Range extension-battery power cooperative distribution method based on mine truck working condition

Abstract

The invention discloses a range extending-battery power cooperative distribution method based on a mine truck working condition, and the method comprises the following steps: S1, collecting mine truck operation parameters in real time, including a gradient, a load coefficient, a vehicle speed, a battery remaining capacity SOC and a battery temperature; s2, judging the real-time operation condition of the mine truck, and performing standardized pretreatment on the operation parameters of the mine truck according to the real-time operation condition of the mine truck to obtain pretreated parameters; s3, enabling the preprocessed parameters to form an input vector, inputting the input vector into the trained neural network model, and obtaining the discharge power Pd and the charge power Pc of the battery; and S4, predicting a feature sequence of a future working condition through the trained neural network model, redistributing the discharge power Pd and the charge power Pc of the battery and the power of the range extender according to the feature sequence of the future working condition, and improving the power adaptation capability of the mine truck under the dynamic working condition.

Description

Technical Field

[0001] This invention belongs to the field of hybrid vehicle technology, and particularly relates to a range extender-battery power collaborative distribution method based on mining truck operating conditions. Background Technology

[0002] Hybrid mining trucks have significant application value in the mining transportation sector. Their power systems typically consist of a range extender and a power battery working together to achieve efficient and stable energy output. The range extender provides continuous and stable power output, suitable for long-term operation scenarios; the power battery has rapid response characteristics, capable of instantly replenishing or absorbing instantaneous power demands. Together, they meet the vehicle's power requirements under complex operating conditions through a power coordination mechanism. In existing technologies, hybrid mining trucks generally employ a power management strategy with a fixed State of Charge (SOC) range. By setting a SOC threshold range for the battery system, the range extender's output power is controlled, allowing the power battery to follow the SOC within a specific range to adapt to power fluctuations during vehicle operation.

[0003] Therefore, it is urgent to design a range extender-battery power collaborative allocation method based on mining truck operating conditions to solve the above-mentioned problems. Summary of the Invention

[0004] The purpose of this invention is to provide a range extender-battery power collaborative allocation method based on mining truck operating conditions, which expands the SOC operating range of the battery system and improves the power adaptability of mining trucks under dynamic operating conditions.

[0005] To achieve the above objectives, the specific technical solution of the range extender-battery power collaborative allocation method based on mining truck operating conditions is as follows: A method for range extender-battery power collaborative allocation based on mining truck operating conditions includes the following steps: S1. Real-time acquisition of mining truck operating parameters, including slope. Load factor Speed Remaining battery charge (SOC) and battery temperature ,in, For real-time load, Maximum load capacity; S2. Determine the real-time operating condition of the mining truck, and perform standardized preprocessing on the operating parameters of the mining truck based on the real-time operating condition of the mining truck to obtain the preprocessed parameters. S3. The preprocessed parameters are used to form an input vector, which is then input into the trained neural network model to obtain the battery's discharge power P. d and charging power P c ; S4. Predict the feature sequence of future operating conditions using the trained neural network model, and adjust the discharge power P of the battery based on the feature sequence of the future operating conditions. d and charging power P c And the power of the range extender is redistributed.

[0006] Further, in step S2, determining the real-time operating condition of the mining truck includes the following steps: when This is considered the initial operating condition; when The condition was determined to be flat ground; when The condition is classified as uphill; when The condition is classified as a downhill slope; when The condition was determined to be a steep slope; when Determined to be in full-load condition. For full load weight; when Determined to be in no-load condition. For the weight of the vehicle; when The condition was determined to be overloaded.

[0007] Furthermore, a preset battery system request mode is matched according to the current operating condition type, and the corresponding SOC working range for the remaining battery capacity is dynamically set: The no-load downhill mode corresponds to the remaining battery charge SOC ∈ [40%, 60%]; The no-load uphill mode corresponds to the remaining battery charge SOC ∈ [60%, 70%]; The full-load uphill mode corresponds to the remaining battery charge SOC ∈ [70%, 75%]; The full-load downhill mode corresponds to the remaining battery charge SOC ∈ [50%, 60%]; The outbound no-load overall adaptation of the battery's remaining SOC ∈ [40%, 70%]; The remaining SOC of the battery during the return trip at full load is ∈ [65%, 75%]. When a steep slope condition is detected and When the steep slope is steep, adaptive control is implemented: the remaining battery charge SOC is anchored to 75%~85% 500 meters in advance.

[0008] Furthermore, slope weights are dynamically configured based on the working condition type. Load Capacity Vehicle speed weight , .

[0009] Furthermore, step S3, which involves constructing an input vector from the preprocessed parameters, includes the following steps: Calculate the weighted fusion feature vector ; The weighted fusion feature vector The remaining battery charge (SOC) and the battery temperature This forms a 5-dimensional input vector.

[0010] Furthermore, the structure of the trained neural network model in step S3 is as follows: The input layer receives the 5-dimensional input vector; The hidden layer contains three layers of bidirectional long short-term memory network, with 64 neurons in each layer; An attention mechanism layer is connected after the output of the bidirectional long short-term memory network. The output layer is a 2D linear layer that directly outputs the battery's discharge power P. d and charging power P c .

[0011] Furthermore, the training of the neural network model in step S3 employs a weighted multi-objective loss function: .

[0012] in, For multi-objective loss functions; For energy loss, This is the energy loss weighting coefficient. Due to battery life loss, This is the weighting coefficient for battery life loss. For dynamic loss, This is the weighting coefficient for dynamic loss.

[0013] Furthermore, during the overload condition, the audible and visual alarm is triggered, and a speed limit command is sent to the vehicle controller.

[0014] Furthermore, when the battery temperature At that time, the mining card is preheated until .

[0015] Furthermore, when the battery temperature At the same time, the proportion of battery power output is reduced, while the proportion of range extender power is increased.

[0016] The range extender-battery power coordinated allocation method based on mining truck operating conditions of the present invention has the following advantages: This scheme collects mining truck operating parameters in real time, including slope. Load factor Speed Remaining battery charge (SOC) and battery temperature This achieves comprehensive perception of the vehicle's operating status. Based on the determined real-time operating conditions of the mining truck, parameters are standardized and preprocessed to ensure the accuracy and consistency of the input data. By constructing an input vector from the preprocessed parameters and feeding it into a trained neural network model, the battery's discharge power P is accurately output. d and charging power P c This design solves the problem that traditional fixed-power strategies cannot adapt to dynamic operating conditions. Furthermore, by using a neural network model to predict future operating condition characteristic sequences, the power allocation ratio between the battery and the range extender is dynamically adjusted, enabling the system to respond to changes in operating conditions in advance. This design improves the adaptability of the battery system in high-power demand scenarios for mining trucks. By dynamically setting the SOC (State of Charge) operating range corresponding to different operating conditions, the effective operating range of the battery system is significantly expanded, and the slope weight is dynamically configured based on the operating condition type. Load Capacity Vehicle speed weight This makes the parameter fusion more closely match the characteristics of actual working conditions. Calculate the weighted fusion feature vector. A 5-dimensional input vector was constructed to improve the representativeness of the model input. A neural network structure including a 3-layer bidirectional long short-term memory network and an attention mechanism was adopted to effectively capture the time-series features and key information of the operating conditions. A weighted multi-objective loss function was used to balance energy efficiency, battery life, and power performance to ensure optimal overall system performance. This scheme ultimately achieved a reasonable extension of the battery system's SOC operating range, avoiding the power regulation limitation problem caused by traditional narrow-range SOC control. Attached Figure Description

[0017] Figure 1 This is a process flow diagram of the range extender-battery power collaborative allocation method based on mining truck operating conditions according to the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0019] Those skilled in the art will understand that although some embodiments herein include certain features included in other embodiments but not others, combinations of features from different embodiments are intended to be within the scope of the invention and form different embodiments. For example, in the claims, any of the claimed embodiments can be used in any combination.

[0020] The following is a reference to the appendix. Figure 1 This invention describes a range extender-battery power collaborative allocation method based on mining truck operating conditions.

[0021] like Figure 1 As shown, a range extender-battery power collaborative allocation method based on mining truck operating conditions includes the following steps: S1. Real-time acquisition of mining truck operating parameters, including slope. Load factor Speed Remaining battery charge (SOC) and battery temperature ,in, For real-time load, Maximum load capacity; in, The longitudinal tilt angle of the mining truck's current travel segment is measured in real time by an onboard tilt sensor, and its unit is degrees, used to characterize the degree of road undulation; load factor. For real-time load With the maximum load The ratio is a dimensionless scalar used to reflect the current load level; vehicle speed The instantaneous linear velocity of the mining truck wheel edge is used to characterize the motion state; the remaining battery charge (SOC) is the percentage of the current usable capacity of the power battery relative to the nominal total capacity, ranging from 0% to 100%, used to characterize the battery's energy reserve state; battery temperature... The average temperature of the battery module is used to characterize the battery's thermal state. These parameters together constitute the set of fundamental physical quantities reflecting the real-time operating status of the mining truck, providing the initial input basis for subsequent operating condition identification and power decision-making.

[0022] Optionally, slope The method for obtaining the pitch angle signal is as follows: a dual-axis tilt sensor installed in the middle of the longitudinal beam of the vehicle frame is used to collect the pitch angle signal, and the signal is then filtered by Kalman filter to eliminate vibration interference before outputting a stable signal. value.

[0023] Optionally, load factor The method for obtaining the data is as follows: based on the pressure sensor array installed in the suspension system, the sprung masses at the four corners are measured and weighted summed, and then compared with... Perform ratio calculations.

[0024] Furthermore, vehicle speed SOC and The acquisition method is as follows: The corresponding message fields are periodically read from the standard CAN bus interfaces of the vehicle controller (VCU) and battery management system (BMS) respectively, with a sampling period of 100ms.

[0025] S2. Determine the real-time operating condition of the mining truck, and perform standardized preprocessing on the operating parameters of the mining truck based on the real-time operating condition of the mining truck to obtain the preprocessed parameters. Specifically, determining the real-time operating condition of the mining truck includes the following steps: Based on the multi-dimensional parameter combinations collected above, identify the typical operating scenario category currently in operation, including but not limited to basic operating condition types such as starting, flat ground, uphill, downhill, steep slope, empty load, full load, and overload; Standardization preprocessing refers to mapping the original parameters to a unified numerical scale, eliminating differences in dimensions and orders of magnitude, and ensuring comparability and weight fairness of different physical quantities in the model input. This preprocessing process does not change the relative relationships between the parameters, but only normalizes the numerical expression.

[0026] Optionally, the standardization preprocessing method is to perform Min-Max normalization on each parameter separately, that is, to calculate (SS) for parameter S. min ) / (S max -S min ), where S min With S max These are the minimum and maximum values ​​of this parameter obtained statistically from the historical running dataset.

[0027] Optionally, the standardization preprocessing method is to use Z-score standardization, that is, to calculate (S-μ) / σ for parameter S, where μ and σ are the mean and standard deviation of the parameter in the calibration condition database, respectively.

[0028] Optionally, the standardized preprocessing method involves setting differentiated normalization intervals for different working condition subclasses, such as for the slope in uphill working conditions. The normalization reference range is [0°, 15°], while [-15°, 0°] is used for downhill conditions to enhance the model's ability to model directional sensitivity.

[0029] S3. The preprocessed parameters are used to form an input vector, which is then input into the trained neural network model to obtain the battery's discharge power P. d and charging power P c ; Specifically, the preprocessed parameters forming the input vector refer to concatenating the normalized parameters of the above output into a fixed-dimensional numerical vector in a predetermined order; the trained neural network model refers to a deep learning model that has completed offline training and online verification, whose structure and parameters are fixed, and which has the ability to non-linearly map from the input vector to the battery power command; the battery's discharge power P d The active power output by the battery to the drive motor or auxiliary load is called the charging power P. A value greater than zero indicates discharge. cThis refers to the active power absorbed by the battery from the regenerative braking or the rectifier side of the range extender; a value greater than zero indicates charging. This step achieves end-to-end mapping from multi-source sensing information to specific power commands.

[0030] S4. Predict the feature sequence of future operating conditions using the trained neural network model, and adjust the discharge power P of the battery based on the feature sequence of the future operating conditions. d and charging power P c And the power of the range extender is redistributed.

[0031] Specifically, predicting the feature sequence of future operating conditions refers to using the built-in time modeling capabilities of neural network models (such as RNN / LSTM / Transformer structures) to infer the changing trends of key parameters such as slope, load, and vehicle speed within several future time steps based on current and historical input vectors, generating a predictive sequence of operating condition features with a time dimension. Reallocation refers to dynamically adjusting the power sharing ratio between the battery and range extender based on the initial power allocation results and the prediction of future road energy demand, so that the overall energy flow tends to be optimal in the time dimension. This step overcomes the limitations of traditional static power allocation and achieves forward-looking energy management for future needs.

[0032] Optionally, the prediction method for the future working condition feature sequence is as follows: the hidden layer state of the above neural network model is used as the initial memory unit, and the model is recursively expanded k steps (k=3~5). At each step, the predicted 5-dimensional parameter values ​​for the next time step are output, forming a feature sequence of length k. Optionally, the prediction method for future working condition feature sequences is to introduce external prior knowledge, such as slope profile data and work cycle path planning information in the mine electronic map, and to perform weighted fusion with the model prediction results to improve the accuracy of long-term prediction.

[0033] Furthermore, the redistribution method is as follows: with the expected power demand at each moment in the predicted sequence as constraints, a rolling optimization objective function is constructed, and the optimal power allocation trajectory in the next N control cycles is solved under the premise of satisfying the battery SOC safety boundary, temperature rise limit and range extender power capability.

[0034] This application constructs physically meaningful input vectors by real-time acquisition of multi-dimensional operating parameters, combined with operating condition identification and standardized preprocessing. A trained neural network model is then used to intelligently map perceived information to battery power commands. Furthermore, the model's temporal modeling capabilities are utilized to predict future operating condition changes, and the power output of the battery and range extender is redistributed in a rolling manner accordingly. This avoids the battery capacity limitations caused by traditional fixed SOC range strategies and overcomes the instantaneous power gap caused by the range extender system's response lag. Thus, while ensuring the vehicle's overall power performance, it significantly slows down battery aging, improves energy utilization efficiency, and enhances the system's adaptability to complex mining operation scenarios.

[0035] Further, in step S2, determining the real-time operating condition of the mining truck includes the following steps: when This is considered the initial operating condition; when The condition was determined to be flat ground; when The condition is classified as uphill; when The condition is classified as a downhill slope; when The condition was determined to be a steep slope; when Determined to be in full-load condition. For full load weight; when Determined to be in no-load condition. For the weight of the vehicle; when The condition was determined to be overloaded.

[0036] Specifically, when =0 is considered the starting condition; among which... The real-time speed of the mining truck is collected by a non-contact speed sensor installed on the drive wheel or drive shaft, with units of km / h or m / s, used to characterize the instantaneous motion state of the vehicle. The starting condition is a transitional operating state where the mining truck enters the acceleration phase from a stationary state. Its technical function is to trigger the initial response of the power request strategy. In this state, the battery system must prioritize providing instantaneous high torque drive power to overcome the vehicle's inertia and rolling resistance. The range extender starts synchronously and gradually builds up output power, thereby avoiding power interruption. =0 is used as the criterion, without introducing a time window or acceleration threshold, to ensure that the judgment logic is simple, the response is timely, and there is no overlap or dead zone with subsequent working condition transitions.

[0037] Specifically, The site was classified as flat ground. The current road slope angle of the mining truck is measured in real time by an onboard tilt sensor, in degrees, reflecting the degree of inclination of the vehicle's longitudinal reference plane relative to the horizontal plane. This threshold is determined based on the design specifications of typical open-pit mine transportation roads and statistical analysis of actual working conditions, covering most gentle slopes and near-horizontal road sections. The technical function of this condition is to define the low dynamic power demand range: within this range, the vehicle's driving force mainly overcomes rolling resistance and air resistance, resulting in small power fluctuations. The range extender can operate stably in the high-efficiency range, and the battery system only needs to make small power follow-up adjustments, which is conducive to maintaining stable SOC. This criterion is independent of load and speed parameters, constituting the first level of coarse layering for working condition identification, providing basic classification anchor points for subsequent subdivision.

[0038] Specifically, The condition was determined to be an uphill driving condition. After excluding flat terrain conditions, this technology specifically identifies steep inclines. Its technical function is to identify operating scenarios requiring continuous high power input: during uphill driving, the vehicle needs to overcome gravity, significantly increasing drive power demand. At this time, the battery system should participate in discharging to replenish power, while the range extender increases output power to maintain vehicle speed. This criterion is consistent with the aforementioned... A symmetrical logical structure is formed to ensure that the boundaries of uphill and downhill working conditions are mutually exclusive and fully covered, avoiding working condition fluctuations caused by sensor noise or data jumps.

[0039] Specifically, when The condition is classified as a downhill condition, indicating that the road slope angle is less than -8°, i.e., it has a significant negative inclination angle. The technical function of this feature is to identify energy feedback-dominated conditions: when going downhill, the vehicle's potential energy is converted into kinetic energy, the braking system or motor is in regenerative braking mode, the battery system needs to have a high power absorption capacity to recover energy, and at the same time suppress the ineffective start-stop of the range extender. This criterion, together with the above, constitutes the slope aspect bipolar criterion, and its threshold (±8°) is strictly connected with the flat ground threshold (|α|≤8°) mentioned above, ensuring seamless division of the three types of slope conditions (flat ground / uphill / downhill) in numerical space.

[0040] Specifically, when The condition is classified as a steep slope condition; it is an independent and enhanced identification of extreme slopes, belonging to the superimposed criteria based on the above. The steep slope condition is not a new condition type independent of uphill or downhill slopes, but rather a condition related to... or The subset with larger slope amplitude is relabeled; the technical function of this feature is to trigger adaptive feedforward control: when it detects... In this case, the system anticipates subsequent high power demand or high energy feedback intensity, and then adjusts the battery SOC target range, range extender preheating status, or transportation management system (TMS) cooling strategy to improve control foresight. This criterion does not replace the original uphill and downhill judgment, but adds a safety redundancy level on top of it to enhance the system's robust response capability to steep sections of the mine (such as the edge of the spoil heap and the crushing station approach road).

[0041] Specifically, when Determined to be in full-load condition. The real-time load of the mining truck is indirectly calculated by the on-board weighing sensor or hydraulic pressure sensor, and the unit is tons. The maximum rated load capacity specified for this vehicle model is a legal parameter for vehicle design and safe operation. The technical function of this feature is to distinguish the changes in dynamic response caused by the difference in the inertia of the whole vehicle mass: under full load, the driving force required for the same slope or acceleration is greater, the braking feedback energy is higher, and the battery system needs to be matched with a higher power level of charging and discharging capacity. As a hard threshold criterion, no tolerance band or sliding window is introduced to ensure the deterministic nature of the full-load identification result, supporting the load factor in the subsequent power allocation model. Accurate normalization processing.

[0042] Specifically, Determined to be in no-load condition. The curb weight of a mining truck, i.e. the weight of the entire vehicle without cargo, is the technical feature that defines the boundary of light-load operation: when unloaded, the vehicle has low inertia and low resistance, and the requirements for instantaneous power response of the battery are relatively low, but the driving range is long, making it suitable for a wide SOC range strategy. and The two load limits are determined by a transition range, which defaults to a more conservative control strategy used in both unloaded and fully loaded conditions to avoid aggressive control.

[0043] Specifically, The condition was determined to be overloaded. The maximum permissible load under the current operating license, which is equal to or slightly greater than the specified value. , It is dynamically issued by the mine dispatching system or the vehicle controller.

[0044] The technical function of this feature is to achieve safety red line monitoring: overload will significantly aggravate tire wear, brake fade and battery thermal stress, triggering the protection mechanism; Criterion independent of This reflects the management distinction between permitted load and designed load, and supports differentiated operating procedures in mines (such as load limits during the rainy season and reduced load on older vehicles); its triggering actions are not included in this scope, but provide a preliminary judgment basis for the audible and visual alarms and speed limit commands in this application.

[0045] This application constructs a system based on the above-mentioned eight-step hierarchical, mutually exclusive, and programmable logical judgment process. , , The real-time operating conditions are input from three core physical quantities; each step follows a priority order of speed, then gradient, and finally load to ensure that a unique operating condition label can still be output in multi-parameter coupled scenarios; all criteria adopt static threshold form, without relying on historical data or sliding windows, to ensure that the embedded controller can execute in real time; on this basis, different operating condition labels are directly mapped to the battery SOC working range defined in this application and the weight configuration scheme defined in this application, thereby realizing end-to-end closed-loop control from operating condition identification to power collaborative allocation.

[0046] Furthermore, a preset battery system request mode is matched according to the current operating condition type, and the corresponding SOC working range for the remaining battery capacity is dynamically set: The no-load downhill mode corresponds to the remaining battery charge SOC ∈ [40%, 60%]; Among them, the no-load downhill mode refers to the condition in which the working condition is determined to be no-load and The operating state; in the relevant technical field, this mode represents the state in which the mining truck is unloaded and traveling downhill along the slope, with its typical energy flow being the conversion of kinetic energy into electrical energy, and the battery system mainly undertaking the task of energy recovery; In this embodiment, the lower limit of SOC is allowed to drop to 40% in this mode. On the one hand, this provides sufficient discharge margin for possible uphill or start-up phases, and on the other hand, it avoids the SOC from rising rapidly to the upper limit due to continuous feedback, thus triggering the charge limit protection and ensuring the continuity of energy recovery.

[0047] The State of Charge (SOC) of a battery is the percentage of its current usable capacity relative to its rated capacity, used to characterize the battery's state of charge. In this embodiment, it is constrained to the range of [40%, 60%], meaning that the control system actively suppresses overcharging of the battery under this condition, while retaining no less than 40% of its discharge capacity to cope with sudden acceleration or gradient changes. This range setting does not rely on fixed threshold judgment, but rather on the vehicle controller to monitor the SOC change trend in real time, and to activate the SOC adjustment strategy before entering an unloaded downhill section, so that it smoothly falls into the target range.

[0048] Optionally, the SOC range maintenance method may be: based on the deviation of the current SOC value from the target range midpoint by 50%, adjust the range extender output power and braking feedback intensity proportionally to bring the SOC to the center of the range. Optionally, the SOC range maintenance method may include: when SOC > 60%, reducing the regenerative braking ratio and simultaneously increasing the range extender's power generation to consume excess energy; when SOC < 40%, suspending regenerative braking or enabling low-power range extender charging to prevent SOC from decreasing further.

[0049] Optionally, the SOC interval maintenance method can also adopt a segmented closed-loop regulation strategy: dividing [40%, 60%] into three sub-intervals: [40%, 45%] is the low SOC response zone, enabling high-priority discharge permission; [45%, 55%] is the steady-state maintenance zone, maintaining the balance between regular feedback and discharge; [55%, 60%] is the high SOC early warning zone, gradually limiting the feedback depth and enhancing power consumption load scheduling.

[0050] The no-load uphill mode corresponds to the remaining battery charge SOC ∈ [60%, 70%]; Among them, the no-load uphill mode refers to the condition in which the working condition is determined to be no-load and The operating state; in the relevant technical field, this mode represents the state in which the mining truck is unloaded and traveling uphill along the slope. Its typical energy requirement is to provide continuous supplementary power to the drive motor, and the battery system needs to have a stable medium power output capability. In this embodiment, the SOC range is shifted to [60%, 70%] in this mode, which avoids the risk of instantaneous power decay caused by low SOC, and does not excessively increase the initial reserve, so as not to affect the subsequent downhill energy recovery space.

[0051] Optionally, the SOC range setting method may be: after identifying an unloaded uphill mode, the controller dynamically adjusts the maximum allowable discharge current limit of the battery based on the difference between the current SOC value and the lower limit of the target range of 60%, so that the SOC slowly rises to around 65% and is maintained. Optionally, the SOC range setting method may include: combining vehicle speed With slope Calculate the theoretical ramp power requirement. If it is predicted that the power requirement will exceed the range extender's continuous output capability within the next 10 seconds, then adjust the SOC to the upper limit of the range in advance (70%). Furthermore, the SOC interval setting method can also adopt a feedforward-feedback coordinated adjustment mode: the slope change rate dα / dt obtained by the feedforward path is used as the SOC pre-adjustment trigger signal, and the actual SOC trajectory is corrected in real time by the feedback path to ensure that it is always within [60%, 70%].

[0052] The full-load uphill mode corresponds to the remaining battery charge SOC ∈ [70%, 75%]; Among them, the full-load uphill mode refers to the condition that is determined to be full-load and The operating state; in the relevant technical field, this mode represents the state in which the mining truck is under maximum load and traveling uphill along a slope. Its typical energy characteristics are high peak drive power demand and long duration, which puts strict requirements on the instantaneous and continuous discharge capabilities of the battery. In this embodiment, the SOC range in this mode is further increased to [70%, 75%], which aims to ensure that the battery has sufficient high-power discharge reserves to support the power continuity and response margin of the whole vehicle during heavy-load climbing.

[0053] Optionally, the SOC range setting method can be as follows: before entering the full-load uphill mode, the controller can activate the SOC pre-rise strategy in advance based on the path planning information or the slope prediction result, and moderately supplement the power of the range extender or reduce the power consumption of unnecessary loads so that the SOC reaches 70% before reaching the bottom of the slope; Optionally, the SOC range setting method may include: when a detection is made... and Immediately freeze the current SOC adjustment logic, allowing only upward adjustment to 75%, and prohibit any regenerative charging behavior; Furthermore, the SOC range setting method can also adopt a dual threshold locking mechanism: 70% is used as the start threshold to trigger the SOC rise action; 75% is used as the hard upper limit, once reached, all charging paths are cut off and the depth of discharge is limited to no more than 85% of the rated capacity to prevent over-discharge.

[0054] The full-load downhill mode corresponds to the remaining battery charge SOC ∈ [50%, 60%]; Among them, the full-load downhill mode refers to the condition that is determined to be full-load and The operating state; in the relevant technical field, this mode represents the state in which the mining truck is under maximum load and traveling downhill along an incline. Its typical energy characteristics are high braking regenerative power and long duration, which can easily lead to increased battery temperature and rapid rise in SOC. In this embodiment, the SOC range in this mode is set to [50%, 60%]. On the one hand, this reserves enough space to absorb a large amount of regenerative energy, and on the other hand, it avoids excessive SOC leading to charging limitation and loss of braking energy. At the same time, it also takes into account the minimum discharge capacity required for possible subsequent flat ground driving or secondary uphill driving.

[0055] Optionally, the SOC range maintenance method may be: after identifying a full-load downhill mode, the controller jointly determines the feedback intensity based on the battery temperature T and the current SOC value, and actively reduces the upper limit of the feedback current when T > 35℃ to slow down the SOC rise rate; Optionally, the SOC range maintenance method may include: dynamically setting the SOC target value to 55% of the range midpoint, and adjusting the ratio of hydraulic braking to motor braking to make the total feedback power decrease linearly as the SOC increases; Furthermore, the SOC range maintenance method can also employ a sliding window constraint strategy: taking the SOC change rate within the last 30 seconds as input, if the rate of increase exceeds a preset threshold, it automatically switches to a high feedback suppression sub-mode, enabling the range extender to assist in power consumption to balance SOC growth.

[0056] The outbound no-load overall adaptation of the battery's remaining SOC ∈ [40%, 70%]; Among them, the outbound empty load refers to the complete operation of the mining truck from the unloading point to the loading point. Its operation characteristics are light overall load, multiple downhill and uphill sections in the path, and periodic fluctuations in energy demand. This mode belongs to a typical low power density and long-term operation scenario in the relevant technical field. In this embodiment, the SOC adaptation range in this mode is widened to [40%, 70%], which is significantly larger than the traditional fixed range (60%~65%). The purpose is to release the energy regulation elasticity of the battery in long-term light load operation and support energy redistribution across operating conditions.

[0057] Optionally, the SOC range adaptation method may be: initialize the SOC target range as [40%, 70%] at the outbound starting point, and dynamically adjust the current recommended SOC value according to the ratio of cumulative downhill mileage to cumulative uphill mileage throughout the journey, so that it is biased towards the lower limit or upper limit of the range; Optionally, the SOC range adaptation method may include: dividing the outbound journey into several geographical segments, configuring differentiated SOC reference values ​​for each segment based on historical calibration data, for example, setting the first downhill segment to [40%, 50%] and the last uphill segment to [65%, 70%]; Furthermore, the SOC interval adaptation method can also adopt a path-aware rolling optimization strategy: based on a high-precision map, the slope sequence of the next 5 kilometers is obtained, and the optimal SOC trajectory is solved in a rolling manner to maximize the energy utilization efficiency throughout the process, while ensuring that the final value is not less than 40%.

[0058] The remaining SOC of the battery during the return trip at full load is ∈ [65%, 75%]. Among them, the return trip full load refers to the complete operation of the mining truck from the loading point to the unloading point. Its operation characteristics are heavy overall load, mainly uphill in the route, and concentrated and high energy demand. This mode belongs to a typical high power density, short-term heavy load operation scenario in the relevant technical field. In this embodiment, the SOC adaptation range is set to [65%, 75%] in this mode, which is significantly higher than that of the outbound trip. This is to ensure that the battery is always in a high available power output window throughout the entire return trip, to meet peak power demand and suppress the risk of voltage drop.

[0059] Optionally, the SOC range adaptation method can be: starting the SOC lifting program at the return start point, continuously charging with low power through the range extender, so that the SOC reaches 65% before the end of the first flat road driving segment; Optionally, the SOC range adaptation method may include: dividing the return route into three response zones according to the slope gradient; maintaining SOC ∈ [65%, 70%] in the gentle slope zone (8°<α≤15°), maintaining SOC ∈ [70%, 75%] in the medium slope zone (15°<α≤25°), and forcibly locking SOC ≥72% in the steep slope preparation zone (α≥25°); Furthermore, the SOC interval adaptation method can also adopt an event-driven adjustment mechanism: GPS positioning is used to match pre-stored path nodes. When the vehicle enters 1 kilometer before the marked key uphill starting point, a rapid SOC increase command is triggered to ensure that the SOC is not lower than 70% before entering the slope.

[0060] When a steep slope condition is detected and When the steep slope is steep, adaptive control is implemented: the remaining battery charge SOC is anchored to 75%~85% 500 meters in advance.

[0061] Among them, steep slope working condition refers to the defined The running status, Specifically referring to steep incline scenarios, this condition represents an extremely high-power demand scenario within the relevant technical field, posing a severe challenge to the battery's single-cycle discharge depth, thermal management capabilities, and voltage stability. In this embodiment, the 500-meter advance time margin is determined based on a combination of the typical mining truck's operating speed (approximately 10 m / s) and the control system's response delay (approximately 500 ms), corresponding to a physical distance of approximately 500 meters. Anchoring refers to actively adjusting and stably maintaining the State of Charge (SOC) within the target range, rather than instantaneously reaching it and then allowing it to naturally decay. This operation does not rely on post-event compensation but rather on proactive energy preparation. The remaining battery SOC is anchored to 75%~85%, with 75% serving as the minimum threshold to ensure basic power output and 85% as the upper limit to prevent overcharging and thermal runaway risks. This range setting balances high power output capability with safety redundancy boundaries.

[0062] Optionally, the adaptive control method for steep slopes can be: when the slope sensor samples and displays data three times consecutively... When the GPS positioning shows that the distance to the starting point of the known steep slope ahead is ≤500 meters, the controller immediately activates the maximum allowable charging power to replenish the battery until the SOC reaches 75% and then switches to constant SOC maintenance mode. Optionally, the steep slope adaptive control method may include: combining the vehicle's current position and real-time vehicle speed. Based on the rate of change of the slope ahead, the time required to reach the top of the steep slope is predicted, and the SOC is used to increase the slope, so as to achieve a smooth and shock-free anchoring process. Furthermore, the steep slope adaptive control method can also adopt a multi-source collaborative triggering mechanism: in addition to slope and location, battery temperature T is introduced as an additional criterion. If T < 15℃, anchoring is initiated at 800 meters in advance; if T > 38℃, liquid cooling enhancement is initiated and the anchoring target is simultaneously reduced to 75%~80% to balance heat load and power demand.

[0063] This application subdivides the operating conditions of mining trucks into specific modes such as no-load downhill, no-load uphill, full-load uphill, and full-load downhill, and configures differentiated SOC operating ranges for each mode, achieving precise matching between battery energy reserves and operating power requirements. Building upon this, a two-level macroscopic adaptation mechanism is introduced for outbound and inbound travel, forming a nested structure of "microscopic mode-SOC range" and "macroscopic travel-SOC range." Finally, for the most challenging steep uphill conditions, a SOC anchoring strategy of 500 meters in advance is designed, ensuring the battery is fully prepared before high power demands arrive. These synergistic effects collectively solve the technical problem of traditional control strategies having excessively narrow SOC ranges and being unable to respond to wide-range dynamic power demands, significantly improving the energy utilization efficiency, power response capability, and service life of the battery system.

[0064] Furthermore, slope weights are dynamically configured based on the working condition type. Load Capacity Vehicle speed weight , .

[0065] Among them, the gradient weight ω1, load weight ω2, and vehicle speed weight ω3 are normalized adjustment coefficients used to characterize the degree of influence of different operating parameters on the power collaborative distribution between the battery and the range extender under the current working conditions; In this technical field, weighting coefficients are typically used to perform importance-weighted fusion of multi-source input features in order to improve the model's sensitivity to key variables and response priority. Among them, the working condition type is any one of the following: starting working condition, flat ground working condition, uphill working condition, downhill working condition, steep slope working condition, fully loaded working condition, unloaded working condition or overloaded working condition. In the relevant technical field, the working condition type is a discrete state identifier that describes the boundary conditions of vehicle dynamics. It does not directly participate in numerical calculations, but can be used as an index to trigger differentiated control strategies. In this embodiment, the working condition type serves as the input condition for the weight configuration module, used to activate the weight combination under the corresponding working condition. For example, when the working condition is determined to be uphill, ω1=0.45, ω2=0.45, ω3=0.10 is called; when the working condition is determined to be downhill, ω1=0.25, ω2=0.15, ω3=0.60 is called; when the working condition is determined to be flat, ω1=0.30, ω2=0.30, ω3=0.40 is called. All of the above combinations satisfy ω1+ω2+ω3=1, and each combination has been verified by real vehicle calibration to enable the P output of the neural network model under the corresponding working condition. d With P c It is more in line with the actual power demand trend.

[0066] Furthermore, step S3, which involves constructing an input vector from the preprocessed parameters, includes the following steps: Calculate the weighted fusion feature vector ; The weighted fusion feature vector The remaining battery charge (SOC) and the battery temperature This forms a 5-dimensional input vector.

[0067] Among them, the gradient weight ω1, load weight ω2, and vehicle speed weight ω3 are inherently designed to achieve adaptive normalization fusion of multi-source operating condition parameters. In this embodiment, they are used to adjust the contribution ratio of each physical quantity to the overall driving intensity under different operating conditions. For example, in uphill conditions, ω1 is increased to strengthen the gradient dominance; in high-speed flat conditions, ω3 is increased to highlight speed response requirements; and in the fully loaded start-up phase, ω2 is increased to enhance mass inertia perception. The weighted fusion feature vector... X It is a scalar generated by linear weighting of the above three factors, which can characterize the comprehensive driving intensity under the current working conditions. Its inherent function is to compress the input dimension while retaining the original physical meaning. In this embodiment, it serves as the core driving factor of the neural network model, bearing the coupled influence of gradient, load, and vehicle speed, and providing a concise and powerful input basis for subsequent power prediction.

[0068] In this embodiment, the gradient weight ω1, load weight ω2, and vehicle speed weight ω3 are not preset as fixed constants. Instead, based on the determined real-time operating condition type, the corresponding combination values ​​are retrieved from the pre-stored mapping relationship table and used as the weighted fusion feature vector in step S3. The calculation basis ensures that the input vector X It can dynamically reflect the physical constraints that dominate the current operating conditions.

[0069] This application establishes a one-to-one mapping relationship between working condition types and weight combinations, enabling the weighted fusion feature vector... X It can automatically adjust the contribution ratio of each parameter according to the change of working conditions; on this basis...X Together with SOC and T, they form the 5-dimensional input vector of step S3, thereby guiding the neural network model to focus on the dominant physical variables under different operating boundaries; ultimately, it realizes differentiated responses to battery discharge / charge power and range extender power requests, avoids feature expression distortion caused by equal weight processing, and improves the adaptability and stability of power allocation strategy in complex mining truck operation cycles.

[0070] Specifically, the normalized slope Load factor Speed Remaining battery charge (SOC) and battery temperature This forms a 5-dimensional column vector. .

[0071] Optionally, the input vector can be constructed by adding a one-dimensional working condition type code (such as one-hot encoding or embedded vector) to the 5-dimensional basic vector to form a 6-dimensional extended input; Optionally, the input vector can be constructed by stacking multiple 5-dimensional vectors from adjacent time steps using a sliding window to generate a three-dimensional tensor input containing temporal dimensions, thereby enhancing the model's ability to remember dynamic processes.

[0072] Furthermore, the structure of the trained neural network model in step S3 is as follows: The input layer receives the 5-dimensional input vector; this input vector serves as the data entry point for the entire neural network model, carrying comprehensive representation information of the current operating status of the mining truck; its components respectively reflect the weighted coupling effect of slope, load, and vehicle speed, battery energy reserve level, and thermodynamic state, which together constitute the basic input space for subsequent time-series modeling.

[0073] The hidden layer contains three layers of bidirectional long short-term memory network, with 64 neurons in each layer; Among them, the Bidirectional Long Short-Term Memory (Bi-LSTM) network is a recurrent neural network structure with two independent LSTM chains, one forward and one backward, which can simultaneously capture the historical dependencies before and after a certain moment in the time series. Its inherent function is to perform context-aware modeling of dynamic operating condition sequences with sequential correlations (such as the periodic changes of empty downhill → uphill → fully loaded uphill → downhill in mining truck round-trip operations). Bi-LSTM is used to process the time series of continuously collected mining truck operating parameters, enabling the model to identify typical operating condition transitions such as start-up acceleration, continuous steep slope climbing, and braking energy recovery. The 3-layer stacked structure design is used to enhance nonlinear fitting ability and feature abstraction level. Each layer has 64 neurons representing a single-layer hidden state dimension. This scale ensures the accuracy of time series modeling while taking into account the real-time inference resource constraints of the vehicle embedded platform.

[0074] Optionally, the Bi-LSTM modeling method is as follows: set the sliding time window length to N=10 sampling points, take the current time t as the end point of the window, input the 5-dimensional input vector from time t-9 to time t in sequence, and after passing through three layers of Bi-LSTM, output the 128-dimensional temporal feature representation obtained by splicing the forward and backward hidden states of the third layer.

[0075] Optionally, the Bi-LSTM modeling method includes: using a gating mechanism to explicitly control the gradient decay path in the input sequence, and using the forget gate, input gate and output gate to collaboratively determine the retention, update and release of information at each time step, thereby stably learning the power evolution law of mining trucks in long-distance gentle slopes or frequent start-stop sections.

[0076] Furthermore, this Bi-LSTM modeling method also adopts a segmented normalization strategy, introducing layer normalization (LayerNorm) after each layer of Bi-LSTM output to suppress training instability caused by input distribution shift and improve the generalization adaptability of the model under different mining area altitudes, temperatures and humidity environments.

[0077] An attention mechanism layer is connected after the output of the bidirectional long short-term memory network. The attention mechanism layer refers to a learnable weight allocation module built on a query-key-value framework. Its core function is to automatically identify and strengthen the features of the time steps that contribute the most to the current power prediction task in the output of multiple time steps. Its inherent function is to improve the model's ability to focus on key events (such as steep slope start point, full load start transient, downhill energy feedback peak) and alleviate the interference of irrelevant working condition segments (such as flat ground constant speed cruise segment). This attention mechanism is applied to the temporal feature sequence output of the third layer of Bi-LSTM. By calculating the attention score of the hidden state of each time step, a weighted aggregated context vector is generated as a highly discriminative representation for the subsequent power regression task.

[0078] Optionally, the attention mechanism method is as follows: the 128-dimensional hidden state sequence output by the third layer of Bi-LSTM is used as the input of Key and Value to construct a learnable 128-dimensional Query vector, the attention weights at each time step are calculated by dot product, and then the Value sequence is weighted and summed according to the weights to obtain a single fixed-dimensional context vector.

[0079] Optionally, the attention mechanism method includes: adopting a multi-head attention structure, dividing the 128-dimensional hidden state into h=4 subspaces, performing the above Query-Key-Value operation independently in each subspace, and then concatenating the outputs of each head and restoring them to 128 dimensions through linear transformation, so as to enhance the model's ability to capture different types of working condition features (such as slope-dominated, load-dominated, and speed-change-type).

[0080] Furthermore, this attention mechanism method also employs a position encoding enhancement strategy, superimposing a learnable position embedding vector at the input of the Bi-LSTM to explicitly inject time step sequence information, thereby compensating for the limitation of Bi-LSTM itself in having a weak perception of absolute temporal position and improving the recognition accuracy of power change patterns near fixed nodes (such as loading and unloading areas) in round-trip operations.

[0081] The output layer is a 2D linear layer that directly outputs the battery's discharge power P. d and charging power P c .

[0082] Here, the 2D linear layer refers to the layer composed of the weight matrix W∈R 2×128 With bias vector b∈R 2 The fully connected mapping takes a 128-dimensional context vector from the attention mechanism layer as input and outputs a two-dimensional real-valued vector. Its inherent function is to perform a regression mapping from a high-order temporal feature space to a physical power dimension space. This linear layer does not introduce nonlinear activation functions to maintain the interpretability and physical consistency of the output power value, ensuring P... d With P cThe numerical ranges correspond to the maximum allowable discharge / charge power boundaries of the battery system; the output results are directly used as the basis for the vehicle controller to issue power commands.

[0083] This application combines Bi-LSTM with an attention mechanism, enabling the neural network model to both model the temporal evolution of mining truck operating parameters and dynamically focus on key operating conditions affecting power allocation decisions. Based on this, a 2D linear layer is used to achieve end-to-end power regression output, avoiding information loss caused by intermediate feature decoupling. The resulting power prediction not only meets the battery system's instantaneous power response requirements under different operating conditions but also supports global power coordination allocation between the range extender and battery systems, effectively mitigating the problem of accelerated battery aging caused by narrow SOC ranges and large power demand fluctuations.

[0084] Furthermore, the training of the neural network model in step S3 employs a weighted multi-objective loss function: .

[0085] in, For multi-objective loss functions; For energy loss, This is the energy loss weighting coefficient. Due to battery life loss, This is the weighting coefficient for battery life loss. For dynamic loss, This is the weighting coefficient for dynamic loss.

[0086] in, This refers to energy loss, specifically the battery discharge power P predicted by the model. d and charging power P c The cumulative energy deviation between the power required under actual operating conditions and the power required under actual operating conditions is the energy loss in the field of hybrid mining truck power allocation. Energy loss usually reflects the degree to which the combined output energy of the range extender system and the battery system deviates from the optimal economic path. In this embodiment, the loss term is used to constrain the neural network model to reduce the overall energy consumption of the vehicle during the training process. Its input comes from the integral of the difference between the measured power demand under historical operating conditions and the model output power, and the output is a scalar form of energy deviation measurement. Battery life loss is a comprehensive quantitative indicator that negatively impacts the state of health (SOH) of a power battery caused by the model's output power strategy. In the field of power battery systems, life loss is usually determined by multiple degradation factors such as cumulative charge-discharge depth, temperature rise rate, and duration of high-rate current. In this embodiment, this loss term is used to guide the model to avoid power allocation patterns that lead to accelerated battery aging. Its inputs include battery temperature T, SOC change rate, current SOC value, and power fluctuation amplitude, and the output is a dimensionless scalar reflecting the battery capacity decay trend. The dynamic loss refers to the degree of degradation in the vehicle's dynamic performance caused by lag in the model's output power response or insufficient instantaneous power supply. In the field of vehicle dynamics control, dynamic loss is usually manifested as observable performance degradation phenomena such as acceleration deviation, exceeding the climbing time limit, and braking feedback delay. In this embodiment, this loss term is used to enhance the model's response capability to sudden operating conditions (such as starting and steep hill entry), and its input is derived from vehicle speed. rate of change, slope step change and load factor The jump amplitude is output as a scalar penalty term representing insufficient dynamic response margin.

[0087] Specifically, This is the energy loss weighting coefficient. This is the weighting coefficient for battery life loss. The weighting coefficients are dynamic loss weights, which are adjustable parameters used to adjust the relative importance of each sub-loss term in the total loss function. In multi-objective machine learning modeling, the weighting coefficients are used to achieve dimensional normalization of loss terms with different physical dimensions and balance the emphasis of the objectives. In this embodiment, these weighting coefficients do not participate in the backpropagation update of the neural network, but are set as hyperparameters before training. Their values ​​are limited to the interval [0,1] and satisfy the following conditions: Different weight combinations can be configured for different operating conditions in different mining areas. For example, when operating in high-temperature mining areas, Set to 0.5–0.7 to enhance battery thermal durability constraints; in long-distance, continuous uphill mining areas, Increase it to above 0.4 to ensure continuous power response; in energy cost-sensitive mining areas, Set as the dominant weight (≥0.6) to optimize fuel economy.

[0088] In this embodiment, a weighted multi-objective loss function is introduced. This allows the neural network model to simultaneously consider three heterogeneous optimization objectives during training: energy efficiency, battery health maintenance, and vehicle dynamic response; leveraging , , right , , Through coordinated regulation, the model not only learns the static mapping relationship of power allocation, but also embeds strategy preferences for different operating scenarios. On this basis, combined with normalization processing and segmented weighting mechanism, it effectively alleviates the differences and conflicts between multiple objectives, and improves the generalization and robustness of the model in the closed-loop control of real mining trucks. Ultimately, it achieves the technical effect of extending battery life while ensuring power performance and simultaneously optimizing the energy consumption level of the whole vehicle.

[0089] Furthermore, during the overload condition, the audible and visual alarm is triggered, and a speed limit command is sent to the vehicle controller.

[0090] Specifically, based on the overload signal, the following two parallel actions are executed simultaneously: driving the on-board audible and visual alarm to emit a continuous warning signal; and sending a speed limit command to the vehicle controller. The audible and visual alarm is an integrated human-machine interface device installed in the driver's cab, possessing dual output capabilities of visible light flashing and audible audio prompts, used to provide the driver with intuitive and unmistakable safety warnings; its role in this embodiment is to achieve localized real-time early warning, ensuring that operators are aware of the overload risk at the first moment. The vehicle controller is the core coordinating unit of the mining truck's electronic control system, receiving input signals from multiple subsystems and performing unified scheduling of the motor controller, engine management system, and braking control system, etc. The speed limit command is a standardized communication message whose content does not contain specific numerical settings, but rather instructs the vehicle controller to dynamically adjust the current vehicle speed limit to below a preset safety threshold; in this embodiment, the function of this command is to actively suppress vehicle power output through upper-level control intervention, reduce transmission system stress, extend the life of braking components, and reduce the risk of loss of control due to overload.

[0091] By using the overload condition identification result as a unified trigger source, the two response paths of audible and visual alarm and speed limit control are driven synchronously, achieving strong coupling between local warning and system intervention. Utilizing three different levels of speed limit command implementation methods—CAN communication, hardware I / O, and software services—reliable execution is ensured under different electronic control architectures. Furthermore, there is no master-slave dependency between the audible and visual alarm and the speed limit action; both are independently triggered by the same overload signal, ensuring that the other path can maintain basic safety functions even if one path fails. Ultimately, this forms a proactive overload protection mechanism for mining trucks covering the entire chain of perception, judgment, and response, effectively supporting the closed-loop implementation of the condition judgment system defined in this application in the safety dimension.

[0092] Furthermore, when the battery temperature At that time, the mining card is preheated until .

[0093] Specifically, when the battery temperature T < -10℃, the mining card is preheated until T ≥ 0℃.

[0094] Among them, battery temperature T refers to the average or highest temperature value collected in real time by multiple temperature sensors arranged inside the battery module or near the cell tabs, which is used to characterize the current thermal state of the power battery system; T < -10℃ is the low-temperature start-up failure risk threshold. Below this temperature, the viscosity of the lithium-ion battery electrolyte increases significantly and the ion migration rate decreases, resulting in increased internal resistance, decreased usable capacity, limited charging and discharging power, and the risk of lithium plating; preheating refers to the process of inputting external heat energy into the battery system to raise its overall temperature, the purpose of which is to restore the battery's electrochemical activity and power response capability in low-temperature environments; until T ≥ 0℃ indicates that the preheating process continues until the monitored battery temperature reaches or exceeds 0℃, the minimum safe operating temperature benchmark. At this time, the battery can meet the basic drive power output and energy feedback absorption requirements.

[0095] By using battery temperature T as the key state criterion, setting T < -10℃ as the preheating trigger condition, and T ≥ 0℃ as the termination condition, a closed-loop low-temperature adaptive control logic is constructed. Based on this, active thermal management of the battery system under extremely cold conditions is achieved by utilizing multiple heat source paths such as PTC direct heating, engine waste heat recovery, or external auxiliary heating. This collaborative mechanism does not depend on the vehicle's driving status and can be executed in advance during the parking and standby phase, effectively avoiding problems such as start-up failure, lack of climbing ability, or interruption of energy feedback caused by limited battery power in low-temperature environments, thereby ensuring the all-weather attendance capability and operational reliability of hybrid mining trucks in high-altitude and cold mining areas.

[0096] Furthermore, when the battery temperature At the same time, the proportion of battery power output is reduced, while the proportion of range extender power is increased.

[0097] Among them, battery temperature T refers to the weighted average temperature of multiple temperature measurement points in the power battery pack or the highest single-cell temperature. Its value is collected in real time by the vehicle battery management system (BMS) and reported to the vehicle controller (VCU); T>42℃ is a preset thermal safety threshold, which is determined based on the thermal runaway initiation temperature range of lithium-ion batteries under continuous high-rate charge and discharge conditions, the temperature distribution of typical mining truck operating environments, and long-term aging test data; reducing the battery power output ratio means reducing the proportion of the absolute value of the power undertaken by the battery system to the total power request under the premise that the current total drive / feedback power demand remains unchanged. The adjustment method does not change the total power demand command, but only reconstructs the power distribution coefficient between the battery and the range extender; simultaneously increasing the range extender power ratio means that while the battery power output ratio is reduced, the proportion of the range extender output power to the total power request is increased by an equal amount to maintain the continuity of vehicle power performance and energy conservation constraints.

[0098] This method monitors whether the battery temperature T exceeds the critical thermal boundary condition of 42°C, dynamically reconstructs the power distribution weight between the battery and the range extender, and actively reduces the thermal load intensity of the battery system while ensuring continuous response to the power demand of the whole vehicle. By leveraging the stable and sustainable power output capability of the range extender to share high-power tasks, it effectively suppresses the battery temperature rise trend and avoids entering the sensitive range of thermal runaway. This not only extends the sustainable operating time of mining trucks in high-temperature environments, but also achieves the coupling and synergy of thermal management strategy and energy distribution strategy, improving the operational safety and service durability of the hybrid system under extreme conditions.

[0099] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A method for coordinated power allocation between range extender and battery based on mining truck operating conditions, characterized in that, Includes the following steps: S1. Real-time acquisition of mining truck operating parameters, including slope. Load factor Speed Remaining battery charge (SOC) and battery temperature ,in, For real-time load, Maximum load capacity; S2. Determine the real-time operating condition of the mining truck, and perform standardized preprocessing on the operating parameters of the mining truck based on the real-time operating condition of the mining truck to obtain the preprocessed parameters. S3. The preprocessed parameters are used to form an input vector, which is then input into the trained neural network model to obtain the battery's discharge power P. d and charging power P c ; S4. Predict the feature sequence of future operating conditions using the trained neural network model, and adjust the discharge power P of the battery based on the feature sequence of the future operating conditions. d and charging power P c And the power of the range extender is redistributed.

2. The range extender-battery power collaborative allocation method based on mining truck operating conditions according to claim 1, characterized in that, In step S2, determining the real-time operating status of the mining truck includes the following steps: when This is considered the initial operating condition; when The condition was determined to be flat ground; when The condition is classified as uphill; when The condition is classified as a downhill slope; when The condition was determined to be a steep slope; when Determined to be in full-load condition. For full load weight; when Determined to be in no-load condition. For the weight of the vehicle; when The condition was determined to be overloaded.

3. The range extender-battery power collaborative allocation method based on mining truck operating conditions according to claim 2, characterized in that, Match the preset battery system request mode according to the current operating condition type, and dynamically set the corresponding SOC working range for the remaining battery power: The no-load downhill mode corresponds to the remaining battery charge SOC ∈ [40%, 60%]; The no-load uphill mode corresponds to the remaining battery charge SOC ∈ [60%, 70%]; The full-load uphill mode corresponds to the remaining battery charge SOC ∈ [70%, 75%]; The full-load downhill mode corresponds to the remaining battery charge SOC ∈ [50%, 60%]; The outbound no-load overall adaptation of the battery's remaining SOC ∈ [40%, 70%]; The remaining SOC of the battery during the return trip at full load is ∈ [65%, 75%]. When a steep slope condition is detected and When the steep slope is steep, adaptive control is implemented: the remaining battery charge SOC is anchored to 75%~85% 500 meters in advance.

4. The range extender-battery power collaborative allocation method based on mining truck operating conditions according to claim 1, characterized in that, Dynamically configure slope weights based on working conditions. Load Capacity Vehicle speed weight , .

5. The range extender-battery power collaborative allocation method based on mining truck operating conditions according to claim 4, characterized in that, Step S3, which involves constructing an input vector from the preprocessed parameters, includes the following steps: Calculate the weighted fusion feature vector ; The weighted fusion feature vector The remaining battery charge (SOC) and the battery temperature This forms a 5-dimensional input vector.

6. The range extender-battery power collaborative allocation method based on mining truck operating conditions according to claim 5, characterized in that, The structure of the trained neural network model in step S3 is as follows: The input layer receives the 5-dimensional input vector; The hidden layer contains three layers of bidirectional long short-term memory network, with 64 neurons in each layer; An attention mechanism layer is connected after the output of the bidirectional long short-term memory network. The output layer is a 2D linear layer that directly outputs the battery's discharge power P. d and charging power P c .

7. The range extender-battery power collaborative allocation method based on mining truck operating conditions according to claim 6, characterized in that, The neural network model trained in step S3 uses a weighted multi-objective loss function: in, For multi-objective loss functions; For energy loss, This is the energy loss weighting coefficient. Due to battery life loss, This is the weighting coefficient for battery life loss. For dynamic loss, This is the weighting coefficient for dynamic loss.

8. The range extender-battery power collaborative allocation method based on mining truck operating conditions according to claim 2, characterized in that, When the overload condition occurs, the audible and visual alarm is triggered and a speed limit command is sent to the vehicle controller.

9. The range extender-battery power collaborative allocation method based on mining truck operating conditions according to claim 1, characterized in that, When the battery temperature At that time, the mining card is preheated until .

10. The range extender-battery power collaborative allocation method based on mining truck operating conditions according to claim 1, characterized in that, When the battery temperature At the same time, the proportion of battery power output is reduced, while the proportion of range extender power is increased.

Images