On-Line HEV Energy Management Using a Fuzzy Logic Yacine Gaoua 1,2,3 , StΓ©phane Caux 1 , Pierre Lopez 2,3 and Josep Domingo Salvany 4 1. Institut National Polytechnique de Toulouse, INPT 2. Laboratoire PLAsma et Conversion d'Energie, LAPLACE 3. Laboratoire d'Analyse et d'Architecture des Systemes, LAAS-CNRS 4. Nexter Electronics, NE ygaoua@laplace.univ-tlse.fr caux@laplace.univ-tlse.fr, lopez@laas.fr, j.domingo@nexter-group.fr
Outline of the presentation I. Introduction to HEV energy chain II. Sources characteristics III. Modeling IV. Solving method V. Off-line optimization VI. Results and performance VII. Conclusion
I. HEV energy chain Hybrid Electrical Vehicle HEV energy chain Battery Super-capacitor Fuel cell
II. Sources characteristics Parameter Meaning π± π π Demand of the powertrain (A) πππ , π± π πππ π± π Min/Max current exiting the PCube converter (A) πππ , π± ππ πππ π± ππ Min/Max current provided by the super-capacitor (A) πππ , π½ ππ πππ , π½ ππ (0) π½ ππ Min/Max/Initial voltage of the super-capacitor (V) πππ , π»π·π« πππ πππ , π»π·π« πππ (0) π»π·π« πππ Min/Max/Initial energy level in the battery pack (%) π«ππ πππ Battery capacity (Ah) π π Time stepsize (s) πΊ ππ Super-capacitor internal resistance (Ξ©) π« ππ Super-capacitor capacity (F) ππ΄πππ πππ Battery energy losses (kW) E π΄πππ π ππ Energy losses of the PCube converter (kW) Input parameters. Battery efficiency. Convertor efficiency.
III. Modeling Mathematical modeling Goal: Minimize battery discharge Under constrains of (system functioning, sources design, safety limitation), π½ πππ’ + π½ π = π½ πβ π½ πβ > 0 π½ πβ β€ π½ πππ’ + π½ π β€ 0 π½ πβ β€ 0 πππ β€ π½ π β€ π½ π πππ¦ π½ π πππ β€ π½ π‘π β€ π½ π‘π πππ¦ π½ π‘π πππ β€ π π‘π β€ π π‘π πππ¦ π π‘π πππ β€ πππ· πππ’ β€ πππ· πππ’ πππ¦ πππ· πππ’ π = π + πΉπππ‘π‘ πππ’ (π ) ( nlp ) π πππ’ πππ’ πππ’ 2 = π + πΉπππ‘π‘ ππ€π‘ π + π π‘π π½ π‘π π π‘π π π Decision variables: 100.πΉ πππ’ πππ· πππ’ = πππ· πππ’ 0 + βπ’ π·ππ πππ’ βπ’ π π‘π = π π‘π 0 + π½ π‘π + π π‘π + π· π‘π πΊ : Real battery current β’ π± πππ π± πππ : battery current β’ π πππ’ = π πππ· πππ’ 0 π»π·π« πππ : Battery State of charge β’ π πΉ πππ’ = π π½ πππ’ π½ πππ : Battery voltage β’ β’ π± ππ : Super-capacitor current β’ π½ ππ : Super-capacitor voltage π : Computation of electrical quantity π± π : Convertor current β’ π : Computation of battery voltage
IV. Solving method using fuzzy logic Super-capacitor voltage. Battery current. Powertrain demand. Parameters setting: Genetic algorithm (off-line - GPS) Control and correction algorithm ππ π± π π = . πππ π½ ππ = . ππππ π± πππ = . ππ Rules generation. Decision surface (centroid method). Rules engine.
V. Off-line optimization Global optimization π΅ππ πππ β π»π·π« πππ πΌ = π΅ππ π»π·π« πππ πΌ π½ πππ’ (π’) + π½ π (π’) = π½ πβ (π’) π½ πβ (π’) > 0 π½ πβ β€ π½ πππ’ + π½ π β€ 0 π½ πβ (π’) β€ 0 πππ β€ π½ π (π’) β€ π½ π πππ¦ π½ π πππ β€ π½ π‘π (π’) β€ π½ π‘π πππ¦ π½ π‘π πππ β€ π π‘π π’ β€ π π‘π πππ¦ π π‘π πππ β€ πππ· πππ’ (π’) β€ πππ· πππ’ πππ¦ πππ· πππ’ ( nlp) π π’ = π πππ’ π’ + πΉπππ‘π‘ πππ’ π πππ’ π’ π πππ’ Mission profile NE (176s). π½ π‘π (π’) 2 (π’) = π (π’) + πΉπππ‘π‘ ππ€π‘ π (π’) + π π‘π π π‘π π π 100.πΉ πππ’ π’ πππ· πππ’ π’ = πππ· πππ’ π’ β 1 + βπ’ π·ππ πππ’ βπ’ π π‘π (π’) = π π‘π π’ β 1 + π½ π‘π (π’) + π π‘π + π· π‘π π πππ’ = π πππ· πππ’ π’ β 1 π (π’) πΉ πππ’ = π π½ πππ’ Optimization using Operations Research methods: AMPL+ IpOpt algorithm (Interior Points)
VI. Results and performance HEV sources/ Number of cycles / HEV sources/ Number of cycles / Battery discharge Battery discharge Method Method HEV battery alone 30 Cycles β 88.3143% HEV battery alone 1 Cycle β 52.3566% HEV with PCube - FL HEV with PCube - FL 34 Cycles β 88.8872% 2 Cycles β 85.7596% HEV with PCube - GAFL 35 Cycles β 87.6296% HEV with PCube - GAFL 2 Cycles β 71.9029% HEV with PCube β IpOpt 39 Cycles β 88.7396% HEV with PCube β IpOpt 3 Cycles β 89.896% Battery discharge (1 cycle) Battery discharge (1 cycle) HEV with PCube HEV with PCube GAFL GAFL 2.546% 36.1712% IpOpt IpOpt 2.29315% 30.49% NE Mission profile 176s. Mission profile 3h 50min.
VII. Conclusions and perspectives Conclusions: β’ Genetic algorithm improve the solution by setting FL parameters off-line, β’ Good quality of the results (in regard to the global optimization), β’ Development of decision support tool in C + + (implementation in a dsp target). Perspectives: β’ Validation of results on a real prototype.
Thank you for your attention ygaoua@laplace.univ-tlse.fr caux@laplace.univ-tlse.fr, lopez@laas.fr, j.domingo@nexter-group.fr
Recommend
More recommend
