Super capacitor hybrid power energy distribution optimization method and system for mining dump truck

Abstract

The invention provides a mining dump truck super capacitor hybrid power energy distribution optimization method and system, and relates to the technical field of energy management, and the method comprises the steps: recognizing a braking recovery and climbing consumption interval according to the position and slope information, and calculating a super capacitor target charge state sequence; the basic power is distributed to the engine, the transient power is distributed to the super capacitor, and rolling optimization is carried out by adopting a particle swarm algorithm to generate a power distribution instruction. The energy recycling efficiency is effectively improved, the service life of the super capacitor is prolonged, the fuel consumption is reduced, and the vehicle operation economy is improved.

Description

Technical Field

[0001] This invention relates to the field of energy management technology, and in particular to a method and system for optimizing energy distribution in a mining dump truck using a supercapacitor hybrid power system. Background Technology

[0002] Mining dump trucks, as the main transportation equipment in mines, are generally driven by high-power diesel engines. Due to their special working environment and operating conditions, they face challenges such as high energy consumption and the need to improve environmental performance. To improve energy efficiency and optimize environmental adaptability, the supercapacitor hybrid power system for mining dump trucks has emerged. This system combines the advantages of traditional engines and supercapacitor energy storage devices, effectively recovering braking energy, smoothing power fluctuations, and achieving rational energy allocation and utilization.

[0003] Currently, energy allocation methods for supercapacitor hybrid power systems in mining dump trucks mainly include rule-based control strategies, fuzzy logic control, and dynamic programming. Rule-based control strategies allocate energy through preset thresholds and logical judgments; fuzzy logic control utilizes fuzzy inference mechanisms for energy management; and dynamic programming aims at global optimization to calculate the optimal energy allocation strategy. These methods have improved the energy utilization efficiency of hybrid power systems to some extent.

[0004] However, existing technologies still have shortcomings and deficiencies in energy distribution optimization of supercapacitor hybrid power systems for mining dump trucks. Existing energy distribution methods often fail to effectively distinguish between transient and fundamental components of power demand, causing the engine to frequently operate in non-economical ranges. This prevents the supercapacitor from fully leveraging its transient power response advantage, reducing overall system efficiency. Traditional energy management strategies fail to fully consider the unique operating path characteristics of mining dump trucks, such as potential energy changes on inclines and declines, and lack the ability to predict future road conditions. They cannot prepare the supercapacitor's state of charge in advance for braking energy recovery and climbing power demands, affecting energy utilization efficiency. Existing methods typically employ offline optimization or simple real-time control, lacking a mechanism to effectively combine global optimization objectives with real-time control. This makes it difficult to achieve precise control of the supercapacitor's state of charge under complex and variable operating conditions, resulting in unreasonable energy distribution and suboptimal system performance. Summary of the Invention

[0005] This invention provides a method and system for optimizing energy distribution in a mining dump truck using a supercapacitor hybrid power system, which can solve the problems in the prior art.

[0006] A first aspect of this invention provides a method for optimizing energy distribution in a mining dump truck using a supercapacitor hybrid power system, comprising:

[0007] Obtain information on the power demand of mining dump trucks, the state of charge of supercapacitors, the path gradient, and the vehicle location;

[0008] The phase space of the power demand information is reconstructed to construct the attractor trajectory. By calculating the curvature tensor eigenvalues ​​of the attractor trajectory and mapping them back to the time domain, the transient power components and the fundamental power components are obtained.

[0009] Based on vehicle location information and path slope information, a potential energy change curve is constructed. The braking energy recovery interval and the climbing power consumption interval are identified from the potential energy change curve. The constraint requirements of each interval on the supercapacitor's state of charge are calculated through the backpropagation algorithm to obtain the target state of charge sequence.

[0010] The basic power component is allocated to the engine, and the transient power component is allocated to the supercapacitor. The particle swarm optimization algorithm is used to perform rolling optimization with the deviation between the supercapacitor's state of charge information and the target state of charge sequence as the optimization objective, so as to obtain the power allocation correction amount and generate the final power allocation command.

[0011] The final power distribution command is sent to the engine control unit and the supercapacitor control unit for execution.

[0012] In one optional embodiment, the power demand information is reconstructed in phase space to construct the attractor trajectory. By calculating the curvature tensor eigenvalues ​​of the attractor trajectory and mapping them back to the time domain, the transient power components and the fundamental power components are obtained, including:

[0013] The power demand information is reconstructed into phase space according to the embedding dimension and time delay. Attractor trajectories are formed in the reconstructed phase space. The attractor trajectories are discretized and sampled to obtain a sequence of trajectory points. A local neighborhood space is constructed for each trajectory point in the sequence of trajectory points. The tangent vector and normal vector of the trajectory points are extracted in the local neighborhood space. The curvature tensor matrix of the trajectory points is constructed based on the tangent vector and normal vector. The curvature tensor matrix is ​​decomposed into eigenvalues ​​to obtain a sequence of curvature tensor eigenvalues.

[0014] Gradient calculation is performed on the curvature tensor eigenvalue sequence to obtain the curvature gradient sequence. Based on the curvature gradient sequence, the positions where the gradient magnitude exceeds the preset abrupt change threshold are identified, and the power abrupt change points are determined. Using the power abrupt change points as boundaries, the attractor trajectory is divided into multiple trajectory segments. The mean curvature tensor eigenvalues ​​of each trajectory segment are calculated. The trajectory segments with the mean curvature tensor eigenvalues ​​greater than the geometric stability threshold are mapped back to the time domain through inverse phase space transformation to obtain the transient power components. The trajectory segments with the mean curvature tensor eigenvalues ​​less than the geometric stability threshold are mapped back to the time domain through inverse phase space transformation to obtain the fundamental power components.

[0015] In an optional embodiment, constructing the curvature tensor matrix of the trajectory points based on the tangent vector and the normal vector includes:

[0016] Construct an initial curvature tensor matrix based on the tangent and normal vectors of each trajectory point;

[0017] Select the forward nearest neighbor and backward nearest neighbor trajectory points of the current trajectory point in the local neighborhood space;

[0018] Calculate the angle between the tangent vector of the current trajectory point and the tangent vector of the nearest forward trajectory point, and the angle between the tangent vector of the nearest backward trajectory point, to obtain the forward tangent angle and the backward tangent angle respectively. At the same time, calculate the angle between the normal vector of the current trajectory point and the normal vector of the nearest forward trajectory point, and the angle between the normal vector of the nearest backward trajectory point, to obtain the forward normal angle and the backward normal angle respectively.

[0019] The tangential coupling component is obtained by weighted summation of the forward tangential angle and the backward tangential angle, and the normal coupling component is obtained by weighted summation of the forward normal angle and the backward normal angle. The tensor coupling coefficient is calculated based on the tangential coupling component and the normal coupling component.

[0020] The initial curvature tensor matrix of the current trajectory point is weighted and fused with the initial curvature tensor matrices of the forward nearest neighbor trajectory points and the backward nearest neighbor trajectory points according to the tensor coupling coefficient to obtain the curvature tensor matrix of the current trajectory point.

[0021] In one optional embodiment, a potential energy change curve is constructed based on vehicle location information and path gradient information. Identifying the braking energy recovery range and the climbing power consumption range from the potential energy change curve includes:

[0022] Based on vehicle location information and path slope information, the location coordinates and slope angles of each road segment in the path are obtained. The slope angles are organized into a slope sequence according to the road segment order. The locations in the slope sequence where the slope change rate exceeds the preset change rate are marked. The slope sequence is segmented at the marked locations. Potential energy sub-curves are constructed for each segment of the slope sequence. At the connection between adjacent potential energy sub-curves, a transition curve segment is generated through an interpolation algorithm. The potential energy sub-curves and the transition curve segment are spliced ​​together to form a complete potential energy change curve.

[0023] A sliding window is set on the potential energy change curve. By moving the sliding window on the potential energy change curve, local potential energy features are extracted. Window regions in the local potential energy features where the potential energy decreases by more than a preset recovery threshold and the duration exceeds a preset recovery distance are merged into braking energy recovery intervals. Window regions in the local potential energy features where the potential energy increases by more than a preset consumption threshold and the duration exceeds a preset climbing distance are merged into climbing power consumption intervals. The interval between the braking energy recovery interval and the climbing power consumption interval is marked as a transition interval.

[0024] In one optional embodiment, the constraint requirements of each interval on the supercapacitor's state of charge are calculated using a backpropagation algorithm to obtain the target state of charge sequence, which includes:

[0025] Establish a reverse traversal sequence from the end of the path. Extract the energy consumption demand value of the climbing power consumption interval and the energy recovery capacity value of the braking energy recovery interval in the reverse traversal sequence. Convert the energy consumption demand value into the state of charge demand of the corresponding position and the energy recovery capacity value into the state of charge compensation of the corresponding position.

[0026] A reverse accumulation and transfer mechanism for state of charge (SOC) demand is established. The SOC demand at the current position is subtracted from the SOC compensation provided by the subsequent regenerative braking range to obtain the SOC demand transferred to the previous position. During the transfer process, the SOC demand is subject to supercapacitor capacity boundary constraints. When the SOC demand exceeds the capacity limit, the excess portion is allocated to the engine. When the SOC demand is below the capacity limit, an additional charging command is inserted.

[0027] After the reverse cumulative transfer is completed, the charge state requirements at each position are arranged in forward order to generate the target charge state sequence.

[0028] In one optional embodiment, the base power component is allocated to the engine, and the transient power component is allocated to the supercapacitor. A particle swarm optimization algorithm is used, with the deviation between the supercapacitor's state of charge information and the target state of charge sequence as the optimization objective, to perform rolling optimization, obtaining the power allocation correction amount. The final power allocation command is generated, including:

[0029] The base power component is allocated to the engine to obtain the engine's base power value. ， The transient power components are distributed to the supercapacitor to obtain the transient power value of the supercapacitor;

[0030] Based on the engine's base power value and the supercapacitor's transient power value, the predicted evolution trajectory of the supercapacitor's state of charge information is calculated. ， Extract the target state of charge (SPO) values ​​at the corresponding time points from the target SPO sequence. ， Calculate the state-of-charge tracking deviation between the predicted evolution trajectory and the target state-of-charge value;

[0031] Initialize the particle positions and velocities of the particle swarm ， Decode the particle position into a power distribution ratio adjustment coefficient. ， The power distribution ratio adjustment coefficient is used to proportionally adjust the engine's base power value and the supercapacitor's transient power value. ， The deviation between the supercapacitor's state of charge information and the target state of charge sequence was recalculated based on the adjusted power value. ，The particle fitness value is obtained by weighting the deviation with the violation of the supercapacitor capacity constraint and the violation of the engine power change rate constraint. ， The particle position and velocity are iteratively updated based on the particle fitness value until convergence. ， Extract the power allocation ratio adjustment coefficient corresponding to the global optimal particle position to determine the power allocation correction amount;

[0032] The power distribution correction is used to adjust the engine's base power value and the supercapacitor's transient power value. ， Generate the final power allocation command.

[0033] In one optional embodiment, calculating the predicted evolution trajectory of the supercapacitor's state of charge information based on the engine's base power value and the supercapacitor's transient power value includes:

[0034] Obtain the current state of charge value of the supercapacitor, determine the prediction starting point, determine whether the supercapacitor is in charging or discharging mode based on the transient power value of the supercapacitor, and extract the charging efficiency curve or discharging efficiency curve corresponding to the current state of charge value.

[0035] Substitute the transient power value of the supercapacitor into the charging efficiency curve or discharging efficiency curve to calculate the actual charging and discharging power of the supercapacitor. Calculate the change in the state of charge of the supercapacitor based on the actual charging and discharging power of the supercapacitor. Add the current state of charge value and the change in the state of charge of the supercapacitor to obtain the state of charge value at the next moment.

[0036] Determine whether the charge interval to which the state of charge value belongs changes at the next moment. If the charge interval changes, extract the charging efficiency curve or discharging efficiency curve corresponding to the state of charge value at the next moment. Use the updated charging efficiency curve or discharging efficiency curve to recalculate the actual charging and discharging power of the supercapacitor and the change in the state of charge of the supercapacitor.

[0037] Within the preset prediction time domain, the state of charge value of the next time step is calculated iteratively, and the state of charge values ​​of each time step are arranged in chronological order to generate the predicted evolution trajectory.

[0038] A second aspect of the present invention provides a supercapacitor hybrid power energy distribution optimization system for mining dump trucks, comprising:

[0039] The data acquisition unit is used to acquire information on the power demand of mining dump trucks, the state of charge of supercapacitors, the path gradient, and the vehicle location.

[0040] The power decomposition unit is used to reconstruct the attractor trajectory from the phase space of the power demand information. By calculating the curvature tensor eigenvalues ​​of the attractor trajectory and mapping them back to the time domain, the transient power components and the fundamental power components are obtained.

[0041] The target planning unit is used to construct the potential energy change curve based on vehicle location information and path slope information, identify the braking energy recovery interval and the climbing power consumption interval from the potential energy change curve, and calculate the constraint requirements of each interval on the supercapacitor's state of charge through the backpropagation algorithm to obtain the target state of charge sequence.

[0042] The optimization control unit is used to allocate the basic power component to the engine and the transient power component to the supercapacitor. The particle swarm optimization algorithm is used to perform rolling optimization with the deviation between the supercapacitor's state of charge information and the target state of charge sequence as the optimization target to obtain the power allocation correction amount and generate the final power allocation command.

[0043] The execution output unit is used to send the final power distribution command to the engine control unit and the supercapacitor control unit for execution.

[0044] A third aspect of the present invention provides an electronic device, comprising:

[0045] processor;

[0046] Memory used to store processor-executable instructions;

[0047] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0048] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0049] In this embodiment of the invention, an innovative method is employed to construct attractor trajectories through phase space reconstruction and to map their curvature tensor eigenvalues ​​back to the time domain. This method achieves precise decomposition of power demand information, effectively distinguishes transient power components from fundamental power components, and makes energy allocation more rational, thereby improving system efficiency. Based on vehicle position and path slope information, a potential energy change curve is constructed, creatively identifying the braking energy recovery interval and the climbing power consumption interval. The constraint requirements of the supercapacitor's state of charge are calculated using the backpropagation algorithm, enabling forward-looking management of supercapacitor energy storage and effectively solving the energy management problem under special working conditions of mining dump trucks. Particle swarm optimization is used for rolling optimization, with the deviation between the supercapacitor's state of charge and the target state sequence as the optimization objective. This dynamically adjusts power allocation, enabling the system to adapt to complex and changing mining road conditions and load conditions, significantly improving energy utilization efficiency and reducing fuel consumption. Attached Figure Description

[0050] Figure 1 This is a flowchart illustrating the energy distribution optimization method for supercapacitor hybrid power systems in mining dump trucks according to an embodiment of the present invention.

[0051] Figure 2 A flowchart for reverse traversal of the energy management strategy. Detailed Implementation

[0052] 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, 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.

[0053] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0054] Figure 1 This is a flowchart illustrating the energy distribution optimization method for supercapacitor hybrid power systems in mining dump trucks according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0055] Obtain information on the power demand of mining dump trucks, the state of charge of supercapacitors, the path gradient, and the vehicle location;

[0056] The phase space of the power demand information is reconstructed to construct the attractor trajectory. By calculating the curvature tensor eigenvalues ​​of the attractor trajectory and mapping them back to the time domain, the transient power components and the fundamental power components are obtained.

[0057] Based on vehicle location information and path slope information, a potential energy change curve is constructed. The braking energy recovery interval and the climbing power consumption interval are identified from the potential energy change curve. The constraint requirements of each interval on the supercapacitor's state of charge are calculated through the backpropagation algorithm to obtain the target state of charge sequence.

[0058] The basic power component is allocated to the engine, and the transient power component is allocated to the supercapacitor. The particle swarm optimization algorithm is used to perform rolling optimization with the deviation between the supercapacitor's state of charge information and the target state of charge sequence as the optimization objective, so as to obtain the power allocation correction amount and generate the final power allocation command.

[0059] The final power distribution command is sent to the engine control unit and the supercapacitor control unit for execution.

[0060] In one alternative implementation ， The attractor trajectory is constructed by reconstructing the phase space of the power demand information. By calculating the curvature tensor eigenvalues ​​of the attractor trajectory and mapping them back to the time domain, the transient power components and the fundamental power components are obtained, including:

[0061] The power demand information is reconstructed into phase space according to the embedding dimension and time delay. Attractor trajectories are formed in the reconstructed phase space. The attractor trajectories are discretized and sampled to obtain a sequence of trajectory points. A local neighborhood space is constructed for each trajectory point in the sequence of trajectory points. The tangent vector and normal vector of the trajectory points are extracted in the local neighborhood space. The curvature tensor matrix of the trajectory points is constructed based on the tangent vector and normal vector. The curvature tensor matrix is ​​decomposed into eigenvalues ​​to obtain a sequence of curvature tensor eigenvalues.

[0062] Gradient calculation is performed on the curvature tensor eigenvalue sequence to obtain the curvature gradient sequence. Based on the curvature gradient sequence, the positions where the gradient magnitude exceeds the preset abrupt change threshold are identified, and the power abrupt change points are determined. Using the power abrupt change points as boundaries, the attractor trajectory is divided into multiple trajectory segments. The mean curvature tensor eigenvalues ​​of each trajectory segment are calculated. The trajectory segments with the mean curvature tensor eigenvalues ​​greater than the geometric stability threshold are mapped back to the time domain through inverse phase space transformation to obtain the transient power components. The trajectory segments with the mean curvature tensor eigenvalues ​​less than the geometric stability threshold are mapped back to the time domain through inverse phase space transformation to obtain the fundamental power components.

[0063] In one specific implementation, the power demand information of the mining dump truck is collected from the on-board controller. Data such as wheel drive power and braking power are acquired through power sensors, with a sampling frequency set to 10 Hz. The collected power demand information is presented as a one-dimensional time series, recording the power demand values ​​of the vehicle at different times.

[0064] Before reconstructing the phase space of the collected power demand information, data preprocessing is required, including outlier removal and noise filtering. Outlier removal uses the three-standard-deviation method, marking data points that deviate more than three standard deviations from the mean as outliers and replacing them with the average of the preceding and following data. Noise filtering uses the median filtering method, selecting a window length of 5 sampling points to effectively remove impulse interference in the power signal while preserving the true characteristics of power changes.

[0065] Phase space reconstruction is the process of mapping a one-dimensional time series to a multi-dimensional space, which requires determining two key parameters: embedding dimension and time delay. The embedding dimension is determined using the false nearest neighbor algorithm. When the embedding dimension increases to 4, the proportion of false nearest neighbors drops below 5%, therefore, an embedding dimension of 4 is chosen. The time delay is determined using the mutual information method. When the delay is 3 sampling points, the mutual information value drops to 1 / e of the initial value, therefore, a time delay of 3 is chosen.

[0066] Based on a defined embedding dimension and time delay, phase space reconstruction is performed on the power demand information. Specifically, the four power values ​​spaced three sampling points apart in the original time series are combined into a 4-dimensional vector, where the phase space coordinates for each time point t are [P(t)]. ， P(t+3)， P(t+6) ， [P(t+9)], where P(t) represents the power value at time t, in kilowatts. For example, when the original power sequence has values ​​of

[120] at a certain time and subsequent times... ， 118 ， 115 ， 113 ， 110 ， 105 ， 102 ， 100 ， 98 ， 95 ， 92 ， 90] kWh, the corresponding phase space point is

[120] ， 113 ， 105 ， 98]、[118 ， 110 ， 102 ， 95]、[115 ， 105 ， 100 ， 92] etc.

[0067] After phase space reconstruction, the original one-dimensional time series is transformed into a set of points in four-dimensional space. These points are connected in time sequence to form a trajectory, namely the attractor trajectory. The attractor trajectory reflects the inherent dynamic characteristics of the mining dump truck's power system and contains pattern information on power changes under various operating conditions.

[0068] The attractor trajectory is discretized and sampled to obtain a sequence of trajectory points. The sampling interval is set to two original sampling points to ensure that the sampling density is sufficient to reflect the trajectory characteristics while avoiding data redundancy. The sampled sequence of trajectory points preserves the geometric shape of the attractor trajectory, providing basic data for subsequent local geometric characteristic analysis.

[0069] For each point in the trajectory point sequence, a local neighborhood space is constructed, containing the set of nearest neighbors around the current point. The number of nearest neighbors is chosen to be 10, determined by calculating the Euclidean distance. For example, for a trajectory point

[105] ， 102 ， 98 ， The local neighborhood of 95 contains the 10 points with the smallest Euclidean distance to it. These points are distributed in phase space near the point and can reflect the local structure of the trajectory at that location.

[0070] In the constructed local neighborhood space, the tangent vector and normal vector of the trajectory points are extracted. The point set within the neighborhood is centered, the covariance matrix of the point set is calculated, and eigenvalue decomposition is performed on the covariance matrix. The eigenvector corresponding to the largest eigenvalue is used as the tangent vector, representing the main direction of motion of the trajectory at that point; the remaining eigenvectors are used as the normal vector, representing the curvature direction of the trajectory. For example, the eigenvalues ​​of the covariance matrix of the point set within the neighborhood are [0.85]. ， 0.12 ， 0.02 ， [0.01], the corresponding eigenvectors are the tangent vector and the three normal vectors, respectively.

[0071] The curvature tensor matrix of the trajectory points is constructed based on the extracted tangent and normal vectors. The curvature tensor matrix has a dimension of (m-1)×(m-1), i.e., a 3×3 matrix, and its elements are calculated by projecting the rates of change of the normal and tangent vectors. The curvature tensor matrix reflects the local curvature of the trajectory at that point; larger matrix elements indicate a higher degree of curvature in the corresponding direction.

[0072] Eigenvalue decomposition is performed on the curvature tensor matrix to obtain a sequence of curvature tensor eigenvalues. Each trajectory point corresponds to three eigenvalues, arranged in descending order. The magnitude of the eigenvalues ​​reflects the curvature of the trajectory in different directions; larger eigenvalues ​​indicate drastic changes in the trajectory along that direction. For example, the curvature tensor eigenvalues ​​of a certain trajectory point are [0.075]. ， 0.032 ， [0.008] indicates that the trajectory at this point has a high degree of curvature in the direction of principal curvature.

[0073] Gradient calculations are performed on the curvature tensor eigenvalue sequence to obtain the curvature gradient sequence. The gradient calculation employs the central difference method. For non-boundary points, the gradient value is the difference between the eigenvalue of the next point and the eigenvalue of the previous point, divided by two. For boundary points, forward or backward differencing is used. The curvature gradient reflects the rate of curvature change; larger gradient values ​​correspond to regions where the trajectory's geometric properties change rapidly.

[0074] The power abrupt change points are identified by using the curvature gradient sequence to pinpoint locations where the gradient amplitude exceeds a preset abrupt change threshold. The preset threshold is set to 2.5 times the standard deviation of the gradient sequence, providing good identification results for typical operating conditions of mining dump trucks. Power abrupt change points typically correspond to vehicle operating condition transition points, such as starting acceleration, transitioning from a hill to a flat road, and the start of braking.

[0075] Using the identified power abrupt change points as boundaries, the attractor trajectory is divided into multiple trajectory segments. Each trajectory segment represents the power demand characteristics of the mining dump truck under specific working conditions. For example, in a certain test, 5 power abrupt change points were identified, and the attractor trajectory was divided into 6 trajectory segments, corresponding to working conditions such as idling, acceleration, constant speed driving, climbing, flat-top driving, and downhill braking.

[0076] The mean eigenvalues ​​of the curvature tensor are calculated for each obtained trajectory segment. The mean is calculated as the arithmetic mean of the eigenvalues ​​of all points within the trajectory segment, resulting in a triplet. The mean eigenvalue reflects the overall geometric stability of the trajectory segment; a larger mean indicates more drastic changes in the trajectory within that segment. For example, the mean eigenvalue of the acceleration segment is [0.068]. ， 0.027 ， The mean eigenvalue of the uniform velocity segment is [0.006], while the mean eigenvalue of the uniform velocity segment is [0.012]. ， 0.005 ， 0.001].

[0077] A geometric stability threshold is set to divide trajectory segments into geometrically stable and geometrically unstable segments. The geometric stability threshold is determined by statistical analysis of historical operating data of mining dump trucks and is set as the maximum eigenvalue of 0.04. Trajectory segments with a maximum eigenvalue greater than the geometric stability threshold are classified as geometrically unstable segments, while those with a eigenvalue less than the threshold are classified as geometrically stable segments.

[0078] By performing an inverse phase space transformation, the geometrically unstable segment is mapped back to the time domain to obtain the transient power components. The inverse transformation is achieved by extracting the original time points and power values ​​corresponding to each point in the reconstructed phase space. The transient power components manifest as power spikes or fluctuations over a short period of time, corresponding to high-dynamic operating conditions such as acceleration, climbing, and braking. The power fluctuation amplitude typically exceeds 50% of the base power, and the duration is between 3 and 15 seconds.

[0079] Similarly, the geometrically stable segment is mapped back to the time domain through an inverse phase space transformation to obtain the fundamental power component. The fundamental power component exhibits a relatively stable power curve, corresponding to steady-state conditions such as constant speed cruising or gentle slope driving. The power fluctuation amplitude is usually no more than 20% of the average value, and the duration is relatively long, typically more than 30 seconds.

[0080] Through power demand decomposition technology, the supercapacitor hybrid power system for mining dump trucks has achieved fine division of power signals, providing a precise basis for energy management strategies.

[0081] In one alternative implementation ， Constructing the curvature tensor matrix of the trajectory points based on the tangent vector and normal vector includes:

[0082] Construct an initial curvature tensor matrix based on the tangent and normal vectors of each trajectory point;

[0083] Select the forward nearest neighbor and backward nearest neighbor trajectory points of the current trajectory point in the local neighborhood space;

[0084] Calculate the angle between the tangent vector of the current trajectory point and the tangent vector of the nearest forward trajectory point, and the angle between the tangent vector of the nearest backward trajectory point, to obtain the forward tangent angle and the backward tangent angle respectively. At the same time, calculate the angle between the normal vector of the current trajectory point and the normal vector of the nearest forward trajectory point, and the angle between the normal vector of the nearest backward trajectory point, to obtain the forward normal angle and the backward normal angle respectively.

[0085] The tangential coupling component is obtained by weighted summation of the forward tangential angle and the backward tangential angle, and the normal coupling component is obtained by weighted summation of the forward normal angle and the backward normal angle. The tensor coupling coefficient is calculated based on the tangential coupling component and the normal coupling component.

[0086] The initial curvature tensor matrix of the current trajectory point is weighted and fused with the initial curvature tensor matrices of the forward nearest neighbor trajectory points and the backward nearest neighbor trajectory points according to the tensor coupling coefficient to obtain the curvature tensor matrix of the current trajectory point.

[0087] In one specific implementation, for each trajectory point, an initial curvature tensor matrix is ​​constructed by calculating its direction of motion and the normal vector perpendicular to the direction of motion. For any point P in the trajectory, assuming its tangent vector is T and its normal vector is N, the initial curvature tensor matrix M can be expressed as the outer product of T and N, that is, a second-order tensor jointly determined by the tangent vector and the normal vector.

[0088] To ensure the continuity and smoothness of the trajectory, the forward nearest neighbor P_forward and the backward nearest neighbor P_backward of the current trajectory point P are selected in the local neighborhood space. The selection of the nearest neighbor can be based on time series or spatial distance, and usually the point that is closest to the current point in time or distance is selected as the nearest neighbor.

[0089] Calculating the directional change between the current trajectory point and its nearest neighbor is an important indicator for measuring curvature change. Specifically, we calculate the angle θ_forward_T between the tangent vector T of the current point P and the tangent vector T_forward of its forward nearest neighbor P_forward, and the angle θ_backward_T between the tangent vector T of the current point P_forward and its backward nearest neighbor P_backward. These two angles represent the degree of change in the trajectory in the forward and backward directions from the current point, respectively.

[0090] Similarly, calculate the angle θ_forward_N between the normal vector N of the current point P and the normal vector N_forward of the forward nearest neighbor, and the angle θ_backward_N between the normal vector P and the normal vector N_backward of the backward nearest neighbor. These normal angles reflect the changes in the direction of the normal to the trajectory surface and are particularly important for describing the three-dimensional characteristics of the trajectory.

[0091] After obtaining this angle information, the tangential coupling component and the normal coupling component are calculated by weighted summation. The tangential coupling component C_T can be expressed as the weighted sum of the forward tangential angle and the backward tangential angle, i.e., C_T = w_forward_T × θ_forward_T + w_backward_T × θ_backward_T, where w_forward_T and w_backward_T are weighting coefficients that can be set according to the specific application scenario and are usually inversely proportional to the distance or time interval.

[0092] Similarly, the normal coupling component C_N can be expressed as the weighted sum of the forward normal angle and the backward normal angle, i.e., C_N = w_forward_N × θ_forward_N + w_backward_N × θ_backward_N, where w_forward_N and w_backward_N are the corresponding weight coefficients.

[0093] The tensor coupling coefficient α is calculated based on the tangential coupling component C_T and the normal coupling component C_N. The tensor coupling coefficient reflects the correlation of the current point with its nearest neighbor in terms of curvature characteristics. It can be obtained by mapping C_T and C_N using a nonlinear function; for example, the sigmoid function can be used to map the coupling components to (0, 1, 2, 3). ， 1) The interval is represented as α = sigmoid(β_T × C_T + β_N × C_N), where β_T and β_N are adjustment parameters. The tensor coupling coefficient α is a scalar with a value range of (0,1), obtained by mapping a linear combination of the tangential coupling component C_T and the normal coupling component C_N using the sigmoid function. This tensor coupling coefficient considers the degree of change in both the tangential and normal directions, where β_T and β_N control the contribution weights of tangential and normal changes to the final coupling effect, respectively. When the tangential and normal changes are large, the corresponding coupling component values ​​increase, and after mapping by the sigmoid function, the α value is close to 1, indicating a strong curvature correlation between the current point and its neighboring points; conversely, when the tangential and normal changes are small, the α value is close to 0, indicating that the curvature characteristics of the current point are mainly determined by itself. The tensor coupling coefficient can adaptively adjust the degree of fusion of the curvature tensor according to the local geometric characteristics of the trajectory, achieving a smooth transition while maintaining the original characteristics of the trajectory.

[0094] The initial curvature tensor matrix of the current trajectory point is weighted and fused with the initial curvature tensor matrices of the forward and backward nearest neighbors. Specifically, if the initial curvature tensor matrix of the current point is M, the forward nearest neighbor is M_forward, and the backward nearest neighbor is M_backward, then the fused curvature tensor matrix M_final can be expressed as: M_final = (1-α) × M + α × (w_forward × M_forward + w_backward × M_backward), where w_forward and w_backward are the weights of the forward and backward nearest neighbors, and satisfy w_forward + w_backward = 1.

[0095] The curvature tensor matrix constructed through the above steps can comprehensively describe the curvature characteristics of the trajectory at various points, providing an important foundation for subsequent trajectory optimization and smoothing. This method considers the local continuity of the trajectory and integrates the curvature information of the nearest neighbor points, making the constructed curvature tensor more accurately reflect the geometric characteristics of the trajectory.

[0096] In one alternative implementation ， Based on vehicle location information and path gradient information, a potential energy change curve is constructed. From the potential energy change curve, the braking energy recovery range and the climbing power consumption range are identified, including:

[0097] Based on vehicle location information and path slope information, the location coordinates and slope angles of each road segment in the path are obtained. The slope angles are organized into a slope sequence according to the road segment order. The locations in the slope sequence where the slope change rate exceeds the preset change rate are marked. The slope sequence is segmented at the marked locations. Potential energy sub-curves are constructed for each segment of the slope sequence. At the connection between adjacent potential energy sub-curves, a transition curve segment is generated through an interpolation algorithm. The potential energy sub-curves and the transition curve segment are spliced ​​together to form a complete potential energy change curve.

[0098] A sliding window is set on the potential energy change curve. By moving the sliding window on the potential energy change curve, local potential energy features are extracted. Window regions in the local potential energy features where the potential energy decreases by more than a preset recovery threshold and the duration exceeds a preset recovery distance are merged into braking energy recovery intervals. Window regions in the local potential energy features where the potential energy increases by more than a preset consumption threshold and the duration exceeds a preset climbing distance are merged into climbing power consumption intervals. The interval between the braking energy recovery interval and the climbing power consumption interval is marked as a transition interval.

[0099] In one specific implementation, the mining dump truck obtains its current location information, including three-dimensional coordinate data such as longitude, latitude, and altitude, through an onboard positioning device. Combined with a pre-established high-precision map database of the mining area, the location coordinates and slope angle information of each segment along the driving path are extracted. Location coordinates are expressed in meters, and slope angles are expressed in degrees or as a percentage.

[0100] The extracted slope angles are organized into a slope sequence according to the road segment order. The slope sequence reflects the elevation change characteristics of the mining dump truck's travel path, providing basic data for constructing potential energy change curves. On the mining area's transportation roads, the slope sequence includes slope data for road segments with different characteristics, such as gentle sections, steep uphill sections, and downhill sections.

[0101] Points in the slope sequence where the rate of slope change exceeds a preset rate of change are marked. The preset rate of change is typically set to 1.0° to 2.0° per 10 meters. The rate of slope change is calculated by dividing the difference in slope between two adjacent points by the distance between the two points. When a rate of slope change exceeding the preset value is detected, the point is marked as a significant point of slope change. Significant points of slope change usually appear at the transition points between gentle road sections and steep slopes; these locations are key points where the path characteristics change significantly.

[0102] The slope sequence is segmented at the marked locations to form multiple slope subsequences. Each slope subsequence represents a relatively gentle slope change in the path, facilitating the subsequent construction of potential energy sub-curves. The segmentation operation decomposes the path into regions with similar slope characteristics, reducing the difficulty of processing complex paths.

[0103] Potential energy sub-curves are constructed for each segment of the slope sequence. These sub-curves reflect the change in potential energy of a vehicle on its corresponding road segment. The calculation method involves converting the slope value into a height change and then combining this with the vehicle's mass to calculate the change in potential energy. The potential energy sub-curves visually reflect the energy characteristics of the road segment; an ascending curve represents an uphill segment that requires energy consumption, while a descending curve represents a downhill segment where energy can be recovered.

[0104] At the junctions of adjacent potential energy sub-curves, an interpolation algorithm is used to generate transition curve segments, ensuring the smoothness and continuity of the overall potential energy change curve. Commonly used interpolation algorithms include cubic spline interpolation, which can generate transition curves with continuous derivatives at the junction points. The smoothness of the transition curve is crucial for subsequent sliding window analysis, avoiding misjudgments caused by abrupt changes in the curve.

[0105] The potential energy sub-curve and transition curve segments are spliced ​​together according to the path sequence to form a complete potential energy change curve. This complete curve visually reflects the potential energy change characteristics of the mining dump truck throughout its entire journey, providing a data foundation for subsequent interval identification. Each point on the curve corresponds to a location on the path, and the vertical axis represents the potential energy value at that location.

[0106] A sliding window is set on the potential energy change curve, with the window width typically set to 100 to 500 meters. The sliding step size is set to 10 to 50 meters, which is the distance the window moves each time. The window size selection needs to consider the characteristics of the mining area's roads; an overly large window may ignore short-distance characteristics, while an overly small window may be overly sensitive to local changes.

[0107] Local potential energy features are extracted by moving a sliding window along the potential energy change curve. For each window position, characteristic parameters such as the potential energy change amplitude, duration, and rate of change are calculated. The potential energy change amplitude is the difference between the maximum and minimum potential energy values ​​within the window; the duration is the distance the elevation changes continuously within the window; and the rate of change is the potential energy change amplitude divided by the duration.

[0108] Window regions where the potential energy reduction exceeds a preset recovery threshold and persists for a distance exceeding a preset recovery distance are merged into braking energy recovery intervals. The preset recovery threshold and recovery distance are set based on the characteristics of mining dump trucks and the features of mining area roads. When adjacent windows all meet the braking energy recovery conditions, these window regions are merged into a continuous braking energy recovery interval. These intervals represent road sections where braking energy can be effectively recovered when mining dump trucks are traveling downhill.

[0109] Window regions where the potential energy increase exceeds a preset consumption threshold and the duration exceeds a preset climbing distance are merged into a climbing power consumption interval. The preset consumption threshold and climbing distance are also set based on vehicle and road characteristics. When adjacent windows all meet the climbing power consumption conditions, these window regions are merged into a continuous climbing power consumption interval. These intervals represent road sections where mining dump trucks require significant power support when traveling uphill.

[0110] The transition zone between the regenerative braking range and the power consumption range for climbing should be clearly marked. This transition zone is typically a flat road section with minimal potential energy change, meeting neither the regenerative braking condition nor the power consumption condition for climbing. Clearly marking the transition zone helps the hybrid power system smoothly switch between different operating conditions, avoiding control instability caused by frequent switching.

[0111] By using path potential energy feature recognition methods, mining dump trucks can identify key energy zones along their travel paths in advance, providing predictive information for energy allocation in the supercapacitor hybrid power system. In practical applications, the system formulates differentiated energy management strategies based on the identification results: before approaching the braking energy recovery zone, the supercapacitor's state of charge is reduced in advance to reserve capacity for the upcoming energy recovery; before approaching the climbing power consumption zone, the supercapacitor's state of charge is appropriately increased to prepare for the upcoming high power demand.

[0112] The method in this embodiment is particularly suitable for fixed-route transportation scenarios in mining areas. By accurately identifying path potential energy characteristics, it significantly improves the energy recovery efficiency and discharge efficiency of supercapacitors, enabling refined management and optimized control of energy distribution in hybrid power systems. Practical applications show that the energy distribution strategy based on path potential energy characteristic identification can effectively reduce fuel consumption of mining dump trucks and improve overall energy utilization efficiency.

[0113] In one alternative implementation ， The constraint requirements of the supercapacitor's state of charge for each interval are calculated using the backpropagation algorithm, resulting in the target state of charge sequence, which includes:

[0114] Establish a reverse traversal sequence from the end of the path. Extract the energy consumption demand value of the climbing power consumption interval and the energy recovery capacity value of the braking energy recovery interval in the reverse traversal sequence. Convert the energy consumption demand value into the state of charge demand of the corresponding position and the energy recovery capacity value into the state of charge compensation of the corresponding position.

[0115] A reverse accumulation and transfer mechanism for state of charge (SOC) demand is established. The SOC demand at the current position is subtracted from the SOC compensation provided by the subsequent regenerative braking range to obtain the SOC demand transferred to the previous position. During the transfer process, the SOC demand is subject to supercapacitor capacity boundary constraints. When the SOC demand exceeds the capacity limit, the excess portion is allocated to the engine. When the SOC demand is below the capacity limit, an additional charging command is inserted.

[0116] After the reverse cumulative transfer is completed, the charge state requirements at each position are arranged in forward order to generate the target charge state sequence.

[0117] In one specific implementation, information such as elevation and distance of the mining dump truck's travel path is acquired. The entire path is divided into several discrete segments based on gradient changes. Each segment has a defined start and end point, gradient value, and corresponding distance. A reverse traversal sequence starting from the end point of the path is constructed, recording the location identifier, elevation value, and cumulative travel distance of each discrete point. In the reverse traversal sequence, segments are classified into two types based on gradient: climbing power consumption intervals and braking energy recovery intervals. Segments with a gradient greater than zero are identified as climbing power consumption intervals, while segments with a gradient less than zero are identified as braking energy recovery intervals.

[0118] For the uphill power consumption range, the energy consumption demand during the uphill process is calculated based on the weight of the mining dump truck, the slope of the road section, and the travel speed. For example, when the mining dump truck weighs 120 tons, the slope of the uphill section is 8%, and the travel distance is 500 meters, the calculated energy consumption demand is approximately 15 kWh. This energy consumption demand is converted into the change in state of charge (SOC) of the supercapacitor. Assuming the total capacity of the supercapacitor is 20 kWh, the corresponding SOC change for this section is 75%.

[0119] For the braking energy recovery range, the recoverable energy value during braking is calculated. Considering energy recovery efficiency, the actual recoverable energy value is less than the theoretical maximum. For example, when the mining dump truck has a mass of 120 tons, a downhill slope of -6%, a travel distance of 400 meters, and an energy recovery efficiency of 60%, the calculated energy recovery capacity is approximately 8 kWh. This energy recovery capacity value is converted into the supercapacitor state-of-charge compensation change. Assuming the total supercapacitor capacity is 20 kWh, the corresponding state-of-charge compensation change for this section is 40%.

[0120] A reverse cumulative transfer mechanism for state of charge (SOC) demand is established, calculating the SOC demand value for each location point sequentially from the end point to the beginning point of the path. During the reverse transfer process, the SOC demand value at the current location point equals the sum of the change in SOC at the current location and the SOC demand value transferred to the current location, minus the SOC compensation value provided by the subsequent regenerative braking energy recovery interval.

[0121] During the transfer process, capacity boundary constraints are set for the supercapacitor's state of charge (SOC), including an upper and lower limit. The upper limit is typically set at 95%, and the lower limit at 20%, to ensure the supercapacitor operates within its optimal range and extends its lifespan. When the accumulated SOC demand exceeds the upper limit, the excess energy is allocated to the engine. For example, if the calculated SOC demand at a certain location is 98%, exceeding the 95% upper limit, the excess 3% energy demand will be provided by the engine, and the engine's auxiliary power supply command and corresponding energy value at that location are recorded.

[0122] When the calculated state of charge (SOC) requirement is lower than the lower capacity limit, it indicates that the supercapacitor's energy is insufficient to support subsequent driving needs, requiring an additional charging command. For example, if the calculated SOC requirement at a certain location is 15%, which is lower than the lower capacity limit of 20%, then an engine charging command is inserted at that location to raise the supercapacitor's SOC to a set safety threshold, such as 50%. The charging command and corresponding charging amount at that location are recorded.

[0123] After the reverse accumulation and transfer is completed, the target state of charge value of the supercapacitor, the engine auxiliary power supply command, and the additional charging command corresponding to each discrete point on the path are obtained. This information is then rearranged in the forward driving order to generate a complete target state of charge sequence.

[0124] For example, assume the driving path of a mining dump truck contains 10 discrete points, P1 to P10 from the starting point to the ending point. The reverse traversal sequence is P10 to P1. Calculations show that P8 to P10 is a downhill section, belonging to the braking energy recovery range, with energy recovery capacity values ​​of 2 kWh, 3 kWh, and 4 kWh respectively; P5 to P7 is a flat section, with an energy consumption demand of 1 kWh for each point; P1 to P4 is an uphill section, belonging to the climbing power consumption range, with energy consumption demand values ​​of 5 kWh, 4 kWh, 3 kWh, and 2 kWh respectively.

[0125] Starting from P10, the reverse cumulative calculation is performed: the state of charge (SOC) requirement at P10 is 30% of the preset value at the endpoint; at P9, 4 kWh of energy is recovered, which is converted to a SOC compensation of 20%, and the SOC requirement is 30% - 20% = 10%, which is lower than the lower limit of 20%, so a charging command is inserted to adjust the SOC to 50%; at P8, 3 kWh of energy is recovered, which is converted to a SOC compensation of 15%, and the SOC requirement is 50% - 15% = 35%; at P7, 1 kWh of energy is consumed, which is converted to a change in SOC requirement of 5%, and the SOC requirement is 35% + 5% = 40%; and so on to complete the SOC calculation for all locations.

[0126] After the reverse accumulation and transfer is completed, the required state of charge (SOC) values ​​at each location are arranged in forward order from P1 to P10 to obtain the target SOC sequence. This sequence serves as the control basis for energy distribution in the supercapacitor hybrid power system of the mining dump truck, providing real-time guidance for the supercapacitor discharge depth and engine operating status, thereby achieving efficient energy utilization.

[0127] In this embodiment, the supercapacitor hybrid power system of the mining dump truck can plan the energy distribution strategy in advance according to the path characteristics, make full use of the braking energy recovered on the downhill section, reduce unnecessary engine work, reduce fuel consumption, improve overall energy utilization efficiency, and achieve the goal of energy conservation and emission reduction.

[0128] like Figure 2 The diagram shown illustrates the reverse traversal flowchart of the energy management strategy.

[0129] In one alternative implementation ，The basic power component is allocated to the engine, and the transient power component is allocated to the supercapacitor. A particle swarm optimization algorithm is used, with the deviation between the supercapacitor's state of charge information and the target state of charge sequence as the optimization objective, to perform rolling optimization, obtaining the power allocation correction amount. The final power allocation command includes:

[0130] The base power component is allocated to the engine to obtain the engine's base power value. ， The transient power components are distributed to the supercapacitor to obtain the transient power value of the supercapacitor;

[0131] Based on the engine's base power value and the supercapacitor's transient power value, the predicted evolution trajectory of the supercapacitor's state of charge information is calculated. ， Extract the target state of charge (SPO) values ​​at the corresponding time points from the target SPO sequence. ， Calculate the state-of-charge tracking deviation between the predicted evolution trajectory and the target state-of-charge value;

[0132] Initialize the particle positions and velocities of the particle swarm ， Decode the particle position into a power distribution ratio adjustment coefficient. ， The power distribution ratio adjustment coefficient is used to proportionally adjust the engine's base power value and the supercapacitor's transient power value. ， The deviation between the supercapacitor's state of charge information and the target state of charge sequence was recalculated based on the adjusted power value. ， The particle fitness value is obtained by weighting the deviation with the violation of the supercapacitor capacity constraint and the violation of the engine power change rate constraint. ， The particle position and velocity are iteratively updated based on the particle fitness value until convergence. ， Extract the power allocation ratio adjustment coefficient corresponding to the global optimal particle position to determine the power allocation correction amount;

[0133] The power distribution correction is used to adjust the engine's base power value and the supercapacitor's transient power value. ， Generate the final power allocation instruction.

[0134] In one specific implementation, the power demand of the hybrid power system is decomposed, with the base power component allocated to the engine and the transient power component allocated to the supercapacitor. The base power component typically refers to the relatively stable power demand portion, which can be extracted from the total power demand using a low-pass filter. The transient power component refers to the power demand portion with larger fluctuations, which can be obtained by subtracting the base power component from the total power demand. Through this allocation method, the engine is primarily responsible for providing stable power output, while the supercapacitor is used to cope with rapid changes in power demand, fully leveraging the respective advantages of both power sources.

[0135] Based on the engine's base power value and the supercapacitor's transient power value, the evolution trajectory of the supercapacitor's state of charge (SOC) is predicted. The evolution of the supercapacitor's SOC can be calculated iteratively using the current SOC value and the power change. Assume the current SOC is SOC(t), and the supercapacitor's power is P. sc (t), the capacitance of the supercapacitor is C sc Then the state of charge at the next moment can be expressed as SOC(t+1) = SOC(t) - P sc (t)×Δt / C sc In this way, the trend of supercapacitor state of charge (SOC) changes over a future period can be predicted. Simultaneously, the target SOC values ​​for corresponding moments are extracted from a pre-defined target SOC sequence, and the deviation between the predicted and target values ​​is calculated.

[0136] To optimize power allocation, a particle swarm optimization (PSO) algorithm is used to adjust the power distribution. PSO is a swarm intelligence optimization algorithm that searches for the optimal solution to a problem by simulating the foraging behavior of bird flocks. In this scheme, the particle swarm is first initialized, with each particle representing a set of power allocation ratio adjustment coefficients. The particle positions can be randomly generated within the range [0.8]. ， Within the range of 1.2, this represents the adjustment ratio of the base power allocation result; the particle velocity can be randomly initialized within [-0.1]. ， Within the range of 0.1, it represents the rate of change of the particle's position.

[0137] After decoding the particle positions into power distribution ratio adjustment coefficients, these coefficients are used to adjust the engine's base power value and the supercapacitor's transient power value. The adjusted engine power is P. eng ' = P eng × τ, the supercapacitor power is P sc ' = P sc × (2-τ), where τ is the power distribution ratio adjustment coefficient. This adjustment method changes the power distribution ratio between the engine and the supercapacitor while keeping the total power demand constant.

[0138] Based on the adjusted power value, the predicted evolution trajectory of the supercapacitor's state of charge (SOC) information is recalculated and compared with the target SOC sequence to obtain the SOC tracking deviation. Simultaneously, it is checked whether the adjusted power allocation violates constraints, including the supercapacitor capacity constraint (SOC should be kept within the allowable range) and the engine power change rate constraint (engine power change should not exceed the maximum allowable change rate). The SOC tracking deviation and constraint violation amounts are weighted and combined to obtain the particle's fitness value.

[0139] Particle swarm optimization (PSO) iteratively updates the position and velocity of particles, continuously searching for solutions with lower fitness values. The particle position update formula is: x(t+1) = x(t) + v(t+1), where x represents the particle position and v represents the particle velocity. The particle velocity update formula is: v(t+1) = w·v(t) + c1·r1·(pbest-x(t)) + c2·r2·(gbest-x(t)), where w is the inertia weight, c1 and c2 are learning factors, and r1 and r2 are [0, 0, 1 ... ， The random number between 1], where pbest is the best historical position of the particle and gbest is the best global position.

[0140] The algorithm stops iterating when the number of iterations reaches a preset maximum or the fitness value meets the convergence condition. The power allocation ratio adjustment coefficient corresponding to the globally optimal particle position is extracted, and the power allocation correction is calculated. The engine power correction is ΔP. eng = P eng × (τ-1), the power correction for the supercapacitor is ΔP sc = P sc × (1-(2-τ)) = P sc ×(τ-1).

[0141] The initial power allocation result is corrected using a power allocation correction factor. The corrected engine power is P. eng_final = P eng + ΔP eng The supercapacitor power is P sc_final = P sc + ΔP sc The corrected power value is then sent as the final power allocation instruction to the corresponding controller for execution.

[0142] By using the above-mentioned rolling optimization method based on particle swarm optimization, the state of charge of the supercapacitor can be made as close as possible to the preset target trajectory while meeting the constraints, thereby improving the energy management efficiency and service life of the hybrid power system.

[0143] In one alternative implementation ， Based on the engine's base power value and the supercapacitor's transient power value, the predicted evolution trajectory of the supercapacitor's state of charge information is calculated, including:

[0144] Obtain the current state of charge value of the supercapacitor, determine the prediction starting point, determine whether the supercapacitor is in charging or discharging mode based on the transient power value of the supercapacitor, and extract the charging efficiency curve or discharging efficiency curve corresponding to the current state of charge value.

[0145] Substitute the transient power value of the supercapacitor into the charging efficiency curve or discharging efficiency curve to calculate the actual charging and discharging power of the supercapacitor. Calculate the change in the state of charge of the supercapacitor based on the actual charging and discharging power of the supercapacitor. Add the current state of charge value and the change in the state of charge of the supercapacitor to obtain the state of charge value at the next moment.

[0146] Determine whether the charge interval to which the state of charge value belongs changes at the next moment. If the charge interval changes, extract the charging efficiency curve or discharging efficiency curve corresponding to the state of charge value at the next moment. Use the updated charging efficiency curve or discharging efficiency curve to recalculate the actual charging and discharging power of the supercapacitor and the change in the state of charge of the supercapacitor.

[0147] Within the preset prediction time domain, the state of charge value of the next time step is calculated iteratively, and the state of charge values ​​of each time step are arranged in chronological order to generate the predicted evolution trajectory.

[0148] In one specific implementation, when predicting the state of charge (SOC) of a supercapacitor, the current SOC value of the supercapacitor is obtained through an onboard monitoring device, and the current moment is set as the prediction starting point. The SOC value of a supercapacitor is expressed as the percentage of its current charge to its rated charge, typically ranging from 0% to 100%. In hybrid power systems for mining dump trucks, the initial SOC value of the supercapacitor is usually set between 30% and 70% to ensure sufficient capacity for energy recovery while reserving enough energy to support short-term high-power output demands.

[0149] The current transient power value of the supercapacitor is obtained, and its current operating condition is determined based on the sign of the transient power. When the transient power value is greater than 0, it indicates that the supercapacitor is in a discharging state, providing energy to the vehicle's power system; when the transient power value is less than 0, it indicates that the supercapacitor is in a charging state, recovering or receiving energy from the braking system or engine. During downhill braking of a mining dump truck, the transient power value of the supercapacitor can reach -200 kW to -50 kW; during uphill acceleration, the transient power value of the supercapacitor is typically in the range of 50 kW to 200 kW.

[0150] Based on the current state of charge (SOC) and operating conditions, the corresponding charging or discharging efficiency curves are extracted from a pre-established efficiency curve database. Supercapacitor efficiency curves reflect energy conversion efficiency at different power points and are typically obtained through experimental testing. For example, when the supercapacitor is at 60% SOC and in charging mode, the charging efficiency curve corresponding to 60% SOC is extracted from the efficiency curve database; if it is in discharging mode, the discharging efficiency curve corresponding to 60% SOC is extracted.

[0151] Substituting the transient power value of the supercapacitor into the corresponding efficiency curve, the actual charging and discharging power of the supercapacitor is calculated. Assuming the transient power value of the supercapacitor is -100 kW (charging condition), and the current state of charge is 60%, the charging efficiency curve shows that the charging efficiency at this power point is 92%. Therefore, the actual charging power of the supercapacitor is -100 kW multiplied by 92%, which equals -92 kW. Similarly, if the transient power is 150 kW (discharging condition), the discharge efficiency curve shows that the discharge efficiency at this power point is 95%. Therefore, the actual discharge power of the supercapacitor is 150 kW divided by 95%, which is approximately 158 kW.

[0152] Based on the calculated actual charging and discharging power of the supercapacitor, the change in the state of charge (SOC) of the supercapacitor within a preset time step is determined. With a set time step of 1 second and a total supercapacitor capacity of 5 kWh (18 MJ), when the actual charging power is -92 kWh, the increase in SOC within 1 second is 92 kWh multiplied by 1 second divided by 18 MJ, approximately 0.51%. When the actual discharging power is 158 kWh, the decrease in SOC within 1 second is 158 kWh multiplied by 1 second divided by 18 MJ, approximately 0.88%.

[0153] The current state of charge (SOC) value is added to the calculated change in SOC to obtain the SOC value at the next moment. For example, if the current SOC is 60%, the SOC value at the next moment under charging conditions is 60% plus 0.51%, which equals 60.51%; under discharging conditions, the SOC value at the next moment is 60% minus 0.88%, which equals 59.12%.

[0154] It is necessary to determine whether the state of charge (SOC) value has changed in the next time step. Supercapacitors are typically divided into multiple SOC intervals, each corresponding to a different charge / discharge efficiency curve. When the SOC changes from one interval to another, the efficiency curve needs to be updated. If the current SOC is 79%, and the calculated SOC for the next time step is 80.2%, the SOC interval changes from the medium SOC interval to the high SOC interval. In this case, the charge / discharge efficiency curve corresponding to the high SOC interval needs to be extracted.

[0155] When the charge range changes, the actual charging and discharging power and the change in state of charge of the supercapacitor are recalculated using the updated charging efficiency curve or discharging efficiency curve. For example, if the supercapacitor changes from a medium charge range of 79% to a high charge range of 80.2%, and the charging efficiency drops from 92% to 90%, the actual charging power needs to be recalculated as -100 kW multiplied by 90%, which equals -90 kW. The corresponding increase in state of charge becomes 0.5%.

[0156] Within a preset prediction time domain, the above calculation steps are executed iteratively to calculate the state of charge (SOC) value of the supercapacitor at each moment. The prediction time domain is typically set to 5 to 30 seconds, with a time step of 0.1 to 1 second. For example, if the prediction time domain is set to 10 seconds and the time step is 1 second, the calculation needs to be iterated 10 times to obtain the SOC value at each moment within the next 10 seconds. Considering the changing operating conditions of the mining dump truck, the transient power value will be dynamically updated according to the vehicle's driving status during the prediction process, for example, changing from -80 kW on an uphill section to 120 kW on a downhill section.

[0157] The predicted state of charge (SOC) values ​​at each time point are arranged in chronological order to generate the predicted SOC evolution trajectory of the supercapacitor. An example of the predicted evolution trajectory under a typical driving scenario for a mining dump truck is as follows: Initially, the SOC is 50%. As the vehicle enters an uphill section, the supercapacitor continuously discharges to support power demand, and the predicted SOC drops to 42% after 10 seconds. Conversely, when the vehicle enters a downhill section, the supercapacitor receives regenerative braking energy, with an initial SOC of 30%, and the predicted SOC rises to 38% after 10 seconds.

[0158] The accuracy of the predicted evolution trajectory was verified by comparing actual test data with predicted data. Under the complex road conditions of mining dump trucks, the error between the predicted state of charge (SOC) value and the actual value did not exceed 2%, meeting the requirements of engineering applications. By predicting the SOC evolution trajectory, the hybrid power system of mining dump trucks can determine in advance whether the supercapacitor capacity will meet future driving needs and adjust the energy distribution strategy in a timely manner. For example, when the predicted evolution trajectory shows that the supercapacitor's SOC will drop below 20% in 5 seconds, the engine output power is increased in advance to charge the supercapacitor, avoiding insufficient power due to insufficient supercapacitor energy.

[0159] The supercapacitor state of charge prediction method is applicable to various working environments and driving conditions. Especially in the complex working environment of mining dump trucks that frequently go up and down slopes, the accurate prediction results provide key decision-making basis for the energy management controller of the hybrid power system, significantly improving the energy utilization efficiency of the hybrid power system of mining dump trucks.

[0160] The supercapacitor hybrid power energy distribution optimization system for mining dump trucks according to embodiments of the present invention includes:

[0161] The data acquisition unit is used to acquire information on the power demand of mining dump trucks, the state of charge of supercapacitors, the path gradient, and the vehicle location.

[0162] The power decomposition unit is used to reconstruct the attractor trajectory from the phase space of the power demand information. By calculating the curvature tensor eigenvalues ​​of the attractor trajectory and mapping them back to the time domain, the transient power components and the fundamental power components are obtained.

[0163] The target planning unit is used to construct the potential energy change curve based on vehicle location information and path slope information, identify the braking energy recovery interval and the climbing power consumption interval from the potential energy change curve, and calculate the constraint requirements of each interval on the supercapacitor's state of charge through the backpropagation algorithm to obtain the target state of charge sequence.

[0164] The optimization control unit is used to allocate the basic power component to the engine and the transient power component to the supercapacitor. The particle swarm optimization algorithm is used to perform rolling optimization with the deviation between the supercapacitor's state of charge information and the target state of charge sequence as the optimization target to obtain the power allocation correction amount and generate the final power allocation command.

[0165] The execution output unit is used to send the final power distribution command to the engine control unit and the supercapacitor control unit for execution.

[0166] A third aspect of the present invention provides an electronic device, comprising:

[0167] processor;

[0168] Memory used to store processor-executable instructions;

[0169] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0170] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0171] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0172] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing energy distribution in a supercapacitor hybrid power system for mining dump trucks, characterized in that: include: Obtain information on the power requirements of mining dump trucks, the state of charge of supercapacitors, path gradient, and vehicle location. ； By reconstructing the attractor trajectory from the phase space of the power demand information, and by calculating the curvature tensor eigenvalues ​​of the attractor trajectory and mapping them back to the time domain, the transient power components and the fundamental power components are obtained. ； Based on vehicle location and path gradient information, a potential energy change curve is constructed. The braking energy recovery interval and the climbing power consumption interval are identified from this curve. The constraint requirements of each interval on the supercapacitor's state of charge are calculated using a backpropagation algorithm, resulting in the target state of charge sequence. ； The basic power component is allocated to the engine, and the transient power component is allocated to the supercapacitor. The particle swarm optimization algorithm is used to perform rolling optimization with the deviation between the supercapacitor's state of charge information and the target state of charge sequence as the optimization objective, so as to obtain the power allocation correction amount and generate the final power allocation command. The final power distribution command is sent to the engine control unit and the supercapacitor control unit for execution.

2. The method according to claim 1, characterized in that, The attractor trajectory is constructed by reconstructing the phase space of the power demand information. By calculating the curvature tensor eigenvalues ​​of the attractor trajectory and mapping them back to the time domain, the transient power components and the fundamental power components are obtained, including: The power demand information is reconstructed into phase space according to the embedding dimension and time delay. Attractor trajectories are formed in the reconstructed phase space. The attractor trajectories are discretized and sampled to obtain a sequence of trajectory points. A local neighborhood space is constructed for each trajectory point in the sequence of trajectory points. The tangent vector and normal vector of the trajectory points are extracted in the local neighborhood space. The curvature tensor matrix of the trajectory points is constructed based on the tangent vector and normal vector. The curvature tensor matrix is ​​decomposed into eigenvalues ​​to obtain a sequence of curvature tensor eigenvalues. Gradient calculation is performed on the curvature tensor eigenvalue sequence to obtain the curvature gradient sequence. Based on the curvature gradient sequence, the positions where the gradient magnitude exceeds the preset abrupt change threshold are identified, and the power abrupt change points are determined. Using the power abrupt change points as boundaries, the attractor trajectory is divided into multiple trajectory segments. The mean curvature tensor eigenvalues ​​of each trajectory segment are calculated. The trajectory segments with the mean curvature tensor eigenvalues ​​greater than the geometric stability threshold are mapped back to the time domain through inverse phase space transformation to obtain the transient power components. The trajectory segments with the mean curvature tensor eigenvalues ​​less than the geometric stability threshold are mapped back to the time domain through inverse phase space transformation to obtain the fundamental power components.

3. The method according to claim 2, characterized in that, Constructing the curvature tensor matrix of the trajectory points based on the tangent vector and normal vector includes: Construct an initial curvature tensor matrix based on the tangent and normal vectors of each trajectory point; Select the forward nearest neighbor and backward nearest neighbor trajectory points of the current trajectory point in the local neighborhood space; Calculate the angle between the tangent vector of the current trajectory point and the tangent vector of the nearest forward trajectory point, and the angle between the tangent vector of the nearest backward trajectory point, to obtain the forward tangent angle and the backward tangent angle respectively. At the same time, calculate the angle between the normal vector of the current trajectory point and the normal vector of the nearest forward trajectory point, and the angle between the normal vector of the nearest backward trajectory point, to obtain the forward normal angle and the backward normal angle respectively. The tangential coupling component is obtained by weighted summation of the forward tangential angle and the backward tangential angle, and the normal coupling component is obtained by weighted summation of the forward normal angle and the backward normal angle. The tensor coupling coefficient is calculated based on the tangential coupling component and the normal coupling component. The initial curvature tensor matrix of the current trajectory point is weighted and fused with the initial curvature tensor matrices of the forward nearest neighbor trajectory points and the backward nearest neighbor trajectory points according to the tensor coupling coefficient to obtain the curvature tensor matrix of the current trajectory point.

4. The method according to claim 1, characterized in that, Based on vehicle location information and path gradient information, a potential energy change curve is constructed. From the potential energy change curve, the braking energy recovery range and the climbing power consumption range are identified, including: Based on vehicle location information and path slope information, the location coordinates and slope angles of each road segment in the path are obtained. The slope angles are organized into a slope sequence according to the road segment order. The locations in the slope sequence where the slope change rate exceeds the preset change rate are marked. The slope sequence is segmented at the marked locations. Potential energy sub-curves are constructed for each segment of the slope sequence. At the connection between adjacent potential energy sub-curves, a transition curve segment is generated through an interpolation algorithm. The potential energy sub-curves and the transition curve segment are spliced ​​together to form a complete potential energy change curve. A sliding window is set on the potential energy change curve. By moving the sliding window on the potential energy change curve, local potential energy features are extracted. Window regions in the local potential energy features where the potential energy decreases by more than a preset recovery threshold and the duration exceeds a preset recovery distance are merged into braking energy recovery intervals. Window regions in the local potential energy features where the potential energy increases by more than a preset consumption threshold and the duration exceeds a preset climbing distance are merged into climbing power consumption intervals. The interval between the braking energy recovery interval and the climbing power consumption interval is marked as a transition interval.

5. The method according to claim 1, characterized in that, The constraint requirements of the supercapacitor's state of charge for each interval are calculated using the backpropagation algorithm, resulting in the target state of charge sequence, which includes: Establish a reverse traversal sequence from the end of the path. Extract the energy consumption demand value of the climbing power consumption interval and the energy recovery capacity value of the braking energy recovery interval in the reverse traversal sequence. Convert the energy consumption demand value into the state of charge demand of the corresponding position and the energy recovery capacity value into the state of charge compensation of the corresponding position. A reverse accumulation and transfer mechanism for state of charge (SOC) demand is established. The SOC demand at the current position is subtracted from the SOC compensation provided by the subsequent regenerative braking range to obtain the SOC demand transferred to the previous position. During the transfer process, the SOC demand is subject to supercapacitor capacity boundary constraints. When the SOC demand exceeds the capacity limit, the excess portion is allocated to the engine. When the SOC demand is below the capacity limit, an additional charging command is inserted. After the reverse cumulative transfer is completed, the charge state requirements at each position are arranged in forward order to generate the target charge state sequence.

6. The method according to claim 1, characterized in that, The basic power component is allocated to the engine, and the transient power component is allocated to the supercapacitor. A particle swarm optimization algorithm is used, with the deviation between the supercapacitor's state of charge information and the target state of charge sequence as the optimization objective, to perform rolling optimization, obtaining the power allocation correction amount. The final power allocation command includes: The base power component is allocated to the engine to obtain the engine's base power value. ， The transient power components are distributed to the supercapacitor to obtain the transient power value of the supercapacitor; Based on the engine's base power value and the supercapacitor's transient power value, the predicted evolution trajectory of the supercapacitor's state of charge information is calculated. ， Extract the target state of charge (SPO) values ​​at the corresponding time points from the target SPO sequence. ， Calculate the state-of-charge tracking deviation between the predicted evolution trajectory and the target state-of-charge value; Initialize the particle positions and velocities of the particle swarm ， Decode the particle position into a power distribution ratio adjustment coefficient. ， The power distribution ratio adjustment coefficient is used to proportionally adjust the engine's base power value and the supercapacitor's transient power value. ， The deviation between the supercapacitor's state of charge information and the target state of charge sequence was recalculated based on the adjusted power value. ， The particle fitness value is obtained by weighting the deviation with the violation of the supercapacitor capacity constraint and the violation of the engine power change rate constraint. ， The particle position and velocity are iteratively updated based on the particle fitness value until convergence. ， Extract the power allocation ratio adjustment coefficient corresponding to the global optimal particle position to determine the power allocation correction amount; The power distribution correction is used to correct the engine's base power value and the supercapacitor's transient power value, generating the final power distribution command.

7. The method according to claim 6, characterized in that, Based on the engine's base power value and the supercapacitor's transient power value, the predicted evolution trajectory of the supercapacitor's state of charge information is calculated, including: Obtain the current state of charge value of the supercapacitor, determine the prediction starting point, determine whether the supercapacitor is in charging or discharging mode based on the transient power value of the supercapacitor, and extract the charging efficiency curve or discharging efficiency curve corresponding to the current state of charge value. Substitute the transient power value of the supercapacitor into the charging efficiency curve or discharging efficiency curve to calculate the actual charging and discharging power of the supercapacitor. Calculate the change in the state of charge of the supercapacitor based on the actual charging and discharging power of the supercapacitor. Add the current state of charge value and the change in the state of charge of the supercapacitor to obtain the state of charge value at the next moment. Determine whether the charge interval to which the state of charge value belongs changes at the next moment. If the charge interval changes, extract the charging efficiency curve or discharging efficiency curve corresponding to the state of charge value at the next moment. Use the updated charging efficiency curve or discharging efficiency curve to recalculate the actual charging and discharging power of the supercapacitor and the change in the state of charge of the supercapacitor. Within the preset prediction time domain, the state of charge value of the next time step is calculated iteratively, and the state of charge values ​​of each time step are arranged in chronological order to generate the predicted evolution trajectory.

8. A supercapacitor hybrid power energy distribution optimization system for mining dump trucks, used to implement the method of any one of claims 1-7, characterized in that, include: The data acquisition unit is used to acquire information such as the power demand of the mining dump truck, the state of charge of the supercapacitor, the path gradient, and the vehicle location. ； The power decomposition unit is used to reconstruct the attractor trajectory from the phase space of the power demand information. By calculating the curvature tensor eigenvalues ​​of the attractor trajectory and mapping them back to the time domain, the transient power components and the fundamental power components are obtained. ； The target planning unit is used to construct a potential energy change curve based on vehicle location information and path slope information. It then identifies the braking energy recovery interval and the climbing power consumption interval from the potential energy change curve. Using a backpropagation algorithm, it calculates the constraint requirements of each interval on the supercapacitor's state of charge, obtaining the target state of charge sequence. ； The optimization control unit is used to allocate the basic power component to the engine and the transient power component to the supercapacitor. The particle swarm optimization algorithm is used to perform rolling optimization with the deviation between the supercapacitor's state of charge information and the target state of charge sequence as the optimization target to obtain the power allocation correction amount and generate the final power allocation command. The execution output unit is used to send the final power distribution command to the engine control unit and the supercapacitor control unit for execution.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images