Particle Filter Battery Parameter Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing battery parameter estimation systems, such as those using Kalman filter designs, face challenges in adapting to changes due to battery age and operating conditions, struggle with concurrent estimation of battery states and parameters, and often require extensive calibration and fail to accurately recover from initial measurement inaccuracies.
Innovation Solution
The use of a particle filter method or Sequential Monte Carlo method that generates a plurality of particles to simulate battery systems, allowing for the joint estimation of state of charge (SOC), state of health (SOH), and open-circuit voltage by propagating particles through a system model, weighting them based on measurements, and iteratively refining estimates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If Kalman filter design is used for battery parameter estimation, then the estimation process can be implemented, but the system cannot adapt to changes resulting from battery age and operating conditions
Solution Approach 1:
The particle filter method dynamically adapts to changing battery conditions by generating multiple particles representing different possible states and parameters. Each particle is propagated through the system model and weighted based on how well it matches actual measurements, allowing the estimation to automatically adapt to battery aging and operating condition changes without requiring manual recalibration.
Solution Approach 2:
The system changes the estimation approach from a single deterministic state (Kalman filter) to a distribution of possible states (particles). By representing battery parameters as a set of particles with different weights, the system can capture uncertainty and adapt to parameter changes over time, improving both adaptability and reliability.
2Extent of automation
If classic least square design or extended Kalman filter design is used, then battery parameters can be estimated, but concurrent estimation of battery states and parameters is not enabled
Solution Approach 1:
The particle filter methodology merges state estimation and parameter estimation into a single unified framework. Both battery states (SOC, temperature) and parameters (capacity, resistance) are represented as particles and updated simultaneously through the same propagation and weighting process, enabling concurrent estimation without requiring separate systems.
Solution Approach 2:
The particle filter system serves multiple functions simultaneously: it estimates battery state of charge, state of health, and various electrical parameters all within one algorithmic framework. This multi-functional approach reduces overall system complexity compared to having separate estimation systems for each parameter type.
3Reliability
If conventional estimation systems are used, then initial parameter estimation can be performed, but the system cannot recover from inaccurate initial SOC or battery capacity measurement
Solution Approach 1:
The particle filter implements continuous feedback by comparing particle predictions with actual measurements at each time step. Particles that diverge from actual measurements receive lower weights, while those that match well receive higher weights. This feedback mechanism allows the system to automatically correct initial estimation errors without manual intervention or calibration time loss.
Solution Approach 2:
The system performs preliminary exploration of the state space by generating multiple particles with different initial values before actual operation begins. This preliminary action ensures that even if the true state is not exactly represented by any single particle, the correct state is likely included in the particle set, enabling automatic recovery without calibration.
4Measurement precision
If particle filter method is used for joint estimation of battery states and parameters, then estimation accuracy improves, but computational complexity increases
Solution Approach 1:
The particle filter methodology segments the estimation problem into discrete particles, each representing a possible state-parameter combination. By dividing the continuous estimation space into discrete particles, the system achieves accurate joint estimation while maintaining computational tractability through parallel processing of individual particles rather than solving a complex continuous optimization problem.
Data Source
AI summary
The present disclosure relates to estimation of battery parameters, including SOC and SOH using a plurality of particles, each of which represents at least one parameter and at least one state of a battery system. A system model may simulate the battery system using processing logic. Each of the plurality of particles may propagate through the system model to generate a plurality of modeled particle values, each of which may be compared to a measurement of an electrical parameter of the battery system. Each particle may be weighted based upon a comparison of the modeled particle value and the measurement. A successive plurality of particles may be generated based upon the weight assigned to each of the plurality of modeled particle values. A number of iterations may be performed to generate a tuned plurality of modeled particle values, upon which an estimated parameter of the battery system may be based.