Technique for estimation of internal battery temperature

Abstract

One embodiment is a method for estimating an internal temperature of a battery, the method comprising obtaining multiple terminal impedance measurements, wherein each of the terminal impedance measurements is taken at a different one of a plurality of frequencies; determining model parameters for a multivariable polynomial regression model; and applying the multivariable polynomial regression model to the multiple terminal impedance measurements to estimate the internal temperature of the battery.

What is claimed is: 
     
         1 . A method for estimating an internal temperature of a battery, the method comprising:
 obtaining multiple terminal impedance measurements for the battery, wherein each of the terminal impedance measurements is obtained at a different one of a plurality of frequencies;   determining model parameters for a multivariable polynomial regression model; and   applying the multivariable polynomial regression model to the multiple terminal impedance measurements to estimate the internal temperature of the battery.   
     
     
         2 . The method of  claim 1 , wherein the battery comprises a rechargeable battery. 
     
     
         3 . The method of  claim 1 , wherein the plurality of frequencies are selected to compensate for effects of at least one of battery state of charge (SOC) and battery state of health (SOH) dependencies of the model. 
     
     
         4 . The method of  claim 1 , wherein the determining the model parameters comprises:
 obtaining training data from a set of training batteries; and   applying a linear least squares fit to the training data.   
     
     
         5 . The method of  claim 4 , wherein the training data comprises alternating current (AC) data and temperature data for the set of training batteries. 
     
     
         6 . The method of  claim 1  further comprising augmenting an equation comprising the multipolynomial regression model using at least one of an additional input and a monomial term. 
     
     
         7 . The method of  claim 6 , wherein the at least one of an additional input and a monomial term includes a battery terminal voltage. 
     
     
         8 . The method of  claim 6 , wherein the at least one of an additional input and a monomial term includes a charge capacity of the battery. 
     
     
         9 . The method of  claim 1  further comprising augmenting an equation comprising the multipolynomial regression model by adding a higher-order cross-term to the equation. 
     
     
         10 . The method of  claim 1  further comprising augmenting an equation comprising the multipolynomial regression model to include an input to compensate for battery state of charge (SOC). 
     
     
         11 . The method of  claim 1  further comprising augmenting an equation comprising the multipolynomial regression model by adding a function of another measurement of the battery to the equation. 
     
     
         12 . The method of  claim 1  further comprising augmenting an equation comprising the multipolynomial regression module to include a memory term. 
     
     
         13 . The method of  claim 1  further comprising calibrating the model parameters using at least one initial set of impedance measurements for the battery. 
     
     
         14 . The method of  claim 13 , wherein the calibrating the model parameters further includes:
 identifying a nearest neighbor dataset in a library of training datasets using the at least one initial set of impedance measurements; and   perturbing the model parameters to fit model parameters of the identified nearest neighbor dataset.   
     
     
         15 . The method of  claim 13  wherein the at least one set of initial impedance measurements is taken at a beginning of a life of the battery. 
     
     
         16 . The method of  claim 13 , wherein the at least one set of initial impedance measurements comprises a plurality of sets of initial impedance measurements. 
     
     
         17 . The method of  claim 1  further comprising calibrating the model parameters using a singular value decomposition (SVD) method. 
     
     
         18 . A method for estimating an internal temperature of a battery from terminal impedance measurements of the battery, the method comprising:
 obtaining multiple terminal impedance measurements for the battery at a plurality of frequencies;   deriving model parameters for a multivariable polynomial regression model, the deriving comprising:
 obtaining training data from a set of training batteries similar to the battery; and 
 applying a linear least squares fit to the training data; and 
   combining the multiple terminal impedance measurements using the multivariable polynomial regression model to produce an estimate of the internal temperature of the battery; wherein the plurality of frequencies span a range of frequencies selected to reduce dependence of the estimate of the internal temperature of the battery on a state of health (SOH) of the battery and a state of charge (SOC) of the battery.   
     
     
         19 . The method of  claim 18  further comprising augmenting an equation comprising the multipolynomial regression model using at least one of an additional input and a monomial term. 
     
     
         20 . The method of  claim 18  further comprising augmenting an equation comprising the multipolynomial regression model by adding a higher-order cross-term to the equation. 
     
     
         21 . The method of  claim 18  further comprising augmenting an equation comprising the multipolynomial regression model to include an input to compensate for battery state of charge (SOC). 
     
     
         22 . The method of  claim 18  further comprising augmenting an equation comprising the multipolynomial regression model by adding a function of another measurement of the battery to the equation. 
     
     
         23 . The method of  claim 18  further comprising augmenting an equation comprising the multipolynomial regression module to include a memory term. 
     
     
         24 . The method of  claim 18  further comprising calibrating the model parameters using at least one initial set of impedance measurements for the battery. 
     
     
         25 . The method of  claim 24 , wherein the calibrating the model parameters further includes:
 identifying a nearest neighbor dataset in a library of training datasets comprising the training data using the at least one initial set of impedance measurements; and   perturbing the model parameters to fit model parameters of the identified nearest neighbor dataset.   
     
     
         26 . The method of  claim 24  wherein the at least one set of initial impedance measurements is taken at a beginning of a life of the battery. 
     
     
         27 . The method of  claim 24 , wherein the at least one set of initial impedance measurements comprises a plurality of sets of initial impedance measurements. 
     
     
         28 . The method of  claim 18  further comprising calibrating the model parameters using a singular value decomposition (SVD) method. 
     
     
         29 . A system for estimating an internal temperature of a battery from a plurality of terminal impedance measurements obtained for the battery, wherein the plurality of terminal impedance measurements are taken at a plurality of frequencies, the system comprising:
 a multivariable polynomial regression model comprising a number of model parameters, the multivariable polynomial regression model combining the multiple terminal impedance measurements to generate an estimate the internal temperature of the battery;   wherein the plurality of frequencies span a range of frequencies selected to reduce dependence of the estimate of the internal temperature of the battery on a state of health (SOH) of the battery and a state of charge (SOC) of the battery.   
     
     
         30 . The system of  claim 29 , wherein the battery comprises a rechargeable battery. 
     
     
         31 . The system of  claim 29 , wherein model parameters of the multivariable polynomial regression model are determined by obtaining training data from a set of training batteries and applying a linear least squares fit to the training data. 
     
     
         32 . The system of  claim 29 , wherein an equation comprising the multipolynomial regression model includes at least one of an additional input and a monomial term. 
     
     
         33 . The system of  claim 32 , wherein the at least one of an additional input and a monomial term includes a battery terminal voltage. 
     
     
         34 . The system of  claim 32 , wherein the at least one of an additional input and a monomial term includes a charge capacity of the battery. 
     
     
         35 . The system of  claim 29 , wherein the equation comprising the multipolynomial regression model includes an input to compensate for battery state of charge (SOC). 
     
     
         36 . The system of  claim 29 , wherein an equation comprising the multipolynomial regression model comprises a function of another measurement of the battery to the equation. 
     
     
         37 . The system of  claim 29 , wherein an equation comprising the multipolynomial regression module includes a memory term. 
     
     
         38 . The system of  claim 29 , wherein the model parameters are calibrated using at least one initial set of impedance measurements for the battery. 
     
     
         39 . The system of  claim 29 , wherein the calibrating the model parameters further includes:
 identifying a nearest neighbor dataset in a library of training datasets using the at least one initial set of impedance measurements; and   perturbing the model parameters to fit model parameters of the identified nearest neighbor dataset.   
     
     
         40 . The system of  claim 39 , wherein the at least one set of initial impedance measurements is taken at a beginning of a life of the battery. 
     
     
         41 . The system of  claim 39 , wherein the at least one set of initial impedance measurements comprises a plurality of sets of initial impedance measurements. 
     
     
         42 . The system of  claim 29 , wherein the model parameters are calibrated using a singular value decomposition (SVD) method.