New energy battery load optimization system and method based on adaptive algorithm

Abstract

The invention discloses a new energy battery load optimization system and method based on an adaptive algorithm, and relates to the technical field of battery control, and the system comprises a multi-source collection module, an analysis decision module and an optimization adjustment module. The multi-source acquisition module acquires driving state data, battery state data and road surface state data; the analysis decision-making module obtains the remaining capacity of the battery through a Kalman filtering algorithm in combination with the battery state data, reconstructs the road surface state data into a road surface three-dimensional model through a fusion three-dimensional reconstruction algorithm, and obtains the slope of the road surface through a sampling analysis method; the optimization adjustment module obtains the driving state data, the battery remaining capacity and the road slope, performs fuzzy reasoning and defuzzification according to a membership function and a fuzzy rule to determine the optimal battery output power, comprehensively evaluates the fuzzy reasoning benefit based on a reward function, optimizes the membership function parameter through reinforcement learning, and obtains the optimal battery output power; and intelligent optimization matching of the battery output power, the driving state and the driving road condition is realized.

Description

Technical Field

[0001] The present invention relates to the field of battery control technology, and in particular to a new energy battery load optimization system and method based on an adaptive algorithm. Background Art

[0002] At present, new energy batteries, as the core power source of electric two-wheelers and three-wheelers, are facing many challenges.

[0003] On the one hand, the limited battery capacity leads to insufficient mileage. On the other hand, the driving conditions are complex and diverse, involving frequent climbing and descending of urban roads and different road conditions. At the same time, users want the vehicle to have good power performance to meet the needs of acceleration and climbing, and also want the battery to have a long mileage to reduce the cost of use.

[0004] Traditional battery load optimization solutions only consider battery capacity to control output power, ignoring the impact of driving status and complex and changeable road conditions on battery output power. It is difficult to achieve intelligent matching of battery output power with driving status and road conditions, and there is room for optimization in battery life and riding experience. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention proposes a new energy battery load optimization system and method based on an adaptive algorithm to achieve intelligent battery output power regulation that comprehensively considers driving status, battery status and road status.

[0006] The technical solutions adopted to achieve the purpose of the present invention are: A new energy battery load optimization system based on an adaptive algorithm, comprising a multi-source acquisition module, an analysis and decision module, and an optimization and adjustment module; The multi-source acquisition module receives the acquisition signal and collects the driving status data, battery status data and road status data in real time; The analysis and decision module uses the Kalman filter algorithm to obtain the remaining battery power based on the battery system status model and battery status data. , using the fusion 3D reconstruction algorithm, the left-eye image of the road surface in the road state data Right eye image with road surface Integrate texture and color information into road surface laser point cloud Generate a 3D road model and obtain the road slope through sampling analysis and road roughness And determine the generation time of the next acquisition signal; Optimize and adjust the module to obtain driving status data and battery remaining power and road slope , perform fuzzy reasoning and defuzzification based on membership functions and fuzzy rules to determine the optimal battery output power , the benefits of fuzzy reasoning are evaluated in multiple aspects based on the reward function, and the membership function parameters are optimized online through reinforcement learning to improve the fuzzy reasoning performance.

[0007] Furthermore, the multi-source acquisition module includes a driving perception unit, a battery acquisition unit and a road surface monitoring unit; The driving sensor unit receives the collected signal and measures the driving speed through the Hall sensor, acceleration sensor and GPS locator. , driving acceleration and single trip distance ; The battery collection unit receives the collection signal and measures the output current through the Hall current sensor and voltmeter and terminal voltage ; The road monitoring unit receives the collected signal and obtains the left-eye image of the road surface through a multifunctional camera composed of a dual-view camera and a laser device. , Road right eye image and road surface laser point cloud .

[0008] Further, the analysis and decision module includes a battery estimation unit, a road surface analysis unit, and an adjustment decision unit; The battery estimation unit obtains the time Output current and terminal voltage , using the Kalman filter algorithm, based on the time Remaining battery power and battery system status model prediction time Estimated battery charge remaining , according to time The estimated error covariance of and observation noise Make corrections and obtain time Remaining battery power ; The road surface analysis unit uses a fusion 3D reconstruction algorithm to and road right eye image Match similar features and combine the internal parameter matrix to generate the road image point cloud , and road surface laser point cloud Perform registration and fusion to obtain a complete point cloud of the road surface , using improved Poisson reconstruction to complete the road point cloud Perform normal vector redirection and model it as a Gaussian Poisson problem to reconstruct and generate a three-dimensional model of the road surface; The adjustment decision unit samples the road surface three-dimensional model through sampling analysis method, finds the approximate plane that best matches the sampling point based on the least squares method, and determines the road surface slope. , select mathematical indicators to quantify the road bumpiness , based on road roughness Determine the generation time of the next acquisition signal, where mathematical indicators include standard deviation, mean error and Gaussian curvature.

[0009] Furthermore, the battery estimation unit uses a Kalman filter algorithm to obtain the first time Remaining battery power The specific steps include: Establish a battery system status model to determine the time Estimated battery charge remaining With time Remaining battery power The difference is equal to time To time The charge and discharge capacity plus the Gaussian distribution of time Process noise , where the charge and discharge capacity and output current There is a proportional relationship; Establish the observation equation, time Terminal voltage Remaining battery power The transformation relationship is obtained by pre-fitting or based on the observation transformation function of the electrochemical model. Adding Gaussian-distributed observation noise Establish; Superposition time The estimated error covariance of With time Process noise Get the time The forecast estimation error covariance of , the observation transformation function About the remaining battery power Derivative to obtain the Jacobian matrix ; Through the Jacobian matrix The first time The forecast estimation error covariance of Mapping to the observation space, the calculation can measure and estimate the remaining battery power With terminal voltage The modified contribution of time Kalman gain ; Through terminal voltage Estimating the remaining battery charge Observed transformation function Transformed mapping value The error, combined with the Kalman gain Correction to obtain time Remaining battery power , through the Kalman gain and the forecast estimate error covariance Update to get the time The estimated error covariance of .

[0010] Furthermore, the pavement analysis unit generates a pavement 3D model using a fusion 3D reconstruction algorithm, including the following specific steps: Based on the left-eye image of the road surface through the stable feature acceleration algorithm and road right eye image Establish the left eye scale space and the right eye scale space, find the local maximum and remove the weak response feature points to achieve feature point positioning, where the weak response feature point is the feature point whose Hessian matrix determinant value is less than the response threshold; The gradient histogram in the neighborhood of a single feature point is counted to determine the main direction of the single feature point, and sixteen area blocks around the single feature point are extracted. The amplitude information of the eight gradient directions in each area block is counted to construct the feature description of the single feature point. The distance measurement method is used to compare and match the feature descriptions of the feature points in the left eye scale space and the right eye scale space to generate multiple sets of matching point pairs, and the basic matrix with the largest number of matching point pairs that meet the geometric constraints is selected through a random sampling consistency algorithm; Calculate the internal parameter matrix, remove the influence of the internal parameter matrix on the basic matrix through bundle adjustment, obtain the external parameter matrix, and combine the triangulation principle to project the matching point pairs to generate the road surface image point cloud , through iterative closest point algorithm and road surface laser point cloud Perform registration and fusion to obtain a complete point cloud of the road surface ; Using improved Poisson reconstruction to reconstruct the complete point cloud of the road surface The normal vector in is redirected and the 3D reconstruction is regarded as a Gaussian Poisson problem, which is solved and restored to generate a 3D model of the road surface.

[0011] Furthermore, the stable feature acceleration algorithm includes the following specific steps: Calculate the Hessian matrix determinant of the road image at different scales to construct the corresponding scale space. The road image includes the left-eye image of the road and road right eye image ; Compare the Hessian matrix determinant values ​​of each pixel with those of its neighboring pixels in the scale space, find the pixel with the largest Hessian matrix determinant value in the neighborhood and regard it as a potential feature point; All potential feature points whose Hessian matrix determinant values ​​are less than the response threshold are eliminated, and the pixel positions of the retained feature points are corrected by interpolation based on the pixel distribution around the feature points using the sub-pixel analysis method.

[0012] Furthermore, the improved Poisson reconstruction is used to complete the point cloud of the road surface. Restoring the road surface 3D model includes the following specific steps: Smoothing the complete point cloud of the road surface by moving least squares method , and calculate the complete point cloud of the road surface through principal component analysis The normal vector and point cloud curvature of each point in the point cloud are taken as the origin and the positive direction is specified; Perform neighborhood search through KD tree, perform scalar product of the searched points with the origin in turn, reverse the direction of the normal vector of the point whose scalar product is less than 0, and mark the point that completes the scalar product; The last point in the neighborhood that completes the scalar product is used as the new origin, and a new round of neighborhood search is performed through the KD tree. The scalar product is continued with the unmarked point to reverse or maintain the normal vector, and the normal vector is marked. Until the road surface is complete point cloud All points are marked, normal vectors are redirected, and the complete point cloud of the road surface after redirection is obtained. Construct octree topology to make the road surface complete point cloud All points in fall into leaf nodes; The node functions of leaf nodes are constructed based on three-dimensional box filtering. All node functions are linearly combined through cubic spline interpolation to construct a vector field. The Gaussian Poisson problem is constructed and converted into a system of linear equations. The three-dimensional scalar field is obtained by solving it through the conjugate gradient method. The marching cubes algorithm is used to extract the isosurface from the three-dimensional scalar field to generate a three-dimensional road model.

[0013] Furthermore, the adjustment decision unit obtains the road slope through sampling analysis method , road bumpiness The generation time of the next acquisition signal includes the following specific steps: Random sampling is performed on the 3D road model to obtain sampling points, is the total number of sampling points; Define the approximate plane equation as The error function is constructed by minimizing the sum of squares of the distances from the sampling points to the approximate plane. ,in, , and are the coefficients of the equation of the approximate plane respectively; Solving for the error function about , and The partial derivative of and set it equal to 0, rewrite it into matrix form and solve it through matrix operations to obtain the equation coefficients of the approximate plane , and ; Determine the normal vector based on the approximate plane equation , then the road slope Equal to the normal vector and The tangent of the angle in the axial direction; The road bumpiness is defined by selecting a mathematical index that can reflect the degree of undulation from each point on the three-dimensional road model to the approximate plane. , mathematical indicators include The standard deviation of the distances of the sampling points relative to the approximate plane, The local slope and road slope of the sampling point The average error and The average Gaussian curvature of the sampling points; Comparison of road bumps and the bumpiness threshold And adaptively determine the collection interval , the time when the next acquisition signal is generated is equal to the time when the current acquisition signal is generated plus the acquisition interval .

[0014] Furthermore, the road bumpiness is compared and the bumpiness threshold And adaptively determine the collection interval , including the following specific steps: Determine the road surface roughness Is it greater than or equal to the bumpiness threshold? ; If greater than or equal to the bumpiness threshold , calculate the road bumpiness and the bumpiness threshold The difference is calculated by using the bumpy acquisition relationship function to calculate the acquisition interval. , the dependent variable of the bump collection relationship function is the collection interval , the independent variable is road roughness and the bumpiness threshold The difference shows a monotonically decreasing trend and the value range is greater than or equal to the minimum collection interval ; If the road is bumpy Less than the bumpiness threshold , then the collection interval Road bumpiness Equal to the bumpiness threshold The bumpy collection relationship function value at that time.

[0015] Further, the optimization and adjustment module includes a fuzzy reasoning unit and an online optimization unit; The fuzzy reasoning unit selects the appropriate membership function according to the characteristics of fuzzy input and fuzzy output during the offline training stage, establishes fuzzy rules and determines the membership function parameters based on domain experience, and calculates the driving speed according to the membership function. , driving acceleration , Remaining battery power and road slope For the membership of fuzzy sets, linguistic fuzzy reasoning (Mamdani reasoning) is performed based on fuzzy rules and the optimal battery output power is generated by defuzzification using the centroid method. ; Online optimization unit obtains driving speed , driving acceleration , Single driving distance and remaining battery charge , a reward function is designed to evaluate the benefits of fuzzy reasoning by comprehensively considering battery energy consumption, driving safety and driving comfort, and reinforcement learning is used to online optimize the membership function parameters based on the reward function to improve the fuzzy reasoning performance.

[0016] Furthermore, in the offline training phase, the fuzzy sets of fuzzy input and fuzzy output are defined and the corresponding membership functions are selected, including: Define driving speed The speed fuzzy set includes low speed , medium speed and high speed , and the driving speed The degree and stability of belonging to a specific speed fuzzy set in a specific speed range are both high, and the trapezoidal membership function is selected to describe the speed fuzzy set; Defining driving acceleration The acceleration fuzzy set includes negative acceleration , zero acceleration and positive acceleration , and the driving acceleration For negative acceleration or positive acceleration The membership degree is high and stable, only the driving acceleration For zero acceleration near 0 The membership degree of shows a decreasing trend from 0 to both sides, and the trapezoidal membership function is selected to describe the negative acceleration. and positive acceleration , choose the triangular membership function to describe zero acceleration ; Define the remaining battery charge The battery fuzzy set includes low battery , China Electric Power and high power , due to the remaining battery power During driving, it presents a monotonically decreasing trend with time, and a unilateral Sigmoid membership function is selected to describe the fuzzy set of electric quantity; Define the road slope The slope fuzzy set includes downhill , flat road and uphill , due to the slope of the road in the plain area Most of them are close to flat roads, and the road slope in mountainous areas is There are also corresponding standards. The greater the road slope, the The smaller the probability of occurrence, the Gaussian membership function is selected to describe the slope fuzzy set; Define battery output power The power fuzzy set includes low power , Medium Power and high power , since the battery output power The transition between different power fuzzy sets is smooth and the battery output power ranges of different models are different. A bell-shaped membership function with a clear center, adjustable width and smooth curve is selected to describe the power fuzzy set.

[0017] Specifically, the fuzzy rules are designed based on domain experience and the relationship between the four fuzzy inputs is combined to determine the output power of the decision battery. , and transform the decision-making thinking mode into computer-processable rules. A total of 81 fuzzy rules are established, which map 81 input combinations to one of the three fuzzy sets corresponding to the fuzzy outputs.

[0018] Furthermore, the fuzzy reasoning unit performs Mamdani reasoning based on fuzzy rules and determines the optimal battery output power by defuzzification using the centroid method. The specific steps include: Get driving speed , driving acceleration , Remaining battery power and road slope ; Traverse the eighty-one fuzzy rules, and for a single fuzzy rule, determine the four fuzzy sets in the corresponding input combination, and convert the driving speed , driving acceleration , Remaining battery power and road slope Calculate the membership of the corresponding fuzzy sets respectively and take the minimum membership as the activation strength of the single fuzzy rule ; Activation intensity according to a single fuzzy rule Modify the membership function of the power fuzzy set corresponding to the fuzzy rule, and synthesize the power membership functions after the 81 rules are modified to generate a synthetic power membership function ; The composite power membership function is defuzzified by using the centroid method. With battery output power The area enclosed by the independent variable coordinate axis is regarded as a plane object, and the horizontal coordinate value of the center of gravity of the plane object is the optimal battery output power , where the synthetic power membership function The area of ​​the horizontal axis is the composite power membership function About battery output power 's points.

[0019] Furthermore, the design of the reward function comprehensively considers battery energy consumption, driving safety and driving comfort, including the following specific steps: Get the time Driving speed , driving acceleration , Single driving distance and remaining battery charge and time Driving acceleration and remaining battery charge ; Consider the time To time The power consumption per unit distance, that is, the power consumption Single trip distance The smaller the power consumption per unit distance, the higher the battery energy consumption reward. time The battery energy consumption reward function Set it to the exponential form of the reciprocal of the power consumption per unit distance; Consider the time Driving speed and driving acceleration Do not exceed the upper limit of safe speed and the upper limit of safe acceleration , when time Driving speed and driving acceleration Less than the upper limit of safety speed and the upper limit of safe acceleration When time Driving safety reward function Set to the inverse of the logarithmic function so that the driving safety reward value increases with the driving speed and driving acceleration decreases with the increase of speed and approaches the upper limit of safe speed and the upper limit of safe acceleration is more sensitive when time Driving speed Greater than or equal to the upper limit of safe speed and / or driving acceleration Greater than or equal to the upper limit of safe acceleration When the penalty factor is introduced The Sigmoid function is combined with the transformation form to make the penalty increase faster when the speeding degree is greater; Calculate the time To time The acceleration change value , when time To time The acceleration change value Exceeding the upper limit of acceleration change When the driving comfort reward is reduced, the comfort value changes with the acceleration. The gradual increase and smooth decrease of time Driving comfort reward function Set to the absolute change value of acceleration Upper limit of acceleration change The percentage error is in the form of a hyperbolic tangent function of the independent variable; No. time The reward function Equal to the battery energy consumption reward function , Driving safety reward function And the driving comfort reward function The linear weight and the linear weight determine the importance ratio of battery energy consumption, driving safety and driving comfort.

[0020] Specifically, reinforcement learning defines states and environmental states, where the states include driving speed , driving acceleration , Single driving distance and remaining battery charge , the environmental state is the road slope , defining the action space as the adjustable battery output power The action is defined as adjusting all the membership function parameters in fuzzy reasoning, and the reward function is taken as the Q value of the deep Q network with experience replay mechanism. The deep Q network is based on a set of environmental states and the probability distribution of state output actions. The proximal policy optimization algorithm is used to select different actions according to the probability distribution of actions, and the impact of different actions is evaluated in parallel based on the Q value. The optimal action is determined by truncated importance sampling, so that the Q value is continuously improved and eventually converges.

[0021] A new energy battery load optimization method based on an adaptive algorithm is implemented based on a new energy battery load optimization system based on an adaptive algorithm, and includes the following specific steps: Using Kalman filter algorithm, based on the battery system state model combined with the time The collected battery status data is obtained time Remaining battery power ; Using the fusion 3D reconstruction algorithm, time Road surface laser point cloud in road condition data On the basis of time Left-eye image of the road surface captured Right eye image of the road The texture information and color information of time The three-dimensional model of the road surface is obtained by sampling analysis method. time The road slope and road roughness And adjust the collection interval ; Using reinforcement learning to evaluate the reward function in multiple aspects time The cycling benefits brought by fuzzy reasoning are used to adjust the membership function parameters using the proximal strategy optimization algorithm; Based on time Collected driving status data, battery remaining power and road slope , fuzzy reasoning and defuzzification are performed through membership functions and fuzzy rules to determine the time Optimal battery output power .

[0022] Compared with the prior art, the present invention has the significant advantage that the driving status data, battery status data and road status data collected in real time are analyzed by the Kalman filtering algorithm and the fusion three-dimensional reconstruction algorithm to obtain accurate battery remaining power and road surface three-dimensional model, analyze the road surface three-dimensional model to adaptively determine the collection interval and infer the road surface slope, perform fuzzy reasoning and defuzzification on the driving status data, battery remaining power and road surface slope according to the membership function and fuzzy rules, determine the optimal battery output power suitable for the current road conditions and driving requirements, and evaluate the benefits of fuzzy reasoning in multiple aspects based on the reward function, optimize the membership function parameters online through reinforcement learning, improve the fuzzy reasoning performance, and ensure that the electric two-wheeled vehicle and / or electric three-wheeled vehicle always takes into account both battery life and riding experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A schematic diagram of a new energy battery load optimization system based on an adaptive algorithm in the present invention; Figure 2 This is a flowchart of the improved Poisson reconstruction in the present invention; Figure 3 is the fuzzy rule table established in the present invention; Figure 4 This is a flow chart of a new energy battery load optimization method based on an adaptive algorithm in the present invention. DETAILED DESCRIPTION

[0024] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments.

[0025] Example 1

[0026] like Figure 1 As shown, a specific embodiment of the present invention discloses a new energy battery load optimization system based on an adaptive algorithm, including a multi-source acquisition module, an analysis and decision module, and an optimization and adjustment module; The multi-source acquisition module receives the acquisition signal and uses the intelligent sensor to collect driving status data, battery status data and road status data in real time. The driving status data includes driving speed , driving acceleration and single trip distance , battery status data includes output current and terminal voltage , the road surface status data includes the left-eye image of the road surface , Road right eye image and road surface laser point cloud , where the single trip distance is Refers to the travel distance of electric two-wheelers or electric three-wheelers from the last collection to the current collection; The analysis and decision module uses the Kalman filter algorithm to accurately obtain the remaining battery power based on the battery system status model and battery status data. , using a fusion 3D reconstruction algorithm to build a road surface laser point cloud As a benchmark, the left image of the road is fused Right eye image of the road The texture information and color information in the image are used to generate a three-dimensional model of the road surface, and the road slope is further obtained through sampling analysis. and road roughness , and based on the road roughness Determine the time to generate the next acquisition signal; Optimize and adjust the module to obtain driving status data and battery remaining power and road slope , fuzzy reasoning is performed based on membership functions and fuzzy rules, and the optimal battery output power is determined by defuzzification Adjustments are made and a reward function is designed to evaluate the driving benefits brought by fuzzy reasoning. Online learning is performed through reinforcement learning to optimize the membership function parameters and improve the fuzzy reasoning performance.

[0027] Furthermore, the multi-source acquisition module includes a driving perception unit, a battery acquisition unit and a road surface monitoring unit; The driving sensor unit receives the collected signal. The Hall sensor generates a pulse signal based on the Hall effect when the wheel rotates. The driving speed is obtained based on the product of the pulse signal frequency and the wheel circumference. The acceleration sensor measures the driving acceleration by sensing the charge signal characteristics generated when the piezoelectric material is deformed by force. The GPS locator records the driving trajectory and adds the recording point when receiving the acquisition signal, obtains the trajectory distance between the recording point and the previous recording point and uses it as the single driving distance ; The battery collection unit receives the collection signal and measures the output current through the Hall current sensor and voltmeter and terminal voltage ; The road monitoring unit receives the collected signal and obtains the left-eye image of the road surface through the multi-functional camera , Road right eye image and road surface laser point cloud The multifunctional camera is composed of a dual-view camera and a laser device. The dual-view camera is used to shoot the road ahead to obtain the left-eye image of the road. Right eye image with road surface The laser device emits a laser beam and receives the reflected laser signal to measure the distance between the laser beam and the road surface, thereby generating a road surface laser point cloud. .

[0028] Further, the analysis and decision module includes a battery estimation unit, a road surface analysis unit, and an adjustment decision unit; The battery estimation unit obtains the time Output current and terminal voltage And the Kalman filter algorithm is used for analysis. time Remaining battery power and battery system status model prediction time Estimated battery charge remaining , according to time The estimated error covariance of and observation noise Make corrections and obtain time Remaining battery power ; The road surface analysis unit is based on road surface laser point cloud The 3D reconstruction of the road surface has problems of holes and texture loss. The fusion 3D reconstruction algorithm is used to and road right eye image Match similar features, calculate the internal parameter matrix and generate the road surface image point cloud , and with road surface laser point cloud Perform registration and fusion to obtain a complete point cloud of the road surface , using improved Poisson reconstruction, the complete point cloud of the road surface The 3D reconstruction is regarded as a Gaussian Poisson problem to reconstruct and generate a 3D model of the road surface; The adjustment decision unit randomly samples the road surface three-dimensional model through the sampling analysis method, finds the approximate plane with the smallest sum of square distances from the sampling points through the least squares method, and determines the road surface slope. , and select appropriate mathematical indicators to quantify the road bumpiness , based on road roughness and the bumpiness threshold The size relationship determines the generation time of the next acquisition signal, where the mathematical indicators include standard deviation, mean error and Gaussian curvature.

[0029] Furthermore, the battery estimation unit uses a Kalman filter algorithm to obtain the first time Remaining battery power The specific steps include: Build a battery system state model to describe the remaining battery capacity The state changes of the battery, considering the remaining battery power during the battery charging and discharging process With output current There is a linear correlation between time Estimated battery charge remaining With time Remaining battery power The difference is approximately equal to time To time The charge and discharge capacity of time To time The power supply output power is not adjusted, so the output current remains unchanged by default and is always equal to the time Output current , then time To time The charge and discharge capacity of time Output current The specific formula is as follows: ; in, and are the rated capacity and charge / discharge efficiency of the battery respectively. For the time To time The collection interval, For the time The process noise describes the uncertainty of the battery charging and discharging process. The mean is 0 and the variance is Gaussian distribution of The observation equation is established to describe the first time Terminal voltage Remaining battery power The transformation relationship is as follows: ,in, For information about the remaining battery power The observation transformation function can be obtained by fitting the pre-observed data or determined based on the electrochemical model of the battery. For the time The observation noise describes the error in the measurement process, The mean is 0 and the variance is Gaussian distribution of Superposition time The estimated error covariance of With time Process noise As the time The forecast estimation error covariance of , the observation transformation function About the remaining battery power Derivative to obtain the Jacobian matrix , the Jacobian matrix Reflects the remaining battery power Opposite voltage the extent of the impact; Through the Jacobian matrix The first time The forecast estimation error covariance of Mapped to the observation space, calculate the time Kalman gain , Kalman gain To measure the time Estimated battery charge remaining With terminal voltage For the contribution of correction, time The forecast estimation error covariance of Greater than observation noise , indicating that the uncertainty of the estimate is greater than the uncertainty of the observation, and the Kalman gain Increase, time Terminal voltage More accurate, more contribution to the correction, when the time The forecast estimation error covariance of Less than the observation noise , indicating that the uncertainty of the estimate is less than the uncertainty of the observation, and the Kalman gain Reduce, time Estimated battery charge remaining The contribution to the revision is higher; is the Jacobian matrix The transpose of Combined with time Kalman gain With terminal voltage Corrected estimation of remaining battery power , get the time Remaining battery power To estimate the remaining battery power Substitute into the observation transformation function After mapping the value, through the terminal voltage With the mapping value The error between the estimated battery remaining capacity Correct the remaining battery power The error is smaller; According to time Kalman gain and the forecast estimate error covariance Update to get the time The estimated error covariance of ,in, represents the unit matrix, when the Kalman gain The smaller the time, the time The estimated error covariance of The smaller it is, the less uncertainty the estimate will have.

[0030] Furthermore, the pavement analysis unit generates a pavement 3D model using a fusion 3D reconstruction algorithm, including the following specific steps: Based on the left-eye image of the road surface through the stable feature acceleration algorithm and road right eye image Establish the corresponding left eye scale space and right eye scale space respectively, find the local maximum value of the scale space and remove the feature points with weak response to achieve sub-pixel feature point positioning; For a single feature point in the left eye scale space or the right eye scale space, the gradient histogram is counted in the neighborhood of the single feature point, and the direction that exceeds 0.8 times the maximum value of the gradient histogram is taken as the main direction of the single feature point. The determination of the main direction makes the single feature point rotationally invariant, that is, no matter the left eye image of the road surface Or road right eye image No matter how it rotates, the feature description in the main direction remains consistent; Extract the surrounding area of ​​a single feature point There are 16 blocks in total. The amplitude information of eight gradient directions in each block is counted to construct The vector of dimension is used as the feature description of a single feature point, and the feature description comprehensively reflects the gradient distribution information of a single feature point; The distance measurement method is used to compare and match the feature descriptions of feature points in the left eye scale space and the right eye scale space. That is, for a single feature point in the left eye scale space, the closest feature point and the second closest feature point are found in the right eye scale space and the distance ratio is calculated. If the distance ratio is less than the matching threshold, the closest feature point in the right eye scale space is matched with the single feature point in the left eye scale space to generate a matching point pair. If the distance ratio is greater than or equal to the matching threshold, it is considered a false match. Use the random sampling consensus algorithm to extract at least eight sets of matching point pairs and calculate the basic matrix, and use the basic matrix to verify the geometric constraints of the remaining matching point pairs, count the total number of matching point pairs that meet the geometric constraints, repeat this step multiple times, and select the basic matrix that makes the total number of matching point pairs that meet the geometric constraints the largest. The basic matrix is ​​used to describe the projection transformation relationship from the right eye perspective to the left eye perspective of the dual-view camera, and is used to unify the left eye scale space and the right eye scale space; The internal parameter matrix is ​​calculated based on the focal length of the left camera of the dual-view camera. The influence of the internal parameter matrix on the basic matrix is ​​removed by bundle adjustment, and the external parameter matrix of the dual-view camera is best estimated. The external parameter matrix describes the change of the position of the right camera relative to the left camera. The bundle adjustment is an existing algorithm. Based on the external parameter matrix, the matching point pairs are projected and transformed using the triangulation principle to calculate the three-dimensional coordinates of the matching point pairs in the world coordinate system to obtain the road surface image point cloud. , and further through the iterative closest point algorithm and road surface laser point cloud Perform registration and fusion to obtain a complete point cloud of the road surface ; Correcting the complete point cloud of the road surface using improved Poisson reconstruction The normal vector ambiguity problem in the point cloud is solved by selecting the point with the minimum curvature as the origin, searching for neighboring points through the KD tree and redefining the normal vector direction, and propagating the neighboring points as the origin to complete the normal vector redirection of the road surface. The 3D reconstruction is regarded as a Gaussian Poisson problem, which is solved and restored to generate a 3D model of the road surface.

[0031] Furthermore, the stable feature acceleration algorithm includes the following specific steps: Calculate the Hessian matrix determinant of the road image at different scales to construct the scale space. The Hessian matrix is ​​used to describe the second-order derivative information of the road image in different directions. The road image includes the left eye image of the road and road right eye image , the scale space includes the left eye scale space and the right eye scale space; In the scale space, the Hessian matrix determinant values ​​of each pixel of the road image are compared with those of the neighboring pixels, and the pixel with the largest Hessian matrix determinant value in the neighborhood is found and regarded as a potential feature point, where the neighborhood refers to a set of pixels that are adjacent to the pixel in scale and position; For potential feature points, all potential feature points whose Hessian matrix determinant values ​​are less than the response threshold are further eliminated. These weakly responding potential feature points may be caused by noise or interference, which has a negative impact on feature point matching and 3D reconstruction. The sub-pixel analysis method is used to correct the pixel position of the retained feature point based on the pixel distribution around the feature point by interpolation to achieve feature point positioning.

[0032] like Figure 2 As shown, further, the improved Poisson reconstruction is used to complete the road point cloud Restoring the road surface 3D model includes the following specific steps: The complete point cloud of the road surface is obtained by moving least squares method Smoothing is performed and the complete point cloud of the road surface is calculated using principal component analysis The normal vector and point cloud curvature of each point in; The point with the smallest curvature of the point cloud is taken as the origin, and the direction of the normal vector of the origin is defined as the positive direction; Perform neighborhood search through KD tree, perform scalar product of the searched points with the origin in turn. If the scalar product is less than 0, reverse the direction of the normal vector of the point. If the scalar product is greater than or equal to 0, keep the original normal vector direction and mark the point that completes the scalar product. The last point in the neighborhood that completes the scalar product is used as the new origin. A new round of neighborhood search is performed through the KD tree, and the already marked points are excluded. The scalar product is continued to reverse or maintain the direction of the normal vector, and the mark is made. Until the road surface is complete point cloud All points are marked, normal vectors are redirected, and the redirected road surface point cloud is further distributed through Poisson distribution. Construct octree topology to make the road surface complete point cloud All points in fall on leaf nodes of appropriate depth in the octree; The node functions of leaf nodes are constructed by using three-dimensional box filtering as basis functions, and all node functions are linearly combined by cubic spline interpolation to construct a vector field so that the vector field matches the normal vector of each point. The three-dimensional box filtering is an existing filter; The three-dimensional reconstruction is restated as a Gaussian Poisson problem, which is further transformed into a system of linear equations through the finite difference or finite element method, and solved by the conjugate gradient method to obtain a three-dimensional scalar field. The marching cube algorithm is further used to extract isosurfaces from the three-dimensional scalar field to form a three-dimensional road model. This step is a commonly used technical means in traditional Poisson reconstruction.

[0033] Furthermore, the adjustment decision unit obtains the road slope through sampling analysis method , road bumpiness The generation time of the next acquisition signal includes the following specific steps: Random sampling is performed on the 3D road model to obtain sampling points, is the total number of sampling points; Define the approximate plane equation , the goal is to find an approximate plane that best fits the overall trend of the road 3D model, that is The sum of the squares of the distances from the sampling points to the approximate plane is minimized, and the error function is constructed ,in, , and are the coefficients of the equation of the approximate plane, Respectively Sampling points axis, Axis and Axis value; Solve the error function separately about , and The partial derivative of and set it equal to 0, simplify and rewrite it into matrix form, as follows: ; Solve the equation coefficients of the approximate plane by matrix operation , and , where matrix operations include Gaussian elimination, LU decomposition, and QR decomposition; Normal vector of the approximate plane , then the normal vector of the approximate plane is and The tangent value of the angle between the two axes can be regarded as the road slope of the three-dimensional road model. , the road slope can be derived based on the trigonometric function relationship ; Defining road roughness It is an arbitrary mathematical index that can reflect the degree of ups and downs from each point on the three-dimensional road model to the approximate plane. The larger the mathematical index, the greater the degree of ups and downs, and the bumpier the three-dimensional road model is relative to the smooth approximate plane. The mathematical index includes The standard deviation of the distances of the sampling points relative to the approximate plane, The local slope and road slope of the sampling point The average error and The average Gaussian curvature of the sampling points; Comparison of road bumps and the bumpiness threshold And adaptively determine the collection interval , then the generation time of the next acquisition signal is equal to the generation time of the current acquisition signal plus the acquisition interval , where the bumpiness threshold According to the definition of road roughness The mathematical index selected at the time of acquisition is used to determine the generation time of the current acquisition signal. time For example, according to time The road slope The collection interval can be determined , then the next time the signal is collected, the time at which it is generated is time .

[0034] Furthermore, the road bumpiness is compared and the bumpiness threshold And adaptively determine the collection interval , including the following specific steps: Determine the road surface roughness Is it greater than or equal to the bumpiness threshold? ; If the road is bumpy Greater than or equal to the bumpiness threshold , calculate the road bumpiness and the bumpiness threshold The difference is calculated by using the bumpy acquisition relationship function to calculate the acquisition interval. , and the bump collection relationship function requires the collection interval About road bumps and the bumpiness threshold The difference shows a monotonically decreasing trend, that is, the bumpier the road, the longer the collection interval. The smaller the value, but it cannot be less than the minimum collection interval. The bump relationship function can be in various forms. In this embodiment, an inverse proportional function is selected. The specific formula is as follows: ; in, and are predetermined constants, which are respectively 2 and 0.4 in this embodiment; If the road is bumpy Less than the bumpiness threshold , the road surface 3D model is considered to be flat, then the collection interval Directly select road bumpiness Equal to the bumpiness threshold The bumpy collection relationship function value at the time, that is, the collection interval .

[0035] Further, the optimization and adjustment module includes a fuzzy reasoning unit and an online optimization unit; The fuzzy reasoning unit selects the corresponding membership function according to the characteristics of fuzzy input and fuzzy output in the offline training stage, establishes fuzzy rules from fuzzy input to fuzzy output based on domain experience, determines the membership function parameters, and obtains the driving speed. , driving acceleration , Remaining battery power and road slope , calculate the driving speed according to the membership function , driving acceleration , Remaining battery power and road slope For the membership of fuzzy sets, Mamdani reasoning is performed based on fuzzy rules, and the optimal battery output power is determined by defuzzification using the centroid method. ; Online optimization unit obtains driving speed , driving acceleration , Single driving distance and remaining battery charge , a reward function is designed to evaluate the driving benefits brought by fuzzy reasoning by comprehensively considering battery energy consumption, driving safety and driving comfort. Reinforcement learning is used to adaptively optimize the membership function parameters determined in the offline training phase based on the reward function. Through online learning, the fuzzy reasoning benefits are continuously improved to approach the upper limit of the benefits.

[0036] Furthermore, in the offline training stage, when the fuzzy input is the driving speed When the driving speed is defined The speed fuzzy set includes low speed , medium speed and high speed , and the speed of the electric two-wheeler and / or electric three-wheeler The degree and stability of belonging to the specific speed fuzzy set within the specific speed range are both high. The speed membership function selects the trapezoidal membership function, and the specific formula is as follows: ; in, and Low speed , medium speed and high speed The corresponding velocity membership function is, and are the speed fuzzy set boundary values ​​respectively. In this embodiment, the actual driving speed range of the electric two-wheeled vehicle and / or the electric three-wheeled vehicle is considered. and Take 15, 20, 30, 35 respectively, the unit is .

[0037] Furthermore, in the offline training stage, when the fuzzy input is the driving acceleration When the driving acceleration is defined The acceleration fuzzy set includes negative acceleration , zero acceleration and positive acceleration , and the driving acceleration of the electric two-wheeled vehicle and / or electric three-wheeled vehicle In the negative acceleration or positive acceleration The degree and stability of the driving acceleration are both high if and only if the driving acceleration In a small range near 0, it belongs to zero acceleration , and the membership degree is centered at 0 and decreases to both sides, so the driving acceleration Negative acceleration and positive acceleration The acceleration membership function selects the trapezoidal membership function, and the zero acceleration The acceleration membership function selects the triangle membership function, and the specific formula is as follows: ; in, and Negative acceleration , zero acceleration and positive acceleration The corresponding acceleration membership function is, and are acceleration fuzzy set boundary values ​​respectively. In this embodiment, the actual driving acceleration range of the electric two-wheeled vehicle and / or the electric three-wheeled vehicle is considered. and They are -0.5, -1, 0.5 and 1 respectively, and the unit is .

[0038] Furthermore, during the offline training phase, when the fuzzy input is the remaining battery power When the battery remaining The battery fuzzy set includes low battery , China Electric Power and high power , due to the remaining battery power During the driving of the electric two-wheeled vehicle and / or electric three-wheeled vehicle, the remaining battery power shows a monotonically decreasing trend over time. The electricity membership function selects the unilateral Sigmoid membership function, and the specific formula is as follows: ; ; ; in, and Low battery , China Electric Power and high power The corresponding electricity membership function is: and are the slope factors, which determine the steepness of the electricity membership function, and the coefficients and Low battery and high power The center value of the power range, and China Electric Power The center value and half of the width of the medium power range. In this embodiment, and The same value of 0.1 is used for both, indicating that the decay trend of the remaining battery power is the same in different power fuzzy sets. and Take separately and is the rated capacity of the battery.

[0039] Furthermore, in the offline training stage, when the fuzzy input is the road slope When the road slope is defined The slope fuzzy set includes downhill , flat road and uphill Based on the clear design standards and specifications in road design, the road slope in plain areas Most of the roads are flat, but the slope of the road in mountainous areas is There are also corresponding standards. The greater the road slope, the The smaller the probability of occurrence, the greater the road slope. The slope membership function selects the Gaussian membership function, and the specific formula is as follows: ; in, and Downhill , flat road and uphill The corresponding slope membership function is, and Downhill , flat road and uphill The typical slope in and Downhill , flat road and uphill The slope variance in determines the slope range of the slope fuzzy set. In this embodiment, and Take -0.05, 0 and 0.05 respectively, and Take 0.02, 0.01 and 0.02 respectively.

[0040] Furthermore, during the offline training phase, the fuzzy output is the battery output power , define the battery output power The power fuzzy set includes low power , Medium Power and high power , since the battery output power The transition between different power fuzzy sets is smooth, so the battery output power The power membership function selects a bell-shaped membership function with a clear center, adjustable width and smooth curve. The specific formula is as follows: ; in, and Low power , Medium Power and high power The corresponding power membership function is, and Low power , Medium Power and high power The center output power, and Low power , Medium Power and high power The width of the power membership function determines the power range of the power fuzzy set. and Low power , Medium Power and high power The shape coefficient of the medium power membership function, in this embodiment, and Take 120 watts, 350 watts and 650 watts respectively. and Take 30, 80 and 120 respectively, and Take 1.5, 1.2 and 1.8 respectively.

[0041] Specifically, in the offline training phase, fuzzy rules are designed considering domain experience to simulate human decision-making thinking and combine the mutual relationship between the four fuzzy inputs to determine the output power of the decision battery. , and convert the decision-making thinking mode into computer-processable rules. Since there are four fuzzy inputs, each of which is divided into three different fuzzy sets, the input combination of fuzzy reasoning is There are 81 fuzzy rules, which map the 81 input combinations to one of the three fuzzy sets corresponding to the fuzzy output. Figure 3 The fuzzy rule table is shown, where the first fuzzy rule is the fuzzy rule in the upper left corner of the fuzzy rule table , number them in the order of columns first and rows later, then the 81st fuzzy rule is the fuzzy rule in the lower right corner , with the first fuzzy rule For example, it shows that when the driving speed Belongs to low speed , driving acceleration Belongs to negative acceleration , remaining battery power Belongs to low battery And the road slope Belong to downhill When the corresponding battery output power Low power .

[0042] Furthermore, the fuzzy reasoning unit performs Mamdani reasoning based on fuzzy rules and determines the optimal battery output power by defuzzification using the centroid method. The specific steps include: Get driving speed , driving acceleration , Remaining battery power and road slope ; Traverse the 81 fuzzy rules determined in the offline training phase. For a single fuzzy rule, determine the four fuzzy sets in the input combination corresponding to the fuzzy rule. , driving acceleration , Remaining battery power and road slope Substitute the membership functions corresponding to their respective fuzzy sets to obtain four memberships and take the minimum membership as the activation intensity of a single fuzzy rule. , with the first fuzzy rule For example, the activation intensity corresponding to the first fuzzy rule is The details are as follows: ; Activation intensity according to a single fuzzy rule The membership function of the power fuzzy set corresponding to the modified fuzzy rule is taken as the first fuzzy rule For example, the activation intensity and low power of the first fuzzy rule The corresponding power membership functions are and , then the low power after the first fuzzy rule correction The corresponding power membership function ; The power membership functions modified by the eighty-one rules are combined to generate a synthetic power membership function ; The centroid method is used to defuzzify the membership function of the composite power. The principle of the centroid method is to transform the membership function of the composite power into With battery output power The area enclosed by the independent variable coordinate axis is regarded as a plane object with a certain mass distribution. The horizontal coordinate value corresponding to the center of gravity of the plane object is the optimal battery output power obtained after defuzzification. , since the battery output power is a continuous value within a specific range, the synthetic power membership function The area of ​​the horizontal axis is the composite power membership function About battery output power The integral within a specific range gives the optimal battery output power The specific formula is as follows: .

[0043] Furthermore, the design of the reward function comprehensively considers battery energy consumption, driving safety and driving comfort, including the following specific steps:

[0044] For time Driving speed , driving acceleration , Single driving distance and remaining battery charge , get the time Driving acceleration and remaining battery charge ; Set a battery energy reward, considering the time To time The power consumption per unit distance is equal to the power consumption Divide by the distance traveled per trip , the smaller the power consumption per unit distance, the time Optimal battery output power determined by fuzzy reasoning The more it helps to reduce energy consumption, the higher the corresponding battery energy consumption reward. time The battery energy consumption reward function Set it to the exponential form of the inverse of the power consumption per unit distance to further enhance the impact of battery energy consumption. The specific formula is as follows: ; in, is the energy consumption influencing factor; Set up driving safety rewards, consider time Driving speed and driving acceleration Do not exceed the upper limit of safe speed and the upper limit of safe acceleration , therefore, when time Driving speed and driving acceleration Less than the upper limit of safety speed and the upper limit of safe acceleration When the driving safety reward value increases with the driving speed and driving acceleration decreases with the increase of The closer you get to the upper limit of safe speed and / or driving acceleration The closer you get to the upper limit of safe acceleration , the faster the driving safety reward value decreases, the time Driving speed Greater than or equal to the upper limit of safe speed and / or driving acceleration Greater than or equal to the upper limit of safe acceleration When the driving safety reward is a large negative value as a penalty, time Driving speed and / or driving acceleration The greater the speeding degree, the faster the penalty increases, so as to drive the reinforcement learning to adjust the membership function parameters in the fuzzy reasoning to avoid speeding. time Driving safety reward function In the time Driving speed and driving acceleration Less than the upper limit of safety speed and the upper limit of safe acceleration When , it is set to the inverse form of the logarithmic function so that the driving safety reward value increases with the driving speed and driving acceleration decreases with the increase of speed, and for those approaching the upper limit of safety speed and the upper limit of safe acceleration is more sensitive when time Driving speed Greater than or equal to the upper limit of safe speed and / or driving acceleration Greater than or equal to the upper limit of safe acceleration When a larger penalty factor is introduced The design is combined with the inverse form of the Sigmoid function, so that the greater the speeding degree, the faster the penalty increases, as follows: ; in, and Respectively represent the "and" relationship and the "and / or" relationship; Set the driving comfort reward and calculate the time To time The acceleration change value , when time To time The acceleration change value Exceeding the upper limit of acceleration change When the electric two-wheeled vehicle and / or electric three-wheeled vehicle has suddenly accelerated or decelerated, the driving comfort reward is reduced, and the comfort value changes with the acceleration. The gradual increase of time Driving comfort reward function Designed to be based on the absolute change in acceleration Upper limit of acceleration change The percentage error is in the form of a hyperbolic tangent function of the independent variable. The hyperbolic tangent function has a smooth curve, as follows: ; The first time The reward function Equal to the battery energy consumption reward function , Driving safety reward function And the driving comfort reward function The linear weights determine the importance of battery energy consumption, driving safety and driving comfort during reinforcement learning optimization. In this embodiment, battery energy consumption and driving safety are given priority, and the linear weights of battery energy consumption, driving safety and driving comfort are set to 0.4, 0.4 and 0.2 respectively, that is, .

[0045] Specifically, reinforcement learning defines states and environmental states, where the states include driving speed , driving acceleration , Single driving distance and remaining battery charge , the environmental state is the road slope , defining the action space as the battery output power adjustable by fuzzy reasoning The action is defined as adjusting all membership function parameters in fuzzy reasoning, and the reward function is used as the Q value of the deep Q network. The deep Q network adopts the experience replay mechanism to store the environment state, state, action, and output function value in each decision-making process in the experience pool. When a set of environment state and state are input into the deep Q network, the probability distribution of the action is output. The proximal policy optimization algorithm is used as the policy gradient algorithm of the deep Q network. Different actions are selected according to the probability distribution of the action, and the influence of different actions on the Q value is evaluated in parallel based on the Q value. The optimal action is determined by truncated importance sampling, so that the Q value is continuously improved and eventually converges.

[0046] Example 2 like Figure 4 As shown, a new energy battery load optimization method based on an adaptive algorithm is implemented based on a new energy battery load optimization system based on an adaptive algorithm, and includes the following specific steps: In the time Collect and obtain driving status data, battery status data and road status data; Using Kalman filter algorithm, based on the battery system state model combined with the time Accurately obtain the battery status data time Remaining battery power ; Using the fusion 3D reconstruction algorithm, time Road surface laser point cloud in road condition data On the basis of time Left-eye image of the road surface captured Right eye image with road surface The texture information and color information of time The three-dimensional model of the road surface is obtained by sampling analysis method. time The road slope and road roughness And adjust the collection interval ; Reinforcement learning is used to evaluate the first vehicle from three aspects: battery energy consumption, driving safety and driving comfort through reward function. time The cycling benefits brought by fuzzy reasoning are used to adjust the membership function parameters using the proximal strategy optimization algorithm; Based on time Driving status data, battery remaining power and road slope , fuzzy reasoning and defuzzification are performed through membership functions and fuzzy rules to determine the time Optimal battery output power .

[0047] The invention discloses a new energy battery load optimization system and method based on an adaptive algorithm, comprising a multi-source acquisition module, an analysis and decision module and an optimization and adjustment module; the multi-source acquisition module acquires driving state data, battery state data and road state data; the analysis and decision module acquires the remaining battery power by combining the battery state data with a Kalman filter algorithm, reconstructs the road state data into a three-dimensional road model by adopting a fusion three-dimensional reconstruction algorithm, acquires the road slope and road bumpiness by a sampling analysis method, and adaptively adjusts the generation time of the next acquisition signal; the optimization and adjustment module acquires the driving state data, the remaining battery power and the road slope, performs fuzzy reasoning and defuzzification according to a membership function and fuzzy rules to determine the optimal battery output power, comprehensively evaluates the fuzzy reasoning benefits based on a reward function, optimizes the membership function parameters by reinforcement learning, and realizes intelligent optimization matching of the battery output power with the driving state and the driving road condition.

[0048] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A new energy battery load optimization system based on an adaptive algorithm, characterized in that: Including analysis and decision-making module and optimization and adjustment module; The analysis and decision module uses the Kalman filter algorithm, combined with the battery system status model and battery status data to obtain the remaining battery power, adopts the fusion 3D reconstruction algorithm, integrates the texture and color information of the left-eye image and the right-eye image of the road surface into the road surface laser point cloud to generate a 3D model of the road surface, obtains the road surface slope and road surface bumpiness through the sampling analysis method and determines the next acquisition time; The optimization and adjustment module obtains driving status data, remaining battery power and road slope, performs fuzzy reasoning based on membership functions and fuzzy rules, and determines the optimal battery output power through centroid method defuzzification, evaluates the fuzzy reasoning benefits through reward functions, and guides reinforcement learning to optimize membership function parameters.

2. A new energy battery load optimization system based on an adaptive algorithm as claimed in claim 1, characterized in that: The optimization and adjustment module includes a fuzzy reasoning unit and an online optimization unit; The fuzzy reasoning unit preselects a membership function according to the characteristics of the fuzzy input and the fuzzy output, establishes fuzzy rules and determines the membership function parameters, calculates the membership of the driving speed, driving acceleration, battery remaining power and road slope in the driving state data to the fuzzy set according to the membership function, performs Mamdani reasoning based on the fuzzy rules and generates the optimal battery output power by defuzzification through the centroid method; The online optimization unit obtains the driving speed, driving acceleration, single driving distance and remaining battery power in the driving status data, comprehensively considers battery energy consumption, driving safety and driving comfort to design a reward function to evaluate the fuzzy reasoning benefit, and guides reinforcement learning to optimize the membership function parameters online to improve the fuzzy reasoning performance.

3. A new energy battery load optimization system based on an adaptive algorithm as claimed in claim 2, characterized in that: The membership function is pre-selected according to the characteristics of the fuzzy input and the fuzzy output, including: The speed fuzzy set is divided into low speed, medium speed and high speed according to the predefined speed fuzzy set boundary value, and a trapezoidal membership function is selected to describe the speed fuzzy set; According to the predefined acceleration fuzzy set boundary value, the acceleration fuzzy set is divided into negative acceleration, zero acceleration and positive acceleration, and the trapezoidal membership function is selected to describe the negative acceleration and positive acceleration, and the triangle membership function is selected to describe the zero acceleration; The power fuzzy set is defined to include low power, medium power and high power, and the unilateral Sigmoid membership function is selected to describe the power fuzzy set. Define the slope fuzzy set including downhill, flat road and uphill, and choose Gaussian membership function to describe the slope fuzzy set; The power fuzzy set is defined as low power, medium power and high power, and a bell-shaped membership function is selected to describe the power fuzzy set.

4. A new energy battery load optimization system based on an adaptive algorithm as claimed in claim 2, characterized in that: The method of performing Mamdani reasoning based on fuzzy rules and defuzzifying the optimal battery output power by the centroid method includes the following specific steps: Traverse all fuzzy rules, and for a single fuzzy rule, determine the specific speed fuzzy set, acceleration fuzzy set, power fuzzy set and slope fuzzy set in the corresponding input combination, calculate the membership of the driving speed, driving acceleration, battery remaining power and road slope with the corresponding fuzzy set respectively, and take the minimum membership as the activation strength of the single fuzzy rule; According to the activation intensity of a single fuzzy rule, the membership function of the power fuzzy set corresponding to the fuzzy rule is modified, and the power membership functions after all fuzzy rules are modified are synthesized to generate a synthetic power membership function. The center of gravity method is used to defuzzify the synthetic power membership function, and the area enclosed by the synthetic power membership function and the corresponding independent variable coordinate axis is taken as a plane object. The horizontal coordinate value of the center of gravity of the plane object is the optimal battery output power.

5. A new energy battery load optimization system based on an adaptive algorithm as described in claim 1 or 2, characterized in that: The reward function is obtained according to battery energy consumption, driving safety and driving comfort, and includes the following specific steps: Get the power consumption per unit distance. According to the principle that the smaller the power consumption per unit distance, the higher the battery energy consumption reward, the battery energy consumption reward function is set to the exponential form of the inverse of the power consumption per unit distance. According to the requirement that both the driving speed and the driving acceleration cannot exceed the upper limit of the safe speed and the upper limit of the safe acceleration, when the driving speed and the driving acceleration are respectively less than the upper limit of the safe speed and the upper limit of the safe acceleration, the driving safety reward function is set to the inverse form of the logarithmic function, so that the driving safety reward value decreases with the increase of the driving speed and the driving acceleration. When the driving speed is greater than or equal to the upper limit of the safe speed and / or the driving acceleration is greater than or equal to the upper limit of the safe acceleration, a penalty factor is introduced to set the driving safety reward function in combination with the transformation form of the Sigmoid function; The acceleration change value is calculated, and the driving comfort decreases as the acceleration change value increases. The driving comfort reward function is set to a hyperbolic tangent function with the percentage error between the absolute acceleration change value and the upper limit of the acceleration change as the independent variable; The reward function is equal to the linear weighted sum of the battery energy consumption reward function, the driving safety reward function, and the driving comfort reward function.

6. A new energy battery load optimization system based on an adaptive algorithm as claimed in claim 1, characterized in that: The analysis and decision module includes a battery estimation unit, a road surface analysis unit and an adjustment decision unit; The battery estimation unit obtains the output current and terminal voltage in the battery status data at a single moment, adopts the Kalman filter algorithm, predicts the estimated battery remaining capacity at a single moment based on the battery remaining capacity at the previous moment and the battery system status model, and makes corrections based on the estimation error covariance and observation noise at a single moment to obtain the battery remaining capacity at a single moment; The pavement analysis unit adopts a fusion 3D reconstruction algorithm to match similar features from the left-eye image and the right-eye image of the pavement and generate a pavement image point cloud in combination with an internal reference matrix, and performs registration and fusion with the pavement laser point cloud to obtain a complete pavement point cloud. The improved Poisson reconstruction is used to redirect the normal vector of the complete pavement point cloud and model it as a Gaussian Poisson problem to reconstruct and generate a 3D pavement model. The adjustment decision unit samples the three-dimensional road model through the sampling analysis method, determines the approximate plane based on the least squares method and calculates the road slope and road bumpiness, and determines the generation time of the next acquisition signal based on the road bumpiness.

7. A new energy battery load optimization system based on an adaptive algorithm as claimed in claim 6, characterized in that: The method of obtaining the remaining battery power at a single moment includes the following specific steps: Establish a battery system state model to determine that the difference between the estimated battery remaining capacity at a single moment and the battery remaining capacity at the previous moment is equal to the charge and discharge capacity from the previous moment to a single moment plus the process noise at a single moment; The observation equation is established. The transformation relationship between the terminal voltage at a single moment and the remaining battery capacity is established by adding the observation transformation function to the observation noise at a single moment. The estimation error covariance of the previous moment is superimposed with the process noise of a single moment to obtain the prediction estimation error covariance of a single moment, and the Jacobian matrix is ​​obtained by taking the derivative of the observation transformation function with respect to the remaining battery power at a single moment; The prediction estimation error covariance at a single moment is mapped to the observation space through the Jacobian matrix, and the Kalman gain at a single moment is calculated; The remaining battery capacity at a single moment is obtained by combining the error between the terminal voltage at a single moment and the mapping value after the estimated remaining battery capacity is transformed by the observation transformation function with the Kalman gain correction at a single moment, and the estimated error covariance at a single moment is obtained based on the prediction estimation error covariance update.

8. A new energy battery load optimization system based on an adaptive algorithm as claimed in claim 6, characterized in that: The fusion 3D reconstruction algorithm includes the following specific steps: The left eye scale space and the right eye scale space are established based on the left eye image and the right eye image of the road surface by using a stable feature acceleration algorithm, local maximum values ​​are found and weak response feature points are eliminated to achieve feature point positioning, wherein the weak response feature point is a feature point whose Hessian matrix determinant value is less than the response threshold; The gradient histogram in the neighborhood of a single feature point is counted to determine the main direction of the single feature point, the area block around the single feature point is extracted and the amplitude information of the gradient direction in the area block is counted to construct the feature description of the single feature point; The distance measurement method is used to compare and match the feature descriptions of the feature points in the left eye scale space and the right eye scale space to generate multiple sets of matching point pairs, and the basic matrix with the largest number of matching point pairs that meet the geometric constraints is selected through a random sampling consistency algorithm; Calculate the internal parameter matrix, remove the influence of the internal parameter matrix on the basic matrix through bundle adjustment, obtain the external parameter matrix, and combine the triangulation principle to perform projection transformation on the matching point pairs to generate the road surface image point cloud, and obtain the complete road surface point cloud by registering and fusing it with the road surface laser point cloud through the iterative closest point algorithm; Improved Poisson reconstruction is used to redirect the normal vectors in the complete point cloud of the road surface, and the three-dimensional reconstruction is regarded as a Gaussian Poisson problem to solve and restore the three-dimensional model of the road surface.

9. A new energy battery load optimization system based on an adaptive algorithm as claimed in claim 8, characterized in that: The improved Poisson reconstruction comprises the following specific steps: The complete point cloud of the road surface is smoothed by the moving least square method, and the normal vector and point cloud curvature of each point in the complete point cloud of the road surface are calculated by the principal component analysis method. The point with the smallest curvature of the point cloud is taken as the origin and the positive direction is specified. Perform neighborhood search through KD tree, perform scalar product of multiple searched points with the origin in turn, reverse the direction of the normal vector of the point whose scalar product is less than 0, and mark the point that completes the scalar product; The last point in the neighborhood that completes the scalar product is used as the new origin, and a new round of neighborhood search is performed through the KD tree. The scalar product is continued with the unmarked point to reverse or maintain the normal vector, and the normal vector is marked. After all points in the complete road point cloud are marked, an octree topology is constructed so that all points in the complete road point cloud fall on leaf nodes. The node functions of leaf nodes are constructed based on three-dimensional box filtering. All node functions are linearly combined through cubic spline interpolation to construct a vector field. The Gaussian Poisson problem is constructed and converted into a system of linear equations. The three-dimensional scalar field is obtained by solving it through the conjugate gradient method. The marching cubes algorithm is used to extract the isosurface from the three-dimensional scalar field to generate a three-dimensional road model.

10. A new energy battery load optimization method based on an adaptive algorithm, characterized in that: The specific steps include: The Kalman filter algorithm is used to obtain the remaining battery power at a single moment based on the battery system state model and the battery state data collected at a single moment; A fusion 3D reconstruction algorithm is used to integrate the texture information and color information of the left and right images of the road surface taken at a single moment into the road surface laser point cloud collected at a single moment to generate a 3D model of the road surface. The road surface slope and road surface bumpiness at a single moment are obtained through sampling analysis and the collection interval is adjusted. Reinforcement learning is used to evaluate the riding benefits brought by the fuzzy reasoning at the previous moment in multiple aspects based on the reward function, and the proximal strategy optimization algorithm is used to adjust the membership function parameters; Based on the driving status data collected at a single moment, the remaining battery power at a single moment, and the road slope at a single moment, fuzzy reasoning and defuzzification are performed through membership functions and fuzzy rules to determine the optimal battery output power at a single moment.

Images