Particle filter and its potential applications in smart grid Zhiguo Shi Outline • • • • • Introduction to Zhejiang University Fundamental concept Particle filter algorithm Application to SOC/SOH of battery charge Discussion Outline • • • • • Introduction to Zhejiang University Fundamental concept Particle filter algorithm Application to SOC/SOH of battery charge Discussion Big picture Observed signal 1 sensor t Observed signal 2 Particle Filter Estimation t t Goal: Estimate a stochastic process given some noisy observations Concepts: – – 2015/4/8 Bayesian filtering Monte Carlo sampling Problem formulations • Estimate a stochastic process given some noisy observations • How? Step 1: Build system dynamic model State equation: xk=fx(xk-1, uk) xk state vector at time instant k fx state transition function uk process noise with known distribution 2015/4/8 Problem formulations • Estimate a stochastic process given some noisy observations • How? Step 2: Build observation model Observation equation: zk=fz(xk, vk) zk observations at time instant k fx observation function vk observation noise with known distribution 2015/4/8 Problem formulations • Estimate a stochastic process given some noisy observations • How? Step 3: Use particle filter Posterior x 2015/4/8 Motivations • The trend of addressing complex problems continues • Large number of applications require evaluation of integrals • Non-linear models • Non-Gaussian noise 2015/4/8 Applications • Signal processing – Image processing and segmentation – Model selection – Tracking and navigation • Communications – Channel estimation – Blind equalization – Positioning in wireless networks 2015/4/8 • Other applications1) – Biology & Biochemistry – Chemistry – Economics & Business – Geosciences – Immunology – Materials Science – Pharmacology & Toxicology – Psychiatry/Psychology – Social Sciences An Example y States: position and velocity xk=[xk, Vxk, yk, Vyk]T Observations: angle yk yk+1 Trajectory zk Observation equation: zk+1 xk xk+1 x State equation: Blue – True trajectory Red – Estimates zk zk=atan(yk/ xk)+vk xk=Fxk-1+ Guk Outline • • • • • Introduction to Zhejiang University Fundamental concept Particle filter algorithm Application to SOC/SOH of battery charge Discussion Basic Idea • Representing belief by sets of samples or particles Bel ( xt ) ~ St { x , w | i 1,..., n} i t • i t are nonnegative weights called importance factors • Updating procedure is sequential importance sampling with re-sampling i t w 2015/4/8 ISEE, ZJU Particle filter illustration Step 0: initialization Each particle has the same weight Step 1: updating weights. Weights are proportional to p(z|x) 2015/4/8 Particle filter illustration (Continued) Step 2: predicting. Predict the new locations of particles. Step 3: updating weights. Weights are proportional to p(z|x) Step 4: predicting. Predict the new locations of particles. Particles are more concentrated in the region where the person is more likely to be 2015/4/8 Particle filtering algorithm Initialize particles New observation Particle generation 1 2 ... M 1 2 ... M Weigth computation Normalize weights Output estimates Resampling Output yes More observations? no Exit 2015/4/8 Resampling M M ~ ( m) 1 x k 1 , M m 1 (m) 1 xk 2 , M m 1 x (m) k 1 , wk( m1) M m 1 M M ~ ( m) 1 xk , M m 1 (m) 1 xk 1 , M m 1 x (m) k , wk( m ) M m 1 M (m) 1 xk 1 , M m 1 x 2015/4/8 Outline • • • • • Introduction to Zhejiang University Fundamental concept Particle filter algorithm Application to SOC/SOH of battery charge Discussion Battery management in Electrical Vehicle [1] • The cost of the power system can reach up to 1/3 of the total cost of the electric vehicle. • The consistency of batteries is essential to the life and safety of the whole vehicle system [1] Gao, M., et al., Battery State of Charge online Estimation based on Particle Filter, Proceeding of the 4th International Congress on Image and Signal Processing, pp. 2233-2236, 2011. 2015/4/8 Battery capacity under different discharging rates 2015/4/8 System model • State Transition function Proportion coefficientt related to discharge rate Nominal capacity of battery Instantaniously discharge current • Observation function 2015/4/8 Simulation results 2015/4/8 Outline • • • • • Introduction to Zhejiang University Fundamental concept Particle filter algorithm Application to SOC/SOH of battery charge Discussion Hope: my crude remarks may draw forth by abler people • Fundamentally, the particle filter can be applied to systems described by state equation representation with state transition function and observation function. 2015/4/8 Battery Charge Management 2015/4/8 Smart Grid Network Status Control 2015/4/8 Short Term Electricity Price Prediction for Home Appliance Control 2015/4/8 2015/4/8