Cooperative control device and method for hybrid power system of new energy automobile

Abstract

The invention discloses a new energy automobile hybrid power system cooperative control device and method.The new energy automobile hybrid power system cooperative control device comprises a multi-source data collection module, a cooperative control center and an execution unit, and the multi-source data collection module transmits collected information data to the cooperative control center; the cooperative control center comprises a data fusion unit, an energy management module, a heat management module and a cooperative decision-making unit, the data fusion unit receives data information of the multi-source data acquisition module, carries out fusion processing, and transmits the data information to the energy management module and the heat management module; the energy management module and the heat management module output a control strategy to the collaborative decision-making unit, and the collaborative decision-making unit judges a correction strategy and then sends a control instruction to the execution unit. The method has the advantages that cooperative control of energy management and heat management is achieved, various dimension data of road conditions, driving behaviors, power states and heat source states are fused, the problem of system unbalance caused by single module optimization is solved, and dynamic cooperation of energy management and heat management is achieved.

Description

TECHNICAL FIELD

[0001] The application relates to a vehicle control device and method, in particular to a new energy vehicle hybrid power system cooperative control device and method, and belongs to the technical field of new energy vehicles. BACKGROUND

[0002] With the continuous development of the automobile industry and the continuous improvement of environmental protection requirements, hybrid power systems have been widely used in commercial vehicles. However, the hybrid power system of commercial vehicles still faces many technical problems, for example, in terms of energy management, it faces the technical problems of poor adaptability to complex road conditions (such as urban congestion and mountain slopes) and driving behavior (such as aggressive acceleration and smooth cruising), and it is difficult for the optimization algorithm to balance accuracy and real-time performance.

[0003] In terms of thermal management, there are problems such as frequent starting of dedicated engines, slow preheating speed, high cold start emissions, and difficulties in coordinating and managing multiple heat sources. More importantly, energy management and thermal management are usually designed separately as independent systems, but the engine speed, load and combustion efficiency of the engine and the output power of the motor will directly affect the heat production of each component, and the temperature of the battery and the engine coolant will in turn restrict the motor output power and engine efficiency, resulting in suboptimal overall energy efficiency of the system. These all restrict the further improvement of the performance of the hybrid power system. SUMMARY

[0004] The purpose of the present application is to solve the problems existing in the prior art and provide a new energy vehicle hybrid power system cooperative control device and method. The two cooperative control of "energy management-thermal management" is realized, and multiple dimension data such as road conditions, driving behavior, power state and heat source state are fused to solve the system imbalance problem caused by single module optimization and realize the dynamic cooperation of energy management and thermal management.

[0005] Technical scheme: A new energy vehicle hybrid power system cooperative control device comprises a multi-source data acquisition module, a cooperative control center and an execution unit. The multi-source data acquisition module transmits the collected information data to the cooperative control center. The cooperative control center sends control instructions to the execution unit after processing and analyzing the collected data information. The cooperative control center comprises a data fusion unit, an energy management module, a thermal management module and a cooperative decision unit. The data fusion unit receives and fuses the data information of the multi-source data acquisition module. The data fusion unit transmits the fused data to the energy management module and the thermal management module respectively. The energy management module and the thermal management module output control strategies to the cooperative decision unit according to the data information. The cooperative decision unit sends control instructions to the execution unit after judging and correcting the strategies.

[0006] The present application overcomes the defects of the existing hybrid power system in energy management and thermal management, realizes intelligent collaborative control of the power system and efficient use of energy, and improves the reliability and performance of the vehicle.

[0007] Preferably, in order to realize the collection of multi-source data, the multi-source data collection module comprises a road condition sensor, a driving behavior sensor, a power component state sensor and a heat source state sensor; the road condition sensor collects vehicle speed, slope and front vehicle distance information in real time; the driving behavior sensor collects throttle opening and brake frequency information in real time; the power component state sensor collects engine parameters, motor parameters and battery parameters in real time; the heat source state sensor collects temperature parameters and cooling liquid parameters of the engine, the motor and the battery in real time, and the collected data is transmitted to the collaborative control hub.

[0008] Preferably, in order to obtain the optimal energy distribution strategy, the energy management module comprises a working condition dynamic prediction model, a driving style recognition model and a multi-target prediction rolling optimization model; the optimal engine / motor torque distribution scheme, engine intervention / exit timing and energy recovery scheme are generated according to the data fusion unit fused data.

[0009] Preferably, in order to obtain an adaptive thermal management strategy, the thermal management module comprises a high-low temperature double-loop cooling system and a heat flow evolution law model, and an adaptive thermal control module.

[0010] The high temperature loop of the high-low temperature double-loop cooling system is connected with the engine cylinder / block, adopts an overall reverse surrounding water jacket and a double-layer water jacket structure of the cylinder head; the low temperature loop is connected with the motor, the battery and the electronic control, adopts a piston high-position internal cooling oil way; the two loops are connected and switched through an electronic three-way valve, and are equipped with an electronic water pump, an electronic thermostat and an intelligent fan; according to the heat source state data, an adaptive thermal management strategy is generated, including an electronic water pump speed control scheme, an electronic thermostat opening degree and a fan speed regulation scheme.

[0011] A new energy vehicle hybrid power system collaborative control method, comprising the following steps:

[0012] Step one, data collection and fusion, the multi-source data collection module collects road condition, driving behavior, power component state and heat source state data in real time, and transmits them to the data fusion unit of the collaborative control hub; the data fusion unit filters and calibrates the original data, integrates multi-source information through a weighted fusion algorithm, and generates a fused data set;

[0013] Step two, multi-energy energy management optimization, the energy management module receives the fused data set, predicts the road condition in the next 5-10s through a working condition dynamic prediction model, and classifies the driving behavior into aggressive / flat / stable types through a driving style recognition model;

[0014] Based on the prediction results and driving style, a multi-objective prediction rolling optimization algorithm is invoked to calculate the optimal energy distribution strategy with the goals of optimal fuel economy, longest battery life and fastest power response.

[0015] Step 3: Multi-parameter thermal management adaptive adjustment. The thermal management module receives the fused dataset and, using the current component temperature, predicted future power demand, and ambient temperature as inputs, predicts the temperature evolution trend, heat generation and transfer paths of key heat source components in the power system within the same time domain through a heat flow evolution law model. Based on the above prediction results, the thermal management module drives the adaptive thermal control algorithm to generate a system regulation strategy. The regulation strategy is divided into a cold start warm-up stage, a normal operation temperature control stage, and a high-load enhanced heat dissipation stage, and performs differentiated control according to the real-time thermal state of the powertrain.

[0016] Step 4: Collaborative Control and Feedback

[0017] The collaborative control center receives energy allocation strategies and thermal management strategies, determines whether the two conflict, and corrects the strategies through collaborative decision-making units.

[0018] After receiving the energy distribution strategy, the collaborative control center determines whether there is a conflict between the strategy and the thermal management strategy. If there is no conflict, the energy distribution strategy with motor torque compensation is directly confirmed. If there is a conflict, the motor torque compensation mechanism is retained first when correcting the strategy by adjusting the engine load and maintaining the motor compensation torque.

[0019] The corrected control command, including motor torque compensation, is sent to the execution unit, and the feedback data from the execution unit is collected and updated to the fusion dataset.

[0020] Preferably, to achieve the fusion of multi-source data, the weighted fusion algorithm is as follows:

[0021] The data fusion unit of the collaborative control center performs weighted fusion of data on road conditions, driving behavior, power status, and heat source status.

[0022]

[0023] Where: F is the fused feature vector; D i Let w be the i-th type of data source; N is the number of data source types; i For the corresponding weights, satisfying ∑w i =1;

[0024] The base weights are dynamically adjusted in real time as follows:

[0025] Basic weight of traffic data: 0.3; Dynamic adjustment: urban congested road sections increased to 0.4; mountainous slope road sections increased to 0.45; highways decreased to 0.25;

[0026] Basic weighting of driving behavior data: 0.25; Dynamic adjustment: Aggressive driving recognition increased to 0.35; Emergency braking events increased to 0.4; Smooth cruise reduced to 0.2;

[0027] Basic weight of power status data: 0.25; Dynamic adjustment: increased to 0.35 when SOC is low (<20%); increased to 0.3 when demand is high; maintained at 0.25 under normal operating conditions;

[0028] Basic weight of heat source status data: 0.2; Dynamic adjustment: cold start phase: increased to 0.35; overheating risk (>85℃): increased to 0.4; normal temperature: decreased to 0.15.

[0029] In a preferred embodiment, to optimize energy management, the dynamic prediction model based on the LSTM algorithm predicts changes in vehicle speed and gradient over the next 5-10 seconds; the driving style recognition model based on the random forest algorithm classifies driving behavior into aggressive, stable, or conservative types.

[0030] The Long Short-Term Memory (LSTM) algorithm is a recurrent neural network, and its workflow is as follows:

[0031] First, the data input layer: The input data comes from the aggregated dataset, and two types of time-series feature data are extracted: historical road condition time-series data: continuous vehicle speed within the past 30-60 seconds, with a sampling frequency of 10Hz, i.e., 10 data points per second; gradient changes, collected in real time by GPS positioning device, updated every 200ms; distance to the vehicle in front, collected by millimeter-wave radar, updated every 100ms; and associated state time-series data: the state of power components strongly correlated with road conditions, including real-time motor torque and engine speed, and driving behavior including throttle opening change rate, which helps correct road condition prediction deviations, including predicting a rapid increase in vehicle speed when the throttle opening suddenly increases.

[0032] Second, the model computation layer: with the prediction time domain being 5-10 seconds in the future, the computation is completed through the following three-layer gating structure:

[0033] The forget gate determines the information to be discarded from the cell state: it decides to discard invalid information in historical data; including in urban congestion conditions, the forget gate discards high-speed driving data from 10 seconds ago and only retains low-speed fluctuation data within the last 5 seconds; in mountain slope conditions, it retains historical trend data of continuous rise or fall of slope and discards invalid slope information in flat road phases.

[0034] The input gate decides whether to store new information into the cell state: it filters and stores the effective features in the new data; when the GPS detects that the slope increases from 2% to 5% 100m ahead, the input gate converts the slope increase rate and the current vehicle speed features into updated values ​​of the cell state and integrates them into long-term memory.

[0035] The output gate, based on the current cell state, determines the next output: it generates a prediction result based on the current cell state; combining historical time series patterns and new input features, it outputs the vehicle speed-time curve and slope-time curve for the next 5-10 seconds, predicting that the vehicle speed will decrease from 35km / h to 20km / h within the next 3 seconds and the slope will increase from 3% to 6% within the next 5 seconds.

[0036] Third, the result output layer: the prediction results are directly transmitted to the MPC framework of the multi-objective prediction rolling optimization module of the energy management module, serving as the core input of the energy allocation strategy;

[0037] The random forest algorithm works as follows:

[0038] First, random sampling: a training subset is generated from the historical driving behavior database by sampling with replacement to ensure the diversity of the samples;

[0039] Second, parallel tree construction: a decision tree is trained independently for each training subset; when splitting at a node of each tree, only 3 features are randomly selected from 6 features to find the best split point. The node is split based on the throttle opening change rate > 5% / s, and the samples are divided into "aggressive candidates" and "non-aggressive candidates".

[0040] Third, the integrated results policy tree classifies the current driving behavior and outputs three types of results: aggressive, stable and conservative. Finally, the category with the most votes is taken as the final identification result.

[0041] For regression tasks, to predict a specific value, we make a prediction for every tree in the forest, and then take the average of the predictions for all trees as the final prediction result.

[0042] Based on the prediction results and driving style, a multi-objective predictive rolling optimization model based on the MPC framework is invoked to calculate the optimal energy distribution strategy with the objectives of optimal fuel economy, longest battery life, and fastest power response.

[0043] With the overall optimization objectives of minimizing total system energy loss and maximizing thermal management efficiency, the objective function J is defined as follows:

[0044]

[0045] Where: Np is the prediction time domain; Fuel(t) is the fuel consumption; BatteryDeg(t) is the battery life loss; ResponsePenalty(t) is the power response delay penalty term; α, β, and γ are the weighting coefficients, satisfying α+β+γ=1;

[0046] Stable vehicle speed, gradient <3%, SOC >40%: α=0.5, β=0.3, γ=0.2;

[0047] Rapid acceleration, gradient >5%, high torque demand: α=0.2, β=0.3, γ=0.5;

[0048] Temperature > 90℃, SOC < 15%, continuous high load: α = 0.3, β = 0.4, γ = 0.3;

[0049] In low-speed congestion conditions, when the vehicle speed is <30km / h: the electric motor takes priority, the engine stops, and braking energy is recovered;

[0050] During medium-speed cruising, 30km / h≤vehicle speed≤80km / h: the engine operates at its highest power point, and the electric motor assists in adjusting torque;

[0051] High-speed / climbing conditions, vehicle speed > 80km / h or gradient > 5%: engine outputs full load, electric motor supplements torque;

[0052] Engine intervention timing: Smooth intervention when battery SOC < 20% or required torque > motor maximum torque;

[0053] When generating the optimal energy distribution strategy, it is necessary to determine the timing of engine intervention in conjunction with real-time torque compensation of the electric motor. The electric motor is pre-charged with torque, and the compensation response time is controlled within <50ms to offset the torque fluctuation at the moment of engine intervention. At the same time, based on the motor's compensability coefficient, the torque distribution ratio between the engine and the electric motor is adjusted to optimize the overall energy distribution and ensure a smooth intervention process.

[0054] In a preferred embodiment, to achieve adaptive adjustment in thermal management, the heat flow evolution model is based on a fully connected deep neural network (FC-DNN): it constructs a nonlinear mapping between input and output through multiple fully connected layers, eliminating reliance on traditional physical mechanism formulas. This allows it to learn heat flow patterns from massive amounts of thermal state data, adapting to the thermal management needs of systems with multiple heat sources and strong coupling. The specific process is as follows:

[0055] Variable definition: Inputs include current component temperature, future power requirements, environmental and system parameters; outputs include temperature evolution trend, heat transfer path, heat load level, and anchored core heat flow parameters.

[0056] Model training: First, offline pre-training is performed using historical thermal data under all operating conditions, and the weights are optimized using gradient descent. After deployment, online fine-tuning is performed using real-time feedback data to improve fitting accuracy.

[0057] Predictive output: Real-time reception of input variables, output of component temperature changes, heat flow distribution ratio and heat load level in the next 5-10 seconds, quantitative description of heat flow evolution and qualitative assessment of risk;

[0058] Closed-loop control: The prediction results drive the adaptive thermal control algorithm to generate a strategy. After execution, the actual temperature feedback data is fed back to the model to fine-tune the parameters and optimize the next round of prediction, thus forming a closed loop.

[0059] The adaptive thermal control algorithm aims to achieve the highest temperature control accuracy deviation (<2℃), the lowest heat dissipation energy consumption, and the lowest component life loss. It optimizes the algorithm using the gradient descent method and iteratively updates the prediction weights of the heat flow evolution model and the threshold parameters of the adaptive control strategy in real time. For the thermal state characteristics under different road conditions and driving styles, the algorithm establishes a mapping relationship library between operating conditions and thermal control parameters through online training. When the same or similar operating conditions are identified in the future, the optimized control parameters can be directly called to reduce control delay.

[0060] The online training of the adaptive thermal control algorithm uses the following two types of real-time data as input to continuously optimize the control parameters:

[0061] First, thermal status feedback data: The actual component temperature, coolant flow rate and temperature data collected by heat source status sensors, including engine cylinder head temperature sensor and battery pack inlet and outlet temperature sensor, are compared with the predicted temperature of the heat flow evolution model to calculate the difference between the predicted engine cylinder block temperature and the actual temperature.

[0062] Second, control effect feedback data: collect the deviation between the actual operating parameters of the thermal management actuator and the control command, as well as energy consumption data and component life-related parameters under different control parameters;

[0063] Based on the real-time thermal state of the powertrain, the strategy divides the operating conditions into three typical stages and implements differentiated control:

[0064] First, during the cold start warm-up phase, when the engine block temperature is below 60°C, the control system rapidly increases the engine operating temperature and optimizes combustion efficiency. The actuators take coordinated actions: the electric water pump maintains a low-speed operation, limiting its flow rate to less than 5L / min; the electronic thermostat remains closed, cutting off the circulation path with the large radiator, allowing the coolant to circulate in a small loop inside the engine, using the engine's own heat to achieve rapid warm-up.

[0065] Secondly, during normal operation and temperature control, the engine block temperature is between 60°C and 90°C, and the battery temperature is within the ideal range of 25°C to 35°C: maintaining all power components within their optimal operating temperature window; the control system enters a balanced mode: the electronic water pump is upgraded to a medium-speed operation, precisely controlling the flow rate within the range of 5 to 10 L / min; the electronic thermostat adjusts proportionally based on the water temperature signal, partially opening to rationally distribute the large and small circulation flow rates; simultaneously, the system decouples and independently controls the high-temperature cooling circuit in the engine compartment from the low-temperature cooling circuits of the battery and electric drive, ensuring that each operates within its most efficient range;

[0066] Third, during the high-load enhanced cooling phase, when the engine block temperature exceeds 90°C or the battery temperature exceeds the safety threshold of 35°C: when the system faces high power output or thermal load caused by harsh environment, the strategy switches to full cooling mode; the electric water pump switches to high-speed operation, increasing the flow rate to over 10L / min; the electronic thermostat turns fully open, guiding the maximum flow through the main radiator; at the same time, the fan is activated to provide forced convection cooling based on the temperature difference between the condenser and the radiator.

[0067] In a preferred embodiment, to achieve real-time diagnostics and dynamic correction capabilities, the control system possesses real-time diagnostics and dynamic correction functions. When the temperature sensor directly detects that the battery temperature instantly exceeds the emergency threshold of 40°C, the system will immediately trigger protective intervention: controlling the output of the low-temperature circuit cooling pump to increase the flow rate of coolant through the battery pack; simultaneously, sending a power limiting request to the vehicle controller to temporarily reduce the torque output of the drive motor; this intervention signal and execution result will be synchronously fed back to the upper-level collaborative control center to trigger the rebalancing and adaptive adjustment of the global energy distribution strategy, forming a complete thermal safety closed-loop management from components to the system.

[0068] In a preferred embodiment, to achieve coordinated control and feedback, the energy distribution strategy includes a torque distribution scheme that involves engine intervention or withdrawal and mode switching.

[0069] The update cycle to the fusion dataset is 100ms per control cycle before entering the next cycle. The motor compensation torque and energy distribution ratio are continuously fine-tuned based on feedback data to ensure that the power impact is reduced by 40% when the engine intervenes or exits or the mode is switched.

[0070] Beneficial effects: This invention achieves coordinated control of energy management and thermal management, integrates data from multiple dimensions such as road conditions, driving behavior, power status, and heat source status, solves the system imbalance problem caused by single module optimization, and realizes dynamic coordination between energy management and thermal management.

[0071] By employing an LSTM dynamic prediction model and a random forest driving style recognition model, combined with the MPC multi-objective prediction rolling optimization algorithm, we can improve the adaptability to complex operating conditions and balance control accuracy and real-time performance through short-term rolling optimization.

[0072] The design incorporates a high-low temperature dual-loop system, including whole-engine reverse surround cooling, cylinder head double-layer cooling, and piston high-level internal cooling. Combined with an artificial intelligence neural network heat flow evolution model and an online-trained adaptive algorithm, it achieves differentiated control of multiple heat sources, solving the problems of slow engine preheating and difficulty in coordinating multiple heat sources. This reduces cold start time by 30% and cold start emissions by 25%.

[0073] When the engine engages / disengages or switches modes, the electric motor compensates for torque in real time to offset engine torque fluctuations. At the same time, the energy distribution is optimized based on the electric motor's compensability coefficient, overcoming the dynamic coordination problem of the hybrid powertrain and reducing power impact by 40%. Attached Figure Description

[0074] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0075] Figure 1 This is a block diagram of the overall structure of the collaborative control device of the present invention;

[0076] Figure 2 This is a schematic diagram of the high and low temperature dual-loop cooling system of the present invention;

[0077] Figure 3 This is a schematic diagram of the collaborative control method of the present invention. Detailed Implementation

[0078] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0079] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0080] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can include direct contact between the first and second features, or contact between the first and second features through another feature between them. Furthermore, "above," "over," and "on top" of the second feature includes the first feature directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature includes the first feature directly below or diagonally below the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.

[0081] like Figure 1 As shown, a collaborative control device for a hybrid power system of a new energy vehicle includes a multi-source data acquisition module, an energy management module, a thermal management module, and a collaborative control center. The modules are connected via CAN bus or Ethernet communication.

[0082] 1. Multi-source data acquisition module

[0083] The multi-source data acquisition module includes road condition sensors: a GPS locator, a millimeter-wave radar, and a camera, which can collect vehicle speed, slope, and distance to the vehicle in front in real time;

[0084] Driving behavior sensors: accelerator pedal displacement sensor and brake pedal pressure sensor, which can collect accelerator opening and braking frequency in real time;

[0085] Power component status sensors: engine speed / load sensor, motor torque / speed sensor, battery SOC / temperature sensor, which can collect engine parameters, motor parameters and battery parameters in real time;

[0086] Heat source status sensors: engine cylinder head temperature sensor, motor stator temperature sensor, battery pack inlet and outlet temperature sensor, coolant flow / temperature sensor, can collect the temperature of each component and coolant parameters in real time.

[0087] All collected data is transmitted to the collaborative control center.

[0088] 2. Energy Management Module

[0089] The energy management module includes a dynamic prediction model based on the LSTM algorithm, a driving style recognition model based on the random forest algorithm, and a multi-objective prediction rolling optimization algorithm based on the MPC framework. Based on the input from the multi-source data acquisition module, it generates an optimal energy allocation strategy, including engine / motor torque allocation, engine intervention / deactivation timing, and energy recovery intensity, and transmits this information to the collaborative control center.

[0090] 3. Thermal Management Module

[0091] The thermal management module includes a high and low temperature dual-loop cooling system (such as...). Figure 2 (As shown) and a heat flow evolution model and adaptive thermal control algorithm. High and low temperature dual-loop cooling system: The high-temperature loop connects to the engine block / cylinder head, employing a reverse-flow water jacket structure for the entire engine and a double-layer water jacket structure for the cylinder head; the low-temperature loop connects to the motor, battery, and electronic control system, using a high-position internal cooling oil passage for the piston. The two loops are connected and switched via an electronic three-way valve and are equipped with an electronic water pump, electronic thermostat, and intelligent fan. Based on heat source status data, an adaptive thermal management strategy is generated, including electronic water pump speed, electronic thermostat opening, and intelligent fan speed adjustment, which is transmitted to the collaborative control center.

[0092] 4. Coordination Control Center

[0093] The collaborative control center includes a data fusion unit and a collaborative decision-making unit. It receives raw data from the multi-source data acquisition module and uses the data fusion unit to fuse multi-source information such as road conditions, driving behavior, power status, and heat source status. Based on the fused data, it coordinates the output strategies of the energy management module and the thermal management module to generate final control commands, which are then sent to the engine controller, motor controller, battery management system, and thermal management actuators (electronic water pump, thermostat, fan).

[0094] like Figure 3 As shown, a collaborative control method for a hybrid power system in a new energy vehicle, based on the aforementioned device, includes the following steps:

[0095] 1. Data Acquisition and Fusion

[0096] S1: The multi-source data acquisition module collects real-time data on road conditions, driving behavior, power component status, and heat source status, and transmits it to the collaborative control center;

[0097] S2: The data fusion unit of the collaborative control center filters and calibrates the raw data, and integrates multi-source information through a weighted fusion algorithm to generate a fused dataset of "operating condition-driving-power-thermal state".

[0098] The collaborative control center performs weighted fusion of multi-source data (road conditions, driving behavior, power status, and heat source status):

[0099]

[0100] Where: F is the fused feature vector; D i Let w be the i-th type of data source; N is the number of data source types; i For the corresponding weights, satisfying ∑w i =1.

[0101] Based on the principle of safety > performance > economy, and with real-time dynamic adjustments, the basic weights are allocated as follows: Road condition data basic weight: 0.3 (dynamically adjusted: urban congested road sections increase to 0.4; mountainous slope sections increase to 0.45; highways decrease to 0.25); Driving behavior data basic weight: 0.25 (dynamically adjusted: aggressive driving recognition increases to 0.35; emergency braking events increase to 0.4; smooth cruise decreases to 0.2); Power status data basic weight: 0.25 (dynamically adjusted: low SOC state <20% increases to 0.35; high load demand increases to 0.3; normal operating conditions remain at 0.25); Heat source status data basic weight: 0.2 (dynamically adjusted: cold start stage: increases to 0.35; overheating risk (>85℃): increases to 0.4; normal temperature: decreases to 0.15).

[0102] 2. Energy Management Optimization

[0103] S3: The energy management module receives the fused dataset, predicts road conditions (vehicle speed, gradient changes) in the next 5-10 seconds through the dynamic prediction model of the working condition (based on the LSTM algorithm), and classifies driving behavior into aggressive / stable / conservative types through the driving style recognition model (based on the random forest algorithm).

[0104] LSTM (Long Short-Term Memory) algorithm: This is a special type of recurrent neural network capable of learning and remembering long-term dependencies, suitable for processing time series data. Its core is a concept called "cell state," which uses a "gating mechanism" to solve the long-term memory problem. Its workflow is as follows:

[0105] First, the data input layer: The input data comes from the "multi-source data acquisition module" and is fused into a "working condition-driving-power-thermal state" fusion dataset by the collaborative control center. It focuses on extracting two types of time-series feature data: historical road condition time-series data: continuous vehicle speed (sampling frequency 10Hz, i.e., 10 data points per second) within the past 30-60 seconds, gradient change (collected in real time by GPS positioning device, updated every 200ms), and distance to the vehicle in front (collected by millimeter-wave radar, updated every 100ms); and related state time-series data: the state of power components strongly correlated with road conditions (such as real-time motor torque and engine speed) and driving behavior (such as the rate of change of throttle opening), which is used to help correct road condition prediction deviations (for example, when the throttle opening suddenly increases, it can be predicted that the vehicle speed will increase rapidly).

[0106] Second, the model computation layer: using "the next 5-10 seconds" as the prediction time domain, the computation is completed through the following three-layer gating structure:

[0107] Forget Gate (determines which information to discard from the cell state): Decides to discard invalid information from historical data. For example, in urban congestion conditions, the forget gate will "forget" high-speed driving data from 10 seconds ago, retaining only low-speed fluctuation data from the last 5 seconds; in mountainous slope conditions, it will focus on retaining historical trend data of continuous rise / fall in slope, discarding invalid slope information from flat road sections;

[0108] Input gate (determines which new information to store in the cell state): filters and stores valid features from new data. For example, when GPS detects that the slope increases from 2% to 5% 100m ahead, the input gate will convert features such as "slope increase rate" and "current vehicle speed" into updated values ​​for the cell state and incorporate them into long-term memory;

[0109] Output gate (determines the next output based on the current cell state): Generates prediction results based on the current cell state. Combining historical time series patterns and new input features, it outputs the "vehicle speed-time curve" and "gradient-time curve" for the next 5-10 seconds. For example, it predicts that the vehicle speed will decrease from 35km / h to 20km / h in the next 3 seconds (increased congestion) and the gradient will increase from 3% to 6% in the next 5 seconds (the slope becomes steeper).

[0110] Third, the result output layer: the prediction results are directly transmitted to the "Multi-Objective Prediction Rolling Optimization Algorithm (MPC Framework)" of the energy management module as the core input of the energy allocation strategy.

[0111] Random Forest Algorithm: Random Forest is an ensemble learning algorithm that constructs multiple decision trees and "votes" or "averages" their predictions to obtain a more accurate and stable model. Its workflow is as follows:

[0112] First, random sampling: From the historical driving behavior database (containing 100,000+ sets of feature data of different driving styles), 200 training subsets (each subset contains 5,000 sets of data) are generated through "sampling with replacement" to ensure the diversity of the samples;

[0113] Second, parallel tree construction: a decision tree is independently trained for each training subset (200 trees in total). When splitting at a node of each tree, only 3 features are randomly selected from 6 features to find the best split point (for example, a node is divided into "aggressive candidate" and "non-aggressive candidate" based on the split condition of "throttle opening change rate > 5% / s").

[0114] Third, the integrated results: 200 decision trees classify the current driving behavior (outputting three categories: "aggressive," "moderate," and "conservative"), and the category with the most votes is used as the final identification result. For example, if 150 trees identify it as "aggressive," then the current driving style is determined to be aggressive.

[0115] For regression tasks (predicting a specific value): make a prediction for every tree in the forest, and then take the average of the predictions for all trees as the final prediction result.

[0116] S4: Based on the prediction results and driving style, the multi-objective predictive rolling optimization algorithm (based on the MPC framework) is invoked to calculate the optimal energy distribution strategy with the objectives of "optimal fuel economy, longest battery life, and fastest power response".

[0117] With the overall optimization objectives of minimizing total system energy loss (including fuel and electricity consumption, converted to equivalent values) and maximizing thermal management efficiency (such as rapid warm-up and maintaining components within the optimal temperature window), the objective function J is defined as follows:

[0118]

[0119] Where: Np: prediction time domain; Fuel(t): fuel consumption; BatteryDeg(t): battery life loss (e.g., capacity decay model); ResponsePenalty(t): power response delay penalty term; α, β, γ: weighting coefficients, satisfying α+β+γ=1

[0120] Stable vehicle speed, gradient <3%, SOC >40%: α=0.5, β=0.3, γ=0.2;

[0121] Rapid acceleration, gradient >5%, high torque demand: α=0.2, β=0.3, γ=0.5;

[0122] Temperature > 90℃, SOC < 15%, continuous high load: α = 0.3, β = 0.4, γ = 0.3.

[0123] In low-speed congestion conditions (vehicle speed < 30km / h): the electric motor takes priority, the engine stops, and braking energy is recovered;

[0124] Medium-speed cruising condition (30km / h≤vehicle speed≤80km / h): The engine operates at its high-efficiency power point (based on the universal characteristic curve), and the electric motor assists in adjusting torque;

[0125] High-speed / climbing conditions (vehicle speed > 80km / h or gradient > 5%): engine outputs full load, electric motor supplements torque;

[0126] Engine intervention timing: When the battery SOC is less than 20% or the required torque is greater than the maximum torque of the motor, the engine will smoothly intervene (motor pre-charge torque compensation to avoid shock).

[0127] When generating the optimal energy distribution strategy, it is essential to determine the timing of engine intervention in conjunction with real-time torque compensation from the electric motor. The electric motor pre-charges torque, and the compensation response time is controlled to be less than 50ms. This counteracts torque fluctuations at the moment of engine intervention, avoiding power shocks. Simultaneously, based on the motor's compensability coefficient, the torque distribution ratio between the engine and the electric motor is adjusted to optimize overall energy distribution and ensure a smooth intervention process.

[0128] 3. Thermal management adaptive adjustment

[0129] S5: The thermal management module receives the fused dataset and uses a heat flow evolution law model (based on artificial intelligence, such as a fully connected deep neural network FC-DNN) to predict the temperature evolution trend, heat generation and transfer path of key heat source components (engine, motor, battery, power electronics) of the power system in the same time domain in the future, with the current component temperature, the predicted future power demand (as heat source input) and the ambient temperature as input.

[0130] The heat flux evolution law model (based on artificial intelligence, fully connected deep neural network FC-DNN) is a "data-driven" model that uses deep neural networks as a "universal function approximator" to directly learn from massive amounts of data and construct the nonlinear mapping relationship between the input and output of complex heat flux systems, thereby achieving accurate and rapid prediction of the spatiotemporal evolution law of heat flux density.

[0131] Fully Connected Deep Neural Networks (FC-DNN): These networks construct a nonlinear mapping between input and output through multiple fully connected layers. They do not rely on traditional physical mechanisms or formulas and can learn heat flow patterns from massive amounts of thermal state data, adapting to the thermal management needs of systems with multiple heat sources and strong coupling. The specific process is as follows:

[0132] Variable definition: Inputs include current component temperature, future power requirements, and environmental and system parameters; outputs include temperature evolution trend, heat transfer path, heat load level, and anchored core heat flow parameters.

[0133] Model training: First, offline pre-training is performed using more than 100,000 sets of historical thermal data under all working conditions, and the weights are optimized using gradient descent. After deployment, online fine-tuning is performed using real-time feedback data to improve fitting accuracy.

[0134] Predictive output: Real-time reception of input variables, output of component temperature changes, heat flow distribution ratio and heat load level in the next 5-10 seconds, quantitative description of heat flow evolution and qualitative assessment of risk.

[0135] Closed-loop control: The prediction results drive the adaptive thermal control algorithm to generate a strategy. After execution, the actual temperature and other feedback data are fed back to the model to fine-tune the parameters and optimize the next round of prediction, thus forming a closed loop.

[0136] S6: Based on the above prediction results, the thermal management module drives the adaptive thermal control algorithm to generate a refined system regulation strategy. This strategy divides the powertrain's real-time thermal state into three typical operating conditions for differentiated control:

[0137] First, the cold start warm-up phase (engine block temperature below 60°C): During this phase, the control system prioritizes rapidly increasing the engine operating temperature and optimizing combustion efficiency. The actuators work in concert: the electric water pump maintains a low-speed operation, limiting its flow rate to less than 5L / min; the electronic thermostat remains closed, cutting off the circulation path to the radiator, allowing the coolant to circulate within the engine in a small loop, thus fully utilizing the engine's own heat for rapid warm-up.

[0138] Second, during the normal operating temperature control phase (engine block temperature between 60℃ and 90℃, and battery temperature within the ideal range of 25℃ to 35℃): the core task at this stage is to maintain all power components within their optimal operating temperature window. The control system enters a balanced mode: the electric water pump is upgraded to a medium-speed operation, precisely controlling the flow rate within the range of 5 to 10 L / min; the electronic thermostat adjusts proportionally based on the water temperature signal, partially opening to rationally distribute the large and small circulation flow rates. Simultaneously, the system decouples and independently controls the high-temperature cooling circuit in the engine compartment from the low-temperature cooling circuits of the battery and electric drive, ensuring that each operates within its most efficient range.

[0139] Third, the high-load enhanced cooling phase (engine block temperature exceeds 90°C, or battery temperature exceeds the safety threshold of 35°C): When the system faces high power output or thermal loads caused by harsh environments, the strategy switches to full-power cooling mode. The electric water pump switches to high-speed operation, increasing the flow rate to over 10L / min; the electronic thermostat turns fully open, guiding the maximum flow through the main radiator; at the same time, the intelligent fan is activated, providing forced convection cooling based on the temperature difference between the condenser and the radiator to maximize heat dissipation efficiency.

[0140] Furthermore, the system possesses real-time diagnostic and dynamic correction capabilities. For example, when the temperature sensor directly detects that the battery temperature instantly exceeds the emergency threshold of 40°C, the system will immediately trigger protective intervention: on the one hand, it instructs the cryogenic circuit cooling pump to increase its output, thereby increasing the coolant flow through the battery pack; on the other hand, it sends a power limiting request to the vehicle control unit (VCU) to temporarily reduce the torque output of the drive motor, reducing heat generation at the source. This intervention signal and execution result will be synchronously fed back to the upper-level collaborative control center to trigger the rebalancing and adaptive adjustment of the global energy distribution strategy, thus forming a complete thermal safety closed-loop management from components to the system.

[0141] The adaptive thermal control algorithm aims to achieve the highest temperature control accuracy (deviation <2℃), lowest heat dissipation energy consumption, and minimum component lifespan loss. It uses optimization algorithms such as gradient descent to iteratively update the prediction weights of the heat flow evolution model and the threshold parameters of the adaptive control strategy (such as the water pump speed threshold during cold start warm-up and the temperature trigger threshold during high-load heat dissipation). For different road conditions (such as urban congestion and mountain slopes) and different driving styles (such as aggressive acceleration and smooth cruising), the algorithm establishes a "working condition-thermal control parameter" mapping library through online training. When the same or similar working conditions are subsequently identified, the optimized control parameters can be directly called, reducing control latency.

[0142] The online training of the adaptive thermal control algorithm uses the following two types of real-time data as input to continuously optimize the control parameters:

[0143] First, thermal status feedback data: The actual component temperature, coolant flow / temperature, and other data collected by heat source status sensors (engine cylinder head temperature sensor, battery pack inlet and outlet temperature sensor, etc.) are compared with the predicted temperature of the heat flow evolution model to calculate the temperature deviation value (such as the difference between the predicted engine block temperature and the actual temperature).

[0144] Second, control effect feedback data: collect the deviation between the actual operating parameters (such as the actual speed of the water pump and the actual opening degree of the thermostat) of the thermal management actuators (electronic water pump, electronic thermostat, intelligent fan) and the control commands, as well as the energy consumption data under different control parameters (such as the power consumption of the water pump when running at high speed) and the component life-related parameters (such as the capacity decay rate of the battery at different temperatures).

[0145] 4. Collaborative Control and Feedback

[0146] S7: The collaborative control center receives energy allocation strategies and thermal management strategies, determines whether the two conflict (such as the balance between energy demand and heat dissipation demand under high load), and corrects the strategy through the collaborative decision-making unit (such as prioritizing heat dissipation, reducing engine load in the short term, and supplementing torque by the motor).

[0147] After receiving the energy distribution strategy (including torque distribution schemes for engine intervention / exit and mode switching), the collaborative control center determines whether there is a conflict between the strategy and the thermal management strategy. If there is no conflict, the effectiveness of the energy distribution strategy including motor torque compensation is directly confirmed; if there is a conflict (such as a contradiction between energy demand and heat dissipation demand under high load), the motor torque compensation mechanism is retained first when correcting the strategy. By adjusting the engine load in the short term and maintaining the motor compensation torque, the power smoothness during the switching process is ensured.

[0148] S8: The corrected control command, including motor torque compensation, is sent to actuators such as the engine controller and motor controller. At the same time, the actuator feedback data, including actual torque output value, power impact, and coolant temperature-related parameters, is collected and updated to the fusion dataset. The control cycle is 100ms, and the next cycle is entered. The motor compensation torque and energy distribution ratio are continuously fine-tuned according to the feedback data to ensure that the power impact is stably reduced by 40% when the engine intervenes / exits or the mode is switched.

[0149] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0150] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A collaborative control device for a hybrid power system of a new energy vehicle, comprising a multi-source data acquisition module, a collaborative control center, and an execution unit, wherein the multi-source data acquisition module transmits the acquired information data to the collaborative control center, and the collaborative control center processes and analyzes the acquired data information before issuing control commands to the execution unit, characterized in that: The collaborative control center includes a data fusion unit, an energy management module, a thermal management module, and a collaborative decision-making unit. The data fusion unit receives and processes data information from the multi-source data acquisition module. The processed data is then transmitted to the energy management module and the thermal management module. The energy management module and the thermal management module output control strategies to the collaborative decision-making unit based on the data information. The collaborative decision-making unit determines and corrects the strategy before issuing control commands to the execution unit.

2. The new energy vehicle hybrid power system coordinated control device according to claim 1, characterized in that: The multi-source data acquisition module includes a road condition sensor, a driving behavior sensor, a power component status sensor, and a heat source status sensor; the road condition sensor collects vehicle speed, gradient, and distance to the vehicle in front in real time; the driving behavior sensor collects throttle opening and braking frequency information in real time. The power component status sensor collects engine parameters, motor parameters, and battery parameters in real time; The heat source status sensor collects the temperature parameters of the engine, motor and battery as well as the coolant parameters in real time, and transmits the collected data to the collaborative control center.

3. The new energy vehicle hybrid power system coordinated control device according to claim 1, characterized in that: The energy management module includes a dynamic operating condition prediction model, a driving style recognition model, and a multi-objective prediction rolling optimization model; it generates the optimal engine / motor torque distribution scheme, engine intervention / exit timing, and energy recovery scheme based on the data fused and processed by the data fusion unit.

4. The new energy vehicle hybrid power system coordinated control device according to claim 1, characterized in that, The thermal management module includes a high and low temperature dual-loop cooling system, a heat flow evolution law model, and an adaptive thermal control module. The high-temperature circuit of the dual-circuit cooling system connects to the engine block / cylinder head and adopts a double-layer water jacket structure with reverse-flow water jackets around the entire engine and cylinder head. The low-temperature circuit connects to the motor, battery, and electronic control system and adopts a high-position internal cooling oil passage for the piston. The two circuits are connected and switched via an electronic three-way valve and are equipped with an electronic water pump, electronic thermostat, and intelligent fan. Based on the heat source status data, an adaptive thermal management strategy is generated, including an electronic water pump speed control scheme, an electronic thermostat opening scheme, and a fan speed adjustment scheme.

5. A cooperative control method for a hybrid power system in a new energy vehicle, characterized in that, Includes the following steps: Step 1: Data Acquisition and Fusion. The multi-source data acquisition module collects real-time data on road conditions, driving behavior, power component status, and heat source status, and transmits it to the data fusion unit of the collaborative control center. The data fusion unit filters and calibrates the raw data, and integrates the multi-source information through a weighted fusion algorithm to generate a fused dataset. Step 2: Multi-energy management optimization. The energy management module receives the fused dataset, predicts road conditions in the next 5-10 seconds through the dynamic prediction model, and classifies driving behavior into aggressive / stable / conservative types through the driving style recognition model. Based on the prediction results and driving style, a multi-objective prediction rolling optimization algorithm is invoked to calculate the optimal energy distribution strategy with the goals of optimal fuel economy, longest battery life and fastest power response. Step 3: Multi-parameter thermal management adaptive adjustment. The thermal management module receives the fused dataset and uses the current component temperature, predicted future power demand, and ambient temperature as inputs through the heat flow evolution law model to predict the temperature evolution trend of key heat source components of the power system, as well as the heat generation and transfer paths within the same time domain in the future. Based on the above prediction results, the thermal management module drives the adaptive thermal control algorithm to generate a system regulation strategy; the regulation strategy is divided into a cold start warm-up stage, a normal operation temperature control stage, and a high load enhanced heat dissipation stage according to the real-time thermal state of the powertrain, and performs differentiated control accordingly. Step 4: Collaborative Control and Feedback The collaborative control center receives energy allocation strategies and thermal management strategies, determines whether the two conflict, and corrects the strategies through collaborative decision-making units. After receiving the energy allocation strategy, the collaborative control center determines whether there is a conflict between the strategy and the thermal management strategy. If there is no conflict, directly confirm the energy distribution strategy that includes motor torque compensation; If a conflict exists, when correcting the strategy, the motor torque compensation mechanism will be retained first, by adjusting the engine load and maintaining the motor compensation torque. The corrected control command, including motor torque compensation, is sent to the execution unit, and the feedback data from the execution unit is collected and updated to the fusion dataset.

6. The cooperative control method for a new energy vehicle hybrid power system according to claim 5, characterized in that, The weighted fusion algorithm is as follows: The data fusion unit of the collaborative control center performs weighted fusion of data on road conditions, driving behavior, power status, and heat source status. ； Where: F is the fused feature vector; D i Let w be the i-th type of data source; N is the number of data source types; i For the corresponding weights, satisfying ∑w i =1; The base weights are dynamically adjusted in real time as follows: Basic weight of traffic data: 0.3; Dynamic adjustment: urban congested road sections increased to 0.4; mountainous slope road sections increased to 0.45; highways decreased to 0.25; Basic weighting of driving behavior data: 0.25; Dynamic adjustment: Aggressive driving recognition increased to 0.35; Emergency braking events increased to 0.4; Smooth cruise reduced to 0.2; Basic weight of power status data: 0.25; Dynamic adjustment: increased to 0.35 when SOC is low (<20%); increased to 0.3 when demand is high; maintained at 0.25 under normal operating conditions; Basic weight of heat source status data: 0.2; Dynamic adjustment: cold start phase: increased to 0.35; overheating risk (>85℃): increased to 0.4; normal temperature: decreased to 0.

15.

7. The cooperative control method for a new energy vehicle hybrid power system according to claim 5, characterized in that: The dynamic prediction model based on the LSTM algorithm predicts changes in vehicle speed and gradient over the next 5-10 seconds; the driving style recognition model based on the random forest algorithm classifies driving behavior into aggressive, stable, or conservative types. The Long Short-Term Memory (LSTM) algorithm is a recurrent neural network, and its workflow is as follows: First, the data input layer: The input data comes from the aggregated dataset, and two types of time-series feature data are extracted: historical road condition time-series data: continuous vehicle speed within the past 30-60 seconds, with a sampling frequency of 10Hz, i.e., 10 data points per second; gradient changes, collected in real time by GPS positioning device, updated every 200ms; distance to the vehicle in front, collected by millimeter-wave radar, updated every 100ms; and associated state time-series data: the state of power components strongly correlated with road conditions, including real-time motor torque and engine speed, and driving behavior including throttle opening change rate, which helps correct road condition prediction deviations, including predicting a rapid increase in vehicle speed when the throttle opening suddenly increases. Second, the model computation layer: with the prediction time domain being 5-10 seconds in the future, the computation is completed through the following three-layer gating structure: The forget gate determines the information to be discarded from the cell state: it decides to discard invalid information in historical data; including in urban congestion conditions, the forget gate discards high-speed driving data from 10 seconds ago and only retains low-speed fluctuation data within the last 5 seconds; in mountain slope conditions, it retains historical trend data of continuous rise or fall of slope and discards invalid slope information in flat road phases. The input gate decides whether to store new information into the cell state: it filters and stores the effective features in the new data; when the GPS detects that the slope increases from 2% to 5% 100m ahead, the input gate converts the slope increase rate and the current vehicle speed features into updated values ​​of the cell state and integrates them into long-term memory. The output gate, based on the current cell state, determines the next output: it generates a prediction result based on the current cell state; combining historical time series patterns and new input features, it outputs the vehicle speed-time curve and slope-time curve for the next 5-10 seconds, predicting that the vehicle speed will decrease from 35km / h to 20km / h within the next 3 seconds and the slope will increase from 3% to 6% within the next 5 seconds. Third, the result output layer: the prediction results are directly transmitted to the MPC framework of the multi-objective prediction rolling optimization module of the energy management module, serving as the core input of the energy allocation strategy; The random forest algorithm works as follows: First, random sampling: a training subset is generated from the historical driving behavior database by sampling with replacement to ensure the diversity of the samples; Second, parallel tree construction: a decision tree is trained independently for each training subset; when splitting at a node of each tree, only 3 features are randomly selected from 6 features to find the best split point. The node is split based on the throttle opening change rate > 5% / s, and the samples are divided into "aggressive candidates" and "non-aggressive candidates". Third, the integrated results policy tree classifies the current driving behavior and outputs three types of results: aggressive, stable and conservative. Finally, the category with the most votes is taken as the final identification result. For regression tasks, to predict a specific value, we make a prediction for every tree in the forest, and then take the average of the predictions for all trees as the final prediction result. Based on the prediction results and driving style, a multi-objective predictive rolling optimization model based on the MPC framework is invoked to calculate the optimal energy distribution strategy with the objectives of optimal fuel economy, longest battery life, and fastest power response. With the overall optimization objectives of minimizing total system energy loss and maximizing thermal management efficiency, the objective function J is defined as follows: ； Where: Np is the prediction time domain; Fuel(t) is the fuel consumption; BatteryDeg(t) is the battery life loss; ResponsePenalty(t) is the power response delay penalty term; α, β, and γ are the weighting coefficients, satisfying α+β+γ=1; Stable vehicle speed, gradient <3%, SOC >40%: α=0.5, β=0.3, γ=0.2; Rapid acceleration, gradient >5%, high torque demand: α=0.2, β=0.3, γ=0.5; Temperature > 90℃, SOC < 15%, continuous high load: α = 0.3, β = 0.4, γ = 0.3; In low-speed congestion conditions, when the vehicle speed is <30km / h: the electric motor takes priority, the engine stops, and braking energy is recovered; During medium-speed cruising, 30km / h≤vehicle speed≤80km / h: the engine operates at its highest power point, and the electric motor assists in adjusting torque; High-speed / climbing conditions, vehicle speed > 80km / h or gradient > 5%: engine outputs full load, electric motor supplements torque; Engine intervention timing: Smooth intervention when battery SOC < 20% or required torque > motor maximum torque; When generating the optimal energy distribution strategy, it is necessary to determine the timing of engine intervention in conjunction with real-time torque compensation of the electric motor. The electric motor is pre-charged with torque, and the compensation response time is controlled within <50ms to offset the torque fluctuation at the moment of engine intervention. At the same time, based on the motor's compensability coefficient, the torque distribution ratio between the engine and the electric motor is adjusted to optimize the overall energy distribution and ensure a smooth intervention process.

8. The cooperative control method for a new energy vehicle hybrid power system according to claim 5, characterized in that: The heat flow evolution model is based on a fully connected deep neural network (FC-DNN): it constructs a nonlinear mapping between input and output through multiple fully connected layers, without relying on traditional physical mechanism formulas, and can learn heat flow laws from massive amounts of thermal state data, adapting to the thermal management needs of systems with multiple heat sources and strong coupling; the specific process is as follows: Variable definition: Inputs include current component temperature, future power requirements, environmental and system parameters; outputs include temperature evolution trend, heat transfer path, heat load level, and anchored core heat flow parameters. Model training: First, pre-train offline using historical hot data under all operating conditions, and then optimize the weights using gradient descent. After deployment, online fine-tuning is performed using real-time feedback data to improve fitting accuracy; Predictive output: Real-time reception of input variables, output of component temperature changes, heat flow distribution ratio and heat load level in the next 5-10 seconds, quantitative description of heat flow evolution and qualitative assessment of risk; Closed-loop control: The prediction results drive the adaptive thermal control algorithm to generate a strategy. After execution, the actual temperature feedback data is fed back to the model to fine-tune the parameters and optimize the next round of prediction, thus forming a closed loop. The adaptive thermal control algorithm aims to achieve the highest temperature control accuracy deviation (<2℃), the lowest heat dissipation energy consumption, and the lowest component life loss. It optimizes the algorithm using the gradient descent method and iteratively updates the prediction weights of the heat flow evolution model and the threshold parameters of the adaptive control strategy in real time. For the thermal state characteristics under different road conditions and driving styles, the algorithm establishes a mapping relationship library between operating conditions and thermal control parameters through online training. When the same or similar operating conditions are identified in the future, the optimized control parameters can be directly called to reduce control delay. The online training of the adaptive thermal control algorithm uses the following two types of real-time data as input to continuously optimize the control parameters: First, thermal status feedback data: The actual component temperature, coolant flow rate and temperature data collected by heat source status sensors, including engine cylinder head temperature sensor and battery pack inlet and outlet temperature sensor, are compared with the predicted temperature of the heat flow evolution model to calculate the difference between the predicted engine cylinder block temperature and the actual temperature. Second, control effect feedback data: collect the deviation between the actual operating parameters of the thermal management actuator and the control command, as well as energy consumption data and component life-related parameters under different control parameters; Based on the real-time thermal state of the powertrain, the strategy divides the operating conditions into three typical stages and implements differentiated control: First, during the cold start warm-up phase, when the engine block temperature is below 60°C, the control system rapidly increases the engine operating temperature and optimizes combustion efficiency. The actuators take coordinated actions: the electric water pump maintains a low-speed operation, limiting its flow rate to less than 5L / min; the electronic thermostat remains closed, cutting off the circulation path with the large radiator, allowing the coolant to circulate in a small loop inside the engine, using the engine's own heat to achieve rapid warm-up. Secondly, during normal operation and temperature control, the engine block temperature is between 60°C and 90°C, and the battery temperature is within the ideal range of 25°C to 35°C: maintaining all power components within their optimal operating temperature window; the control system enters a balanced mode: the electronic water pump is upgraded to a medium-speed operation, precisely controlling the flow rate within the range of 5 to 10 L / min; the electronic thermostat adjusts proportionally based on the water temperature signal, partially opening to rationally distribute the large and small circulation flow rates; simultaneously, the system decouples and independently controls the high-temperature cooling circuit in the engine compartment from the low-temperature cooling circuits of the battery and electric drive, ensuring that each operates within its most efficient range; Third, during the high-load enhanced cooling phase, when the engine block temperature exceeds 90°C or the battery temperature exceeds the safety threshold of 35°C: when the system faces high power output or thermal load caused by harsh environment, the strategy switches to full cooling mode; the electric water pump switches to high-speed operation, increasing the flow rate to over 10L / min; the electronic thermostat turns fully open, guiding the maximum flow through the main radiator; at the same time, the fan is activated to provide forced convection cooling based on the temperature difference between the condenser and the radiator.

9. The collaborative control method for a new energy vehicle hybrid power system according to claim 8, characterized in that: The control system has real-time diagnostic and dynamic correction functions. When the temperature sensor directly detects that the battery temperature instantly exceeds the emergency threshold of 40°C, the system will immediately trigger protective intervention: control the low-temperature circuit cooling pump to increase the output and increase the flow of coolant through the battery pack; at the same time, send a power limiting request to the vehicle controller to temporarily reduce the torque output of the drive motor; this intervention signal and execution result will be synchronously fed back to the upper-level collaborative control center to trigger the rebalancing and adaptive adjustment of the global energy distribution strategy, forming a complete thermal safety closed-loop management from components to the system.

10. The cooperative control method for a new energy vehicle hybrid power system according to claim 5, characterized in that: The energy distribution strategy includes torque distribution schemes for engine intervention or deactivation and mode switching. The update cycle to the fusion dataset is 100ms per control cycle before entering the next cycle. The motor compensation torque and energy distribution ratio are continuously fine-tuned based on feedback data to ensure that the power impact is reduced by 40% when the engine intervenes or exits or the mode is switched.

Images