of the recharging period of the battery.

29. A host device fed by a rechargeable battery and comprising a monitoring system as claimed in claim 28, coupled to this battery.

30. Method of producing neural networks of the monitoring system as claimed in claim 12, comprising in a learning period:

the learning by the first neural network, of discharging time curves (t) as a function of the discharge voltage (Vt) during which discharge voltages (Vt) are imposed on the second port of the first neural network normally intended for the voltage value, and the corresponding instants (t) are imposed on the third port for forming a data base formed by vectors of the synaptic coefficients (WjA) of this first neural network,

the learning by the second neural network of relations between the initial values (Vo, Δ

Vo, No) and the synaptic coefficients (WjA) of the first neural network, determined in its learning procedure,

the learning of the third neural network to make it determine its own adaptive synaptic coefficients (WkC), this learning comprising the steps of;

1) in the discharging period (PD1) of a discharging/charging cycle of the battery, storing the initial values (Vo, Δ

Vo, No) and instantaneous real values of the battery voltage (Vt) as a function of the corresponding current instant (t), and approximate parameters (WjB) produced by the second neural network (NN2),

2) in the subsequent charging period (PC1), this learning comprising the sub-steps of;

2a) calculating by the first neural network its own real synaptic coefficients (WjA*) when the instantaneous voltage (Vt) and time (t) values are imposed on its second and third ports respectively,

2b) calculating error parameters (WjC*) by calculating the differences between the real parameters (WjA*) and the approximate parameters (WjB),

2c) calculating by the third neural network (NN3) its own synaptic coefficients (WkC) called adaptive parameters when the error parameters (WjC*) are imposed on its fourth port and the initial values (Vo, Δ

Vo, No) are imposed on its first through third ports, and

method of use of the monitoring system comprising the use of said adaptive parameters (WkC) as synaptic coefficients of the third neural network (NN3) in the later discharging period of the next discharging/charging cycle.

31. Method of producing neural networks of the monitoring system as claimed in claim 24, comprising in a learning period:

the learning by the first neural network, of discharging time curves (t) as a function of the discharge voltage (Vt) during which discharge voltages (Vt) are imposed on the second port of the first neural network normally intended for the voltage value, and the corresponding instants (t) are imposed on the third port for forming a data base formed by vectors of the synaptic coefficients (WjA) of this first neural network,

the learning by the second neural network of relations between the initial values (Vo, Δ

Vo, No) and the synaptic coefficients (WjA) of the first neural network, determined in its learning procedure,

the learning of the third neural network to make it determine its own adaptive synaptic coefficients (WkC), this learning comprising the steps of;

1) in the discharging period (PD1) of a discharging/charging cycle of the battery, storing the initial values (Vo, Δ

Vo, No) and instantaneous real values of the battery voltage (Vt) as a function of the corresponding current instant (t), and approximate parameters (WjB) produced by the second neural network (NN2),

2) in the subsequent charging period (PC1), this learning comprising the sub-steps of;

2a) calculating by the first neural network its own real synaptic coefficients (WjA*) when the instantaneous voltage (Vt) and time (t) values are imposed on its second and third ports respectively,

2b) calculating error parameters (WjC*) by calculating the differences between the real parameters (WjA*) and the approximate parameters (WjB),

2c) calculating by the third neural network (NN3) its own synaptic coefficients (WkC) called adaptive parameters when the error parameters (WjC*) are imposed on its fourth port and the initial values (Vo, Δ

Vo, No) are imposed on its first through third ports, and

method of use of the monitoring system comprising the use of said adaptive parameters (WkC) as synaptic coefficients of the third neural network (NN3) in the later discharging period of the next discharging/charging cycle.