Skip to content

README.md

Latest commit

 

History

History
106 lines (89 loc) · 3.51 KB

README.md

File metadata and controls

106 lines (89 loc) · 3.51 KB

🔋 Battery ESC + SPKF Engine (Python + Rust)

Overview

This project was inspired by my coursework in Algorithms for Battery Management Systems at the University of Colorado at Boulder which I completed during my MS in Computer Science. Professor Gregory Plett has published several online resources related to this topic, including http://mocha-java.uccs.edu/BMS1/index.html.

I implemented a battery equivalent circuit (ESC) model and a Sigma-Point Kalman Filter (SPKF) for state-of-charge (SOC) estimation, with a dual-language architecture: Python: orchestration, data handling, testing, visualization Rust: high-performance numerical engine for ESC dynamics

The system is designed to support both offline analysis and future streaming, real-time inference pipelines.

Why This Matters

This project recasts matlab/Octave code into new languages that extend the reach of these valuable BMS related algorithms: hybrid Python + Rust system design numerical modeling + estimation algorithms production-ready validation (parity testing) extensibility toward real-time energy systems

Key Features

  1. ESC Battery Model Temperature-dependent parameter interpolation OCV(SOC, T) and dOCV/dSOC lookup RC branch dynamics and hysteresis modeling Fully validated against reference Python implementation
  2. SPKF Pipeline (Python baseline) Sigma-point generation Nonlinear state propagation Measurement update SOC estimation from voltage/current signals
  3. Rust Acceleration Layer Rust implementation of: interpolation parameter lookup ESC state equation ESC output equation Exposed via PyO3 + maturin Designed as a reusable compute engine
  4. Python ↔ Rust Bridge Clean interface via RustBatterySpkfEngine Model exported as flat dict → Rust struct Runtime toggle: USE_RUST = True / False
  5. Numerical Parity Validation A dedicated test suite verifies that Rust matches Python: ESC state evolution (ir, hk, soc) Predicted voltage Innovation

All comparisons match to machine precision (~1e-16).

Example Results

Python and Rust pipelines produce identical metrics: soc_rmse: 0.0014775 voltage_rmse: 7.98e-05

State trajectories and outputs are numerically equivalent.

Architecture

Python Layer ├── Data loading (MAT files) ├── SPKF orchestration ├── Testing + plotting └── Rust bridge adapter

Rust Layer ├── interp.rs (linear interpolation) ├── params.rs (temperature parameter lookup) ├── ocv.rs (OCV + derivative) ├── esc.rs (state + output equations) └── engine.rs (stateful compute engine)

Design Goals

Separate model orchestration (Python) from numerical compute (Rust) Enable future: real-time streaming pipelines multi-device simulation high-throughput inference Maintain strict numerical parity across implementations

Tech Stack

Python (NumPy, Pandas) Rust (PyO3, maturin) MATLAB data ingestion Battery modeling (ESC, SPKF)

Phase 1 Status

ESC model fully implemented in Rust Python ↔ Rust integration complete Numerical parity validated End-to-end SOC pipeline working

Next Steps

Full SPKF implementation in Rust (sigma-point update) Streaming integration (Kafka / Redpanda) Multi-cell / pack-level modeling GPU / SIMD optimization exploration

Rust Extension Setup

This project includes a Rust-based compute engine exposed to Python using PyO3 and maturin.

Prerequisites

  • Python 3.8+
  • Rust toolchain
  • maturin

Result