Enabling a Shift from Reactive to Predictive Thermal Software and Controls
Automakers are making tremendous investments in new thermal architectures involving both hardware and software. Despite this investment, the systems have great leeway for improvement, particularly in software and controls. For Battery Electric Vehicles (BEVs) the importance of thermal controls may have a great influence on vehicle range and battery charging rates.
Control of first-generation BEV thermal systems was simpler because designs used relatively independent thermal loops and component hardware. The latest generations of electric vehicles involve increasing complexity and interdependence of the coolant, refrigerant, and oil based thermal components and systems. This complexity offers the potential for improving thermal control to transfer heat to the right place at the right time. State-of-the-art thermal systems reduce energy consumption, enable longer range, and allow higher battery charge and discharge rates.
Thermal hardware and software are undergoing rapid development to reduce cost and improve the value of thermal systems. ITB finds that the thermal hardware is maturing faster than software. This may be natural since hardware capability must be “unlocked” with software. Over time, software can be continuously updated to further enhance the value of a given hardware configuration.
Exhibit: Simplified Example of Advanced Neural Network Thermal Control
Sources: The ITB Group and Toyota
Vehicle thermal control systems involve complex, non-linear, multi-dimensional systems. Such systems involve many conductive or insulative materials, thermal transfer fluids, plus heat sources and sinks. Control is made by actuators, such as pumps, valves, louvers, flaps, and fans plus their interactive effects. Furthermore, heat generation varies greatly during driving and charging cycles. Additionally, external influences such as ambient temperature, solar loading and convective air flow have significant influence on thermal conditions and control.
Thermal software uses sensor data and load models to parameterize the control challenge. Thermal control must be predictive, based on driving and ambient conditions to perform cooling or heating functions, while minimizing energy consumption. Traditional vehicle thermal control approaches involve switching between states and control scenarios. Optimizing among many variables and state constraints traditionally involves data centric parameterization and calibration through look-up tables. Traditional control system development relies on validating the control scenarios for specific test cases. Increasingly, the development process uses modeling approaches and hardware in the loop testing.
Disturbances in real complex non-linear systems are difficult to accurately simulate, and therefore predictive models have limitations. Calibrated models are used to estimate current and future thermal demand and heat flow rates. Reactive controllers may use proportional, integral and/or differential methods supplemented by look-up tables to compensate for system changes and disturbances. The parameter sets for these look-up tables are optimized by engineers based on simulations and previous in-car experience. Best practices include using real-world data from vehicles on the road to further optimize objective functions like energy consumption. Companies like Geely, JLR, Hyundai, Porsche, and VW have been led by Tesla toward continuously improving thermal vehicle controls through software updates at dealers and/or Over-The-Air (OTA).
One solution for improving thermal control is data-driven machine learning control parameterization being developed by Porsche. Similarly, AI (artificial Intelligence) driven calibration and thermal control is being developed by Huawei for BAIC and other OEMs. These techniques can be combined with multi-layer neural network approaches being applied by Toyota, GM, and others.
Model-based predictive thermal control development applies to both electric powertrains and passenger comfort systems. Thermal control opportunities currently involve basic improvements. For example, Porsche has studied how model-based predictive control can better balance heat pump refrigerant compressor energy consumption with high-voltage PTC heating consumption. Porsche’s study showed a 21% reduction in cabin energy consumption on a dynamic duty cycle, while improving compressor life significantly. JLR recently demonstrated a 6% savings over a -7°C UDDS driving cycle through improved cabin air recirculation control. Controls development starts with optimization of existing designs, but new hardware and sensors can enable finer control and improved performance for high customer value.
ITB Provides Insights into Changing Automotive Thermal Technologies and Markets
ITB surveys automakers and organizations throughout the value chain to understand unmet needs and innovations for improving value. In our 2023 research, ITB will dive deeply into thermal controls development to identify leading techniques and companies plus barriers to be overcome. Contact The ITB Group to learn more about changing technology markets, and how to influence our research.