There are several different types of models used in the energy storage sector, among them electrochemical, empirical, black box and table look-up models.
In terms of electrochemical modeling, a very large amount of research has been, and is being, performed at many places on several types of batteries. Much of this work is done by electrochemists, and has managed to explain many important phenomena.  However, electrochemical models require extensive and time-consuming calculations, making them unsuitable for battery system development.

As for empirical modeling, several published papers present battery measurements, and then fit simple equations to the data.  The value of such models is mainly in describing the performance of batteries already measured.  There is no guarantee that the empirical equations devised will work with other batteries of the same type, much less that they will work with batteries of different types.

This method has the advantage that its models can be computationally fast.  However, the values they predict are limited in the range of existing measurements. Thus, empirical modeling lacks predictive power. 

Another issue with empirical modeling is that the equations often used in it are developed to model a particular phenomenon, with little or no regard to other phenomena. If a quantity is used as part of the description of two different phenomena, the value for it may have to be different in each case.  This can cause interference between the modeling of different phenomena.  Finally, an effort to combine all the different empirical equations into one big empirical model can result in numerical issues, such as discontinuities of functions or their derivatives, which can result in a lack of accuracy and can have a disastrous effect in computer simulations when it comes to convergence. 
Two other types of models, black box and table-lookup, fit to data even more blindly than is the case for empirical models.  They completely lack predictive power.

What can be useful are physical compact models.  Such models have been used in other sectors, notably for semiconductor development. Compact models start from physical or electrochemical models, but are developed with an eye toward simplification and computational speed.  Their equations may look similar to those of empirical models, with a key difference: the parameters in them have physical / electrochemical significance and as such the values needed for them are close to those dictated by electrochemistry and physics.  Thus such models have the potential to combine the best features of electrochemical and empirical models: predictive power and computational efficiency.

Sendyne’s computationally efficient, quasi-static physical compact model is capable of accurately predicting cell and battery pack performance, life and safety.  This model is based on results from electrochemistry and physics and takes into consideration all the known physical phenomena and their interdependencies such as temperature effects, capacity fade, impedance growth, hysteresis, etc.  This model does not require a proprietary system to run and can interface to a wide range of simulation tools.

Adaptive to a variety of cell chemistries, the first iteration is being designed in conjunction with a major cell maker, who is providing the necessary real world data. Preliminary results are encouraging as the error in prediction falls below the statistical variations provided by the cell manufacturer.

Quoted from Sendyne company website. For the original article, please refer to Sendyne's website: http://www.sendyne.com//Technology/technology.html