Mixed-Fleet Energy Forecasting with Shared and Vehicle-Specific Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Public transit agencies face challenges in optimizing route assignments and charging schedules for mixed fleets of vehicles, requiring accurate trip-level predictions of energy use across different vehicle types and conditions.
Innovation Solution
A system utilizing neural networks to generate predictive functions for energy consumption in mixed-vehicle fleets, incorporating shared and vehicle-specific layers to identify features indicative of energy consumption and account for various vehicle types, traffic, weather, and route conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If a mixed fleet of vehicles (electric, hybrid, internal combustion) is employed to reduce upfront costs, then capital expenditure is reduced, but optimizing route assignments and charging schedules becomes significantly more complex
Solution Approach 1:
The patent segments the optimization problem into vehicle-specific sub-problems, creating separate predictive models for each vehicle type (electric, hybrid, internal combustion). This allows the complex mixed-fleet optimization to be broken down into manageable components, where each vehicle class is optimized according to its specific energy consumption characteristics, thereby reducing overall optimization complexity while maintaining fleet diversity
Solution Approach 2:
The patent changes the optimization parameters based on vehicle type by incorporating vehicle-specific energy consumption models that account for different fuel/energy types, efficiency characteristics, and operational constraints. This enables the optimization system to adapt parameters dynamically according to the mixed fleet composition, making the complex optimization problem solvable through parameter differentiation
2Measurement precision
If accurate trip-level predictions of energy use are made for multiple vehicle classes and conditions, then route assignment and charging schedule optimization is enabled, but data processing and model complexity increase
Solution Approach 1:
The patent creates a universal predictive framework that handles multiple vehicle classes (electric, hybrid, internal combustion) and various operating conditions through a single integrated system. The model uses common input features (route characteristics, traffic conditions, weather) across all vehicle types while incorporating vehicle-specific parameters, enabling accurate predictions for the entire mixed fleet without requiring separate independent models for each vehicle class
Solution Approach 2:
The patent applies local quality by incorporating vehicle-specific layers or parameters within the predictive model that capture the unique energy consumption characteristics of each vehicle class. While the overall framework remains unified, each vehicle type receives customized treatment through vehicle-specific features (e.g., battery characteristics for electric vehicles, fuel injection parameters for combustion engines), thereby achieving high prediction accuracy without excessive complexity
3Use of energy by moving object
If electric vehicles are used to reduce energy costs and environmental impact, then operational energy efficiency improves, but high upfront costs require combination with conventional vehicles
Solution Approach 1:
The patent implements dynamic optimization that adaptively assigns routes and charging schedules based on real-time and historical data, allowing the mixed fleet to dynamically adjust to changing conditions. This enables the system to maximize the utilization of energy-efficient electric vehicles when conditions are favorable while seamlessly integrating conventional vehicles when needed, thereby achieving improved overall energy efficiency without requiring a complete fleet transition
Data Source
AI summary
A system for forecasting energy consumption by vehicles in a mixed-vehicle fleet. The system includes neural network(s) that generate a predictive function for vehicles in each of a number of classes (e.g., electric vehicles, hybrid vehicles, internal combustion vehicles, vehicle models, model years, etc.). To capture both the generalizable patterns that govern energy consumption across all vehicle classes and the features and relationships that are specific to each class, the neural network(s) include a multi-task learning model that includes shared layers for all of the classes of vehicles and a set of vehicle-specific layers for each class. In some of those embodiments, for example to predict to energy consumption of an additional class with limited data, the neural networks further include an inductive transfer learning model that includes the shared layers transferred from the multi-task learning model and vehicle-specific layers for the additional class.