Abstract:Combining machine learning with physics is a trending approach for discovering unknown dynamics, and one of the most intensively studied frameworks is the physics-informed neural network (PINN). However, PINN often fails to optimize the network due to its difficulty in concurrently minimizing multiple losses originating from the system's governing equations. This problem can be more serious when the system's states are unmeasurable, like lithium-ion batteries (LiBs). In this work, we introduce a robust method for training PINN that uses fewer loss terms and thus constructs a less complex landscape for optimization. In particular, instead of having loss terms from each differential equation, this method embeds the dynamics into a loss function that quantifies the error between observed and predicted system outputs. This is accomplished by numerically integrating the predicted states from the neural network(NN) using known dynamics and transforming them to obtain a sequence of predicted outputs. Minimizing such a loss optimizes the NN to predict states consistent with observations given the physics. Further, the system's parameters can be added to the optimization targets. To demonstrate the ability of this method to perform various modeling and control tasks, we apply it to a battery model to concurrently estimate its states and parameters.

What problem does this paper attempt to address?

### Problems Addressed by the Paper This paper attempts to address the problem of effectively training Physics-Informed Neural Networks (PINNs) in physical systems where the state is unmeasurable. Specifically, the authors focus on the state estimation and parameter identification problems in Lithium-Ion Batteries (LiBs). #### Background Issues 1. **Optimization Challenges of PINNs**: - PINNs typically need to minimize multiple loss terms simultaneously, which originate from the system's governing equations. When the system state is unmeasurable, such as the internal state in lithium-ion batteries, this multi-loss optimization problem becomes more complex. - Multiple loss terms can lead to gradient imbalance, making the model difficult to train and prone to getting stuck in local optima or not converging. 2. **State Estimation and Parameter Identification of Lithium-Ion Batteries**: - Lithium-ion batteries are highly nonlinear, multi-scale systems whose internal states (such as State of Charge (SoC) and State of Health (SoH)) cannot be directly measured and must be inferred from available data. - Traditional model-based methods (such as Equivalent Circuit Models (ECM) and electrochemical models) and data-driven methods (such as neural networks) each have their limitations. The former relies on accurate model parameters, while the latter requires a large amount of training data and lacks an understanding of the physical processes. #### Solutions 1. **Simplifying the Loss Function**: - The authors propose a new PINN training framework that converts predicted states into predicted outputs through numerical integration and constructs a loss function to quantify the error between observed and predicted values. - This method reduces the number of loss terms, thereby simplifying the optimization process and avoiding gradient imbalance issues caused by multiple loss terms. 2. **Application to Lithium-Ion Batteries**: - The authors apply this method to a lithium-ion battery model, demonstrating its effectiveness in state estimation and parameter identification tasks. - Using this method, it is possible to estimate the internal state and model parameters of the battery using only measurable signals (such as input current and terminal voltage) without the need to pre-determine model parameters or prepare training data with known states. ### Summary The main contribution of this paper is the proposal of a new PINN training framework that simplifies the optimization process by reducing the number of loss terms, making it particularly suitable for physical systems with unmeasurable states. The authors validate the effectiveness and accuracy of this method in state estimation and parameter identification tasks through a case study on lithium-ion batteries.

Physics-Informed Neural Network for Discovering Systems with Unmeasurable States with Application to Lithium-Ion Batteries

State-space Modeling for Electrochemical Performance of Li-ion Batteries with Physics-Informed Deep Operator Networks

Physics-Informed Neural Networks for Prognostics and Health Management of Lithium-Ion Batteries

Hybrid Modeling of Lithium-Ion Battery: Physics-Informed Neural Network for Battery State Estimation

Physics-Informed Neural Networks for Modeling Li-ion Batteries: Solving the Single Particle Model Without Labeled Data

Inherently Interpretable Physics-Informed Neural Network for Battery Modeling and Prognosis

Coupling Neural Networks and Physics Equations For Li-Ion Battery State-of-Charge Prediction

Physics-Informed Neural Network for Spacecraft Lithium-Ion Battery Modeling and Health Diagnosis

A physics-informed neural network approach to parameter estimation of lithium-ion battery electrochemical model

PHYSICS-INFORMED NEURAL NETWORKS FOR DEGRADATION DIAGNOSTICS OF LITHIUM-ION BATTERIES

A Physics-Informed Machine Learning Approach for Estimating Lithium-Ion Battery Temperature

Physics-informed neural network for lithium-ion battery degradation stable modeling and prognosis

An Enhanced Single-Particle Model Using a Physics-Informed Neural Network Considering Electrolyte Dynamics for Lithium-Ion Batteries

MINN: Learning the dynamics of differential-algebraic equations and application to battery modeling

A Physics-Informed Composite Network for Modeling of Electrochemical Process of Large-Scale Lithium-Ion Batteries

Forward and Inverse Simulation of Pseudo-Two-Dimensional Model of Lithium-Ion Batteries Using Neural Networks

Physics-Informed Neural Networks for State of Health Estimation in Lithium-Ion Batteries

Machine Learning Through Physics–Informed Neural Networks: Progress and Challenges

A Combined Data-Model Method for State-of-Charge Estimation of Lithium-Ion Batteries.

Physics-informed neural networks for electrode-level state estimation in lithium-ion batteries

AutoPINN: When AutoML Meets Physics-Informed Neural Networks