A PSO-based forklift low-voltage management optimization method

Abstract

The present application relates to a kind of PSO-based forklift low voltage management optimization method, comprising the following steps: data acquisition and preprocessing, algorithm model construction and initialization, parallel optimization calculation and automatic optimization, if judging optimization result is effective, then directly output, then parameter deployment is carried out, if it is invalid, then the weight of multi-objective fitness function is dynamically adjusted, and return parallel optimization stage.The PSO-based forklift low voltage management optimization method, by the ternary mapping model constructed accurately quantifies the coupling loss relationship of working condition, the quantitative relationship model of exponential temperature correction factor is integrated to realize the accurate characterization of temperature nonlinear characteristics, multi-objective function realizes demand balance, WOA-PSO efficiently optimizes core parameter, dynamic temperature compensation improves adaptability, engineering determination and targeted weight adjustment form full working condition closed loop optimization, battery safety and operation efficiency are considered, solve the forklift battery working condition coupling and temperature nonlinear characterization deficiency, the problem of single-objective optimization losing one and gaining another.

Description

Technical Field

[0001] This invention relates to the field of forklift battery management system technology, specifically a forklift low-voltage management optimization method based on PSO. Background Technology

[0002] The forklift battery management system prevents forklift batteries from overcharging, over-discharging, overheating, and short-circuiting, ensuring the safety of personnel and equipment. It includes a master control unit, slave control units, high-voltage control unit, low-voltage management, and communication network. Among them, the low-voltage management system sets low-voltage warning thresholds and low-voltage protection thresholds, which are two important lines of defense for protecting battery safety and lifespan. The low-voltage warning threshold means that when the battery voltage drops to this threshold, it means that the battery's safety margin is not much, and continued deep discharge will accelerate battery aging. Its core is preventative protection. The low-voltage protection threshold means that when the battery voltage continues to drop to this threshold, it means that the battery is on the verge of danger, and continued discharge will cause irreversible chemical damage. Its core is mandatory ultimate protection, with priority above all user operations.

[0003] Currently, in the low-voltage management of forklift battery management systems, low-voltage warning thresholds and low-voltage protection thresholds can be divided into static thresholds and dynamic thresholds. Static thresholds are set as fixed values ​​or segmented fixed values ​​based on operating conditions, according to the safe voltage range in the cell specifications and combined with limited laboratory test data. Dynamic thresholds, on the other hand, use fixed compensation coefficients or linear compensation to dynamically adjust the thresholds within a certain range. In practical use, if the low-pressure warning threshold and low-pressure protection threshold are dynamically adjusted using linear compensation, linear correction can be achieved based on a single or a few parameters. However, the coupling loss relationship between load, operating time, and capacity is ignored. This results in the effective available capacity decaying much faster than under no-load or light-load conditions within the same operating time. At the same time, there are also nonlinear effects of temperature and capacity, which cannot accurately characterize the complex characteristics of reduced battery activity and sudden decrease in available capacity at low temperatures, as well as the accelerated decay of capacity, even though it may increase briefly at high temperatures. Furthermore, linear compensation cannot accurately fit the coupling surface of load, operating time, and capacity, nor can it capture the exponential change curves of temperature and capacity. This causes the protection system to be either too sluggish and endanger safety, or too sensitive and affect operation. Therefore, a PSO-based forklift low-pressure management optimization method is proposed to solve the above problems. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a PSO-based optimization method for forklift low-pressure management. This method accurately captures the coupled surface of load, operating time, and capacity, as well as the exponential change curves of temperature and capacity. It solves the problem that if low-pressure warning and protection thresholds are dynamically adjusted using linear compensation, linear correction based on a single or few parameters is possible, but the coupled loss relationship between load, operating time, and capacity is ignored. This results in the effective usable capacity decaying much faster than under no-load or light-load conditions within the same operating time. Furthermore, the non-linear effects of temperature and capacity cannot accurately characterize the complex characteristics of reduced battery activity and sudden decrease in usable capacity at low temperatures, and the accelerated decay of capacity despite a possible brief increase at high temperatures. Linear compensation cannot accurately fit the coupled surface of load, operating time, and capacity, nor can it capture the exponential change curves of temperature and capacity. This leads to protection systems that are either too sluggish, endangering safety, or too sensitive, affecting operation.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a forklift low-pressure management optimization method based on PSO, comprising the following steps; S1. Collect raw data of forklift battery voltage V, ambient temperature T, battery capacity C, load L and working time t, then preprocess the data, establish a mapping model and fit a quantitative relationship model. S2. Build the WOA-PSO parallel hybrid fusion algorithm model, construct the multi-objective fitness function, and then initialize it; S3. Execute the optimized calculation process and automatically find the optimal solution; S4. Determine the optimization result of S3. If yes, output the optimization result; otherwise, dynamically adjust the weights and return to S3. S5. Based on the optimization results, the optimal threshold and compensation coefficient are obtained, and then the compensation voltage is calculated. The calculation logic of the optimal threshold, compensation coefficient and compensation voltage is embedded into the forklift BMS system to complete the deployment of optimization parameters.

[0006] Furthermore, the preprocessing includes cleaning, denoising, and normalizing the collected raw data. To obtain a clean dataset ; Each sample , ; Calculate the battery capacity decay rate for the i-th sample based on a clean dataset R. The calculation formula is as follows: ,in, This refers to the battery's rated capacity. The mapping model is established based on a clean dataset R and the battery capacity decay rate of the i-th sample. And a multivariate nonlinear regression was used to establish the load. Homework duration Battery capacity The ternary mapping model is expressed as follows: ; ; in, , , , , , The fitting parameters were obtained based on a clean dataset R. The fitted quantitative relationship model is based on the above ternary mapping model, incorporating the collected ambient temperature T and battery voltage V as correction factors to form a quantitative relationship model, the expression of which is as follows: ; in, This is the predicted remaining battery capacity after temperature and voltage correction. An exponential nonlinear function is used as the ambient temperature correction factor. The battery voltage correction factor is a linear function. , , For the fitting parameters, The rated operating temperature of the battery. , For the fitting parameters, This is the battery's rated voltage.

[0007] Furthermore, the multi-objective fitness function constructed in S2 is based on the accuracy of low-pressure early warning. Maximize battery life and temperature adaptability The multi-objective fitness function for the objective is expressed as follows: ; in, ,and , To improve the accuracy of low-pressure early warning after normalization, To maximize the normalized battery life, For normalized temperature adaptability, , , These are the weighting coefficients for the corresponding objectives; ; in, This represents the actual number of low-pressure warnings reported. To avoid missing the number of low-pressure warnings, The number of false low-pressure warnings This represents the total number of low-pressure warnings. ; ; in, Let i be the battery's usable time for the i-th sample. This is the battery's rated voltage. For forklift energy conversion efficiency, Let the load power of the i-th sample be given by... And forklift operating speed calculation, This represents the maximum available time in a clean dataset R; ; in, For the battery's rated capacity, Let be the minimum actual capacity in a clean dataset R. ,and , , Both are minimizing objective functions.

[0008] Furthermore, the optimization calculation process in S3 is carried out in a phased and strongly coupled collaborative optimization manner in conjunction with the guidance mechanism; Automatic optimization involves uniformly employing two-dimensional real-number encoding for both WOA individuals and PSO particles, with the optimization variable being a two-dimensional vector. ; in, This is the low-pressure warning threshold. , is the voltage compensation coefficient; The global large-scale search based on the WOA algorithm includes three behaviors for position updates: prey encirclement, bubble net attack, and random search, and is based on the probability of behavior selection. The trigger expression is as follows: ; in, , For the coefficient vector, For random individuals, To find the global optimal solution for variable X, This is a constant specific to the WOA algorithm, and its value is 1. , For the j-th WOA individual in the t-th iteration, ; Local optimization is performed based on the PSO algorithm, and its velocity and position update formulas are adapted to the optimization variable X. The expressions are as follows: ; Where ω is the inertia weight, , As a learning factor, , It is a random vector. This represents the position of the m-th PSO particle. Consistent with the optimization variable X; After parallel iteration of the WOA and PSO algorithms, the global optimal solution is selected based on non-dominated sorting and crowding calculation, and the multi-objective fitness function F(X) is used as the criterion for judging the quality of the solution.

[0009] Furthermore, the step of selecting the globally optimal solution based on non-dominated ranking and crowding calculation, and using the multi-objective fitness function F(X) as the criterion for judging the quality of the solution, includes the following steps: S3.1. The first The solutions from the WOA algorithm and the PSO algorithm in the next iteration are merged into the total solution set. ; S3.2. For any two solutions Based on the accuracy of low-pressure early warning Maximize battery life and temperature adaptability The three optimization objectives determine the dominance relationship, if All sub-objectives are non-inferior to ,Right now And at least one sub-objective is strictly superior to ,but Dominate Solutions that are not dominated by any solution constitute the non-dominated solution set. ; S3.3. Calculate the non-dominated solution set The crowding degree of each solution is then calculated from the non-dominated solution set. Select the solution with the highest congestion. =[ , ], and as the first The global optimal solution in the next iteration; in, To optimize the low-pressure warning threshold, This is the optimized voltage compensation coefficient.

[0010] Furthermore, the optimization result of S3 determined by S4 is based on the standard low-pressure threshold. Battery capacity degradation rate Determine the optimization result of S3 ,Right now like ,and The optimal result is... It is valid and outputs the optimization results. ; like ,or The optimal result is... If invalid, dynamically adjust the weights and then return to S3; in, Allowable deviation for low-pressure early warning threshold For reference capacity decay rate, This is the allowable deviation for capacity decay rate.

[0011] Furthermore, the dynamic adjustment of weights is based on a ternary mapping model or a quantitative relationship model to dynamically adjust the weights of the multi-objective fitness function; like ,and Then, the weights of the multi-objective fitness function are dynamically adjusted based on the quantitative relationship model; like ,and Then, the weights of the multi-objective fitness function are dynamically adjusted based on the ternary mapping model; like ,and First, the weights of the multi-objective fitness function are dynamically adjusted based on the ternary mapping model, and then the weights of the multi-objective fitness function are dynamically adjusted based on the quantitative relationship model.

[0012] Furthermore, the dynamic adjustment of the multi-objective fitness function weights based on the quantitative relationship model includes calculating the weighting factor, i.e. ; The dynamic adjustment of multi-objective fitness function weights based on the ternary mapping model includes calculating the weighting factor, i.e. ; in, This represents the average remaining battery capacity relative to the rated capacity after temperature and voltage correction for the entire sample. This represents the mean deviation of the environmental temperature correction factor for the entire sample. This represents the mean deviation of the battery voltage correction factor across the entire sample. This represents the average deviation of the battery capacity decay rate from the reference value across the entire sample. The mean relative deviation of the fitting function for the full sample load and job duration; Calculation based on the adjustment factor After adjusting the weights and normalizing them, we obtain... , , ; in, , , To adjust the weights of the previous multi-objective fitness function, , , To adjust the weights of the multi-objective fitness function, ,and .

[0013] Furthermore, the optimization calculation process in S3 also includes synchronously correcting the temperature compensation coefficient, which is defined as the ambient temperature correction factor in the quantitative relationship model. Temperature sensitivity coefficient Temperature sensitivity coefficients were obtained by fitting R to a clean dataset. The initial value, and satisfying ; The WOA algorithm employs a global optimal solution-based approach during the global large-scale search phase. The incremental coarse correction is calculated using the following formula: ; in, For global correction step size, This represents the mean deviation of the environmental temperature correction factor for the entire sample. This represents the temperature-adaptive objective function value corresponding to the global optimal solution. The mean of the temperature fitness objective function for all solutions; The PSO algorithm employs a small-step fine correction based on congestion during the local optimization phase. The correction formula is as follows: ; in, For local correction step size, The crowding degree of the global optimal solution. The average crowding degree of the non-dominated solution set. For random disturbance factors; Revised Real-time temperature correction factor updates And in turn, it guides the optimization direction of WOA-PSO.

[0014] Compared with existing technologies, this invention provides a PSO-based method for optimizing low-pressure management of forklifts, which has the following advantages: 1. This PSO-based forklift low-pressure management optimization method accurately quantifies the coupling loss relationship under various working conditions through a constructed ternary mapping model. It also incorporates a quantitative relationship model with an exponential temperature correction factor to accurately characterize the nonlinear temperature characteristics. After fitting with the dual models, the accuracy of battery remaining capacity prediction is significantly improved, providing a high-quality, high-precision quantitative data foundation for algorithm optimization. 2. This PSO-based forklift low-pressure management optimization method achieves a quantitative balance of the three core requirements by constructing a multi-objective fitness function with the objectives of low-pressure early warning accuracy, maximizing battery availability time, and temperature adaptability, thus avoiding the limitations of single-objective optimization. All sub-objective functions are normalized and set as minimization objectives, which is compatible with the optimization logic of the WOA-PSO hybrid algorithm. This provides a clear and quantifiable standard for judging the quality of solutions for algorithm iteration and optimization, transforming multi-objective optimization from qualitative requirements to quantitative calculations. 3. This PSO-based forklift low-pressure management optimization method builds a WOA-PSO parallel hybrid fusion algorithm, combining the advantages of WOA global large-scale search and PSO local fine-grained optimization. It avoids local optima and improves optimization accuracy, achieving efficient and accurate collaborative optimization of low-pressure warning threshold and voltage compensation coefficient. The algorithm is engineered and adapted, and a unified two-dimensional real number encoding is used to focus on core parameters, reducing the computational complexity of the algorithm and adapting to the embedded computing requirements of the forklift BMS system. 4. This PSO-based forklift low-pressure management optimization method first selects the Pareto optimal non-dominated solution set through solution selection rules based on non-dominated sorting and congestion calculation, and then selects the solution with the highest congestion as the global optimal solution. This ensures both the optimality of the optimization results and the diversity and representativeness of the working conditions of the solutions. A synchronous correction mechanism for the temperature compensation coefficient is designed to iteratively correct the temperature compensation coefficient and the core parameters in stages at the same frequency, achieving strong coupling and collaborative optimization between the two. This significantly improves the adaptability of forklift low-pressure management to different ambient temperatures and accurately matches the battery characteristics under high and low temperatures. 5. This PSO-based forklift low-pressure management optimization method sets dual engineering judgment conditions based on standard low-pressure threshold and reference capacity decay rate, defining reasonable boundaries for optimization results. This achieves a balance between the mathematical optimality of algorithm optimization and the engineering practicality of low-pressure management. A scenario-based dynamic weighting strategy is designed, accurately matching a ternary mapping model or quantitative relationship model based on the cause of optimization failure. In dual-failure scenarios, the principle of first correcting the underlying layer and then fine-tuning is followed to achieve targeted weighting. Invalid results trigger re-optimization, forming a closed-loop optimization process of optimization, judgment, weighting, and further optimization. This significantly improves the success rate of optimization results, ultimately achieving full-condition adaptation of forklift low-pressure management to different loads, operating durations, and ambient temperatures, while considering battery safety, operational efficiency, and environmental adaptability. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of a PSO-based low-pressure management optimization method for forklifts proposed in this invention. Detailed Implementation

[0016] 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.

[0017] Example 1: Please refer to Figure 1 The forklift low-pressure management optimization method based on PSO in this embodiment includes the following steps; S1. Collect raw data of forklift battery voltage V, ambient temperature T, battery capacity C, load L and working time t, then preprocess the data, establish a mapping model and fit a quantitative relationship model. S2. Build the WOA-PSO parallel hybrid fusion algorithm model, construct the multi-objective fitness function, and then initialize it; S3. Execute the optimized calculation process and automatically find the optimal solution; S4. Determine the optimization result of S3. If yes, output the optimization result; otherwise, dynamically adjust the weights and return to S3. S5. Based on the optimization results, the optimal threshold and compensation coefficient are obtained, and then the compensation voltage is calculated. The calculation logic of the optimal threshold, compensation coefficient and compensation voltage is embedded into the forklift BMS system to complete the deployment of optimization parameters.

[0018] Preprocessing includes cleaning, denoising, and normalizing the collected raw data. To obtain a clean dataset ; Each sample , ; Calculate the battery capacity decay rate for the i-th sample based on a clean dataset R. The calculation formula is as follows: ,in, This refers to the battery's rated capacity. The mapping model is established based on a clean dataset R and the battery capacity decay rate of the i-th sample. And a multivariate nonlinear regression was used to establish the load. Homework duration Battery capacity The ternary mapping model is expressed as follows: ; ; in, , , , , , The fitting parameters were obtained based on a clean dataset R. The fitted quantitative relationship model is based on the above ternary mapping model, incorporating the collected ambient temperature T and battery voltage V as correction factors to form a quantitative relationship model, the expression of which is as follows: ; in, This is the predicted remaining battery capacity after temperature and voltage correction. An exponential nonlinear function is used as the ambient temperature correction factor. The battery voltage correction factor is a linear function. , , For the fitting parameters, The rated operating temperature of the battery. , For the fitting parameters, This is the battery's rated voltage.

[0019] It should be noted that the normalization of the original data uses the min-max normalization method, and the specific formula is as follows: ; in, This is the original data. , These are the minimum and maximum values ​​for the corresponding data dimensions, normalized to... Eliminate the interference of dimensional differences on modeling; Fitting parameters of the ternary mapping model and the quantitative relationship model - , - , - All values ​​were obtained by least squares fitting using multivariate nonlinear regression. The fitting process was based on a clean dataset R, with the goal of minimizing the sum of squared residuals between the fitted values ​​and the actual values. The rated operating temperature T0 of the battery is taken as the general operating condition reference value of 25 degrees Celsius for forklift batteries, with a fluctuation of ±5 degrees Celsius. The rated voltage V0 of the battery is selected according to the actual configuration of the forklift. By constructing a two-layer model with coupled operating conditions and nonlinear temperature correction, the core solution addresses the problem that existing low-voltage management methods neglect the coupled loss relationship between load, operating time, and battery capacity, and cannot accurately characterize the nonlinear impact of temperature on battery capacity. This enables accurate prediction of remaining battery capacity and provides a high-quality quantitative data foundation for subsequent algorithm optimization.

[0020] In addition, the multi-objective fitness function constructed in S2 is based on the accuracy of low-pressure early warning. Maximize battery life and temperature adaptability The multi-objective fitness function for the objective is expressed as follows: ; in, ,and , To improve the accuracy of low-pressure early warning after normalization, To maximize the normalized battery life, For normalized temperature adaptability, , , These are the weighting coefficients for the corresponding objectives; ; in, This represents the actual number of low-pressure warnings reported. To avoid missing the number of low-pressure warnings, The number of false low-pressure warnings This represents the total number of low-pressure warnings. ; ; in, Let i be the battery's usable time for the i-th sample. This is the battery's rated voltage. For forklift energy conversion efficiency, Let the load power of the i-th sample be given by... And forklift operating speed calculation, This represents the maximum available time in a clean dataset R; ; in, For the battery's rated capacity, Let be the minimum actual capacity in a clean dataset R. ,and , , Both are minimizing objective functions.

[0021] It should be noted that the weight coefficients in the multi-objective fitness function , , The initial value can be set according to the principle of equal weight. It can also be calibrated according to the actual working conditions of the forklift. The standard value range is: , , When prioritizing battery safety protection, the level can be appropriately increased. When focusing on work efficiency, it is appropriate to increase ; Load power of the i-th sample The specific calculation formula is as follows: ; in, The forklift load power conversion factor is calibrated by the forklift model, v is the actual operating speed of the forklift, and the minimum actual capacity in the clean dataset R is the value of the forklift load power conversion factor. To minimize capacity after removing outliers, and to avoid the impact of extreme values ​​on temperature adaptability. Interference in computation; By constructing a multi-objective fitness function, the three core requirements of low-pressure early warning accuracy, maximizing battery availability time, and temperature adaptability are quantitatively integrated, solving the problem of neglecting one aspect in multi-objective optimization and single-objective optimization. At the same time, all sub-objective functions are normalized and set as minimization objectives, which is compatible with the optimization logic of the WOA-PSO hybrid algorithm and provides a clear and quantifiable standard for judging the quality of solutions for algorithm optimization.

[0022] Furthermore, the optimization calculation process in S3 is carried out in a phased and strongly coupled collaborative optimization manner in conjunction with the guidance mechanism; Automatic optimization involves uniformly employing two-dimensional real-number encoding for both WOA individuals and PSO particles, with the optimization variable being a two-dimensional vector. ; in, This is the low-pressure warning threshold. , is the voltage compensation coefficient; The global large-scale search based on the WOA algorithm includes three behaviors for position updates: prey encirclement, bubble net attack, and random search, and is based on the probability of behavior selection. The trigger expression is as follows: ; in, , For the coefficient vector, For random individuals, To find the global optimal solution for variable X, This is a constant specific to the WOA algorithm, and its value is 1. , For the j-th WOA individual in the t-th iteration, ; Local optimization is performed based on the PSO algorithm, and its velocity and position update formulas are adapted to the optimization variable X. The expressions are as follows: ; Where ω is the inertia weight, , As a learning factor, , It is a random vector. This represents the position of the m-th PSO particle. Consistent with the optimization variable X; After parallel iteration of the WOA and PSO algorithms, the global optimal solution is selected based on non-dominated sorting and crowding calculation, and the multi-objective fitness function F(X) is used as the criterion for judging the quality of the solution.

[0023] It should be noted that the coefficient vector in the WOA algorithm It is a linearly decreasing vector with a range of values ​​of 1. The coefficient vector decreases linearly from 2 to 0 during the iteration process. The vector is generated randomly, and its value range is... The learning factor decreases linearly with the iteration process. , Take the classic optimal value Random vectors , It is a uniform random vector, and its value range is 1. ; The number of individuals in the WOA algorithm is consistent with the number of particles in the PSO algorithm, with a typical value of 20-50, which can be adjusted according to the computing power of the forklift BMS system. By performing engineering adaptation design on the WOA-PSO parallel hybrid fusion algorithm, a unified two-dimensional real number encoding is adopted to focus on core optimization parameters. To address the issues of low local convergence accuracy of the single WOA algorithm and the tendency of the single PSO algorithm to get trapped in local optima, a phased collaborative optimization mode combining global large-scale WOA search and local fine-grained PSO search is adopted. This reduces the computational complexity of the algorithms while improving the efficiency and accuracy of optimizing the low-voltage warning threshold and voltage compensation coefficient.

[0024] The process of selecting the global optimal solution based on non-dominated ranking and crowding calculation, and using the multi-objective fitness function F(X) as the criterion for judging the quality of the solution, includes the following steps: S3.1. The first The solutions from the WOA algorithm and the PSO algorithm in the next iteration are merged into the total solution set. ; S3.2. For any two solutions Based on the accuracy of low-pressure early warning Maximize battery life and temperature adaptability The three optimization objectives determine the dominance relationship, if All sub-objectives are non-inferior to ,Right now And at least one sub-objective is strictly superior to ,but Dominate Solutions that are not dominated by any solution constitute the non-dominated solution set. ; S3.3. Calculate the non-dominated solution set The crowding degree of each solution is then calculated from the non-dominated solution set. Select the solution with the highest congestion. =[ , ], and as the first The global optimal solution in the next iteration; in, To optimize the low-pressure warning threshold, This is the optimized voltage compensation coefficient.

[0025] It should be noted that the specific formula for calculating congestion is as follows: ; in, For the first The crowding of each solution , Let m be the function value of the m-th objective function at the (j-1)-th solution. , For the m-th objective function in the non-dominated solution set The maximum and minimum values ​​in Corresponding to , , ; The solutions from the WOA algorithm and the PSO algorithm are merged into a total solution set. Afterwards, the solution deduplication process needs to be performed to remove duplicate solutions with completely identical objective function values, so as to avoid the interference of duplicate solutions on non-dominated sorting and crowding calculation. By using standardized non-dominated sorting and crowding degree calculation for solution selection rules, the problem of multiple non-dominated solutions in multi-objective optimization that cannot be directly determined as optimal is solved. Non-dominated sorting ensures the Pareto optimality of the optimization results, while crowding degree calculation and selection of the solution with the highest crowding degree ensures the diversity of optimal solutions and the representativeness of working conditions, allowing the algorithm's optimization results to be adapted to different forklift operating conditions.

[0026] In addition, the optimization calculation process in S3 also includes synchronous correction of the temperature compensation coefficient, which is defined as the ambient temperature correction factor in the quantitative relationship model. Temperature sensitivity coefficient Temperature sensitivity coefficients were obtained by fitting R to a clean dataset. The initial value, and satisfying ; The WOA algorithm employs a global optimal solution-based approach during the global large-scale search phase. The incremental coarse correction is calculated using the following formula: ; in, For global correction step size, This represents the mean deviation of the environmental temperature correction factor for the entire sample. This represents the temperature-adaptive objective function value corresponding to the global optimal solution. The mean of the temperature fitness objective function for all solutions; The PSO algorithm employs a small-step fine correction based on congestion during the local optimization phase. The correction formula is as follows: ; in, For local correction step size, The crowding degree of the global optimal solution. The average crowding degree of the non-dominated solution set. For random disturbance factors; Revised Real-time temperature correction factor updates And in turn, it guides the optimization direction of WOA-PSO.

[0027] It should be noted that the temperature sensitivity coefficient The reasonable constraint range of the project is , This range is suitable for the temperature characteristics of lead-acid and lithium-ion batteries used in forklifts; Global correction step size Pick Local correction step size Pick Small step size correction can avoid The process of iteration is subject to significant fluctuations, ensuring the stability of the correction. random disturbance factor The values ​​are uniformly randomized and range from 1 to 10. Introducing a random perturbation factor can prevent the correction process from getting trapped in local optima and improve the robustness of the temperature compensation coefficient. The temperature compensation coefficient is defined as the temperature correction factor through a synchronous correction mechanism. Temperature sensitivity coefficient and achieve with , This method employs iterative correction at the same frequency to address the issue in existing low-pressure management methods where temperature compensation uses fixed coefficients or linear compensation and cannot be dynamically corrected according to actual operating conditions. It utilizes a phased correction strategy of global coarse correction (WOA) and local fine correction (PSO) to achieve... The strong coupling and synergistic optimization with low-voltage warning threshold and voltage compensation coefficient significantly improves the adaptability of forklift low-voltage management to different ambient temperatures.

[0028] Furthermore, S4 determines the optimization result of S3 based on the standard low-pressure threshold. Battery capacity degradation rate Determine the optimization result of S3 ,Right now like ,and The optimal result is... It is valid and outputs the optimization results. ; like ,or The optimal result is... If invalid, dynamically adjust the weights and then return to S3; in, Allowable deviation for low-pressure early warning threshold For reference capacity decay rate, This is the allowable deviation for capacity decay rate.

[0029] It should be noted that the standard low-pressure threshold The value is based on the low-voltage warning baseline value specified in the forklift battery cell specification sheet, and the allowable deviation of the low-voltage warning threshold is... The value range is 0.5V-2.0V, and the allowable deviation of the capacity decay rate is... The value range is [0.05, 0.15], and both are calibrated based on the actual operating conditions of the forklift. Reference capacity decay rate The mean of the battery capacity decay rate of all samples in the clean dataset R is calculated using the following formula: Using the full sample mean as a reference can fit the overall battery degradation conditions. By setting dual judgment conditions based on standard low-pressure threshold and battery capacity decay rate, a reasonable engineering boundary is defined for the algorithm optimization results. This solves the problem that the algorithm optimization results may deviate from industrial reality and have extreme thresholds or compensation coefficients. It achieves the unity of mathematical optimality of algorithm optimization and engineering practicality of forklift low-pressure management. At the same time, invalid results trigger the logic of re-optimization to form a closed-loop optimization, thereby improving the effectiveness of the optimization results.

[0030] Among them, the dynamic adjustment weight is based on the ternary mapping model or the quantitative relationship model to dynamically adjust the weight of the multi-objective fitness function; like ,and Then, the weights of the multi-objective fitness function are dynamically adjusted based on the quantitative relationship model; like ,and Then, the weights of the multi-objective fitness function are dynamically adjusted based on the ternary mapping model; like ,and First, the weights of the multi-objective fitness function are dynamically adjusted based on the ternary mapping model, and then the weights of the multi-objective fitness function are dynamically adjusted based on the quantitative relationship model.

[0031] It should be noted that the dynamic weight adjustment operation is executed immediately after the optimization result is determined to be invalid, and the weight adjustment is only performed once each time the optimization is re-optimized, until the optimization result is valid or the maximum number of iterations of the algorithm is reached. The maximum number of iterations of the WOA-PSO hybrid algorithm is usually 100-200 times, which can be adjusted according to the real-time requirements of the forklift BMS system. The matching of weighting models for different failure scenarios follows the principle of using ternary mapping models for low-level problems and quantitative relationship models for refined problems; in, The deviation from the baseline is due to the underlying operating condition-attenuation coupling problem. The deviation from the reference is due to a problem with the fine-tuning of the temperature-voltage correction. By employing a scenario-based dynamic weight adjustment strategy, this approach addresses the issues of existing dynamic weight adjustment methods lacking targeting and being ineffective due to the use of a single model for general adjustment. By accurately tracing the causes of optimization failures and matching corresponding weight adjustment models, it achieves the optimization logic of problem tracing, model matching, and targeted weight adjustment. In dual-failure scenarios, the approach first uses a ternary mapping model to adjust weights to resolve underlying biases, and then uses a quantitative relationship model for refined correction. This follows the basic principles of engineering optimization, reducing the number of algorithm iterations and improving the success rate of optimization.

[0032] In addition, dynamically adjusting the weights of the multi-objective fitness function based on a quantitative relationship model includes calculating the adjustment factor. , , ,Right now ; Dynamically adjusting the weights of the multi-objective fitness function based on a ternary mapping model includes calculating the weighting factor. , , ,Right now ; in, This represents the average remaining battery capacity relative to the rated capacity after temperature and voltage correction for the entire sample. This represents the mean deviation of the environmental temperature correction factor for the entire sample. This represents the mean deviation of the battery voltage correction factor across the entire sample. This represents the average deviation of the battery capacity decay rate from the reference value across the entire sample. The mean relative deviation of the fitting function for the full sample load and job duration; Calculation based on the adjustment factor After adjusting the weights and normalizing them, we obtain... , , ; in, , , To adjust the weights of the previous multi-objective fitness function, , , To adjust the weights of the multi-objective fitness function, ,and .

[0033] It should be noted that the specific calculation formula for weight normalization based on the weighting factor is as follows: ; Among them, when adjusting the weights based on the ternary mapping model, the above formula is... , , Replace with , , The weight calculation can then be completed; It should be noted that, , ; , ; , ; By using specific quantitative calculation formulas for the adjustment factors and adjusted weights, this paper addresses the issues of dynamic weight adjustment having only strategies but no specific calculation methods, and the adjusted weights not meeting constraints. All adjustment factors are derived based on the core parameters of the corresponding models, realizing a direct link between the quantitative representation of model parameters and weight reconstruction. Furthermore, the normalization formula strictly guarantees that the adjusted weights meet the constraints. ,and The constraints are adjusted so that the weights can be directly substituted into the original multi-objective fitness function, seamlessly connecting with the WOA-PSO hybrid algorithm without modifying the algorithm framework.

[0034] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0035] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for optimizing low-pressure management of forklifts based on PSO, characterized in that, Includes the following steps; S1. Collect raw data of forklift battery voltage V, ambient temperature T, battery capacity C, load L and working time t, then preprocess the data, establish a mapping model and fit a quantitative relationship model. S2. Build the WOA-PSO parallel hybrid fusion algorithm model, construct the multi-objective fitness function, and then initialize it; S3. Execute the optimized calculation process and automatically find the optimal solution; S4. Determine the optimization result of S3. If yes, output the optimization result; otherwise, dynamically adjust the weights and return to S3. S5. Based on the optimization results, the optimal threshold and compensation coefficient are obtained, and then the compensation voltage is calculated. The calculation logic of the optimal threshold, compensation coefficient and compensation voltage is embedded into the forklift BMS system to complete the deployment of optimization parameters.

2. The forklift low-pressure management optimization method based on PSO according to claim 1, characterized in that: The preprocessing includes cleaning, denoising, and normalizing the collected raw data. To obtain a clean dataset ; Each sample , ; Calculate the battery capacity decay rate for the i-th sample based on a clean dataset R. The calculation formula is as follows: ,in, This refers to the battery's rated capacity. The mapping model is established based on a clean dataset R and the battery capacity decay rate of the i-th sample. And a multivariate nonlinear regression was used to establish the load. Homework duration Battery capacity The ternary mapping model is expressed as follows: ； ; in, , , , , , The fitting parameters were obtained based on a clean dataset R. The fitted quantitative relationship model is based on the above ternary mapping model, incorporating the collected ambient temperature T and battery voltage V as correction factors to form a quantitative relationship model, the expression of which is as follows: ; in, This is the predicted remaining battery capacity after temperature and voltage correction. An exponential nonlinear function is used as the ambient temperature correction factor. The battery voltage correction factor is a linear function. , , For the fitting parameters, The rated operating temperature of the battery. , For the fitting parameters, This is the battery's rated voltage.

3. The forklift low-pressure management optimization method based on PSO according to claim 2, characterized in that: The multi-objective fitness function constructed in S2 is based on the accuracy of low-pressure early warning. Maximize battery life and temperature adaptability The multi-objective fitness function for the objective is expressed as follows: ； in, ,and , To improve the accuracy of low-pressure early warning after normalization, To maximize the normalized battery life, For normalized temperature adaptability, , , These are the weighting coefficients for the corresponding objectives; ； in, This represents the actual number of low-pressure warnings reported. To avoid missing the number of low-pressure warnings, The number of false low-pressure warnings This represents the total number of low-pressure warnings. ; ； in, Let i be the battery's usable time for the i-th sample. This is the battery's rated voltage. For forklift energy conversion efficiency, Let the load power of the i-th sample be given by... And forklift operating speed calculation, This represents the maximum available time in a clean dataset R; ； in, For the battery's rated capacity, Let be the minimum actual capacity in a clean dataset R. ,and , , Both are minimizing objective functions.

4. The forklift low-pressure management optimization method based on PSO according to claim 3, characterized in that: The optimization calculation process in S3 is carried out in a phased and strongly coupled collaborative optimization manner in conjunction with the guidance mechanism. Automatic optimization involves uniformly employing two-dimensional real-number encoding for both WOA individuals and PSO particles, with the optimization variable being a two-dimensional vector. ; in, This is the low-pressure warning threshold. , is the voltage compensation coefficient; The global large-scale search based on the WOA algorithm includes three behaviors for position updates: prey encirclement, bubble net attack, and random search, and is based on the probability of behavior selection. The trigger expression is as follows: ； in, , For the coefficient vector, For random individuals, To find the global optimal solution for variable X, This is a constant specific to the WOA algorithm, and its value is 1. , For the j-th WOA individual in the t-th iteration, ; Local optimization is performed based on the PSO algorithm, and its velocity and position update formulas are adapted to the optimization variable X. The expressions are as follows: ； Where ω is the inertia weight, , As a learning factor, , It is a random vector. This represents the position of the m-th PSO particle. Consistent with the optimization variable X; After parallel iteration of the WOA and PSO algorithms, the global optimal solution is selected based on non-dominated sorting and crowding calculation, and the multi-objective fitness function F(X) is used as the criterion for judging the quality of the solution.

5. The forklift low-pressure management optimization method based on PSO according to claim 4, characterized in that: The method of selecting the global optimal solution based on non-dominated ranking and crowding calculation, and using the multi-objective fitness function F(X) as the criterion for judging the quality of the solution, includes the following steps: S3.

1. The first The solutions from the WOA algorithm and the PSO algorithm in the next iteration are merged into the total solution set. ; S3.

2. For any two solutions Based on the accuracy of low-pressure early warning Maximize battery life and temperature adaptability The three optimization objectives determine the dominance relationship, if All sub-objectives are non-inferior to ,Right now And at least one sub-objective is strictly superior to ,but Dominate Solutions that are not dominated by any solution constitute the non-dominated solution set. ; S3.

3. Calculate the non-dominated solution set The crowding degree of each solution is then calculated from the non-dominated solution set. Select the solution with the highest congestion. =[ , ], and as the first The global optimal solution in the next iteration; in, To optimize the low-pressure warning threshold, This is the optimized voltage compensation coefficient.

6. The forklift low-pressure management optimization method based on PSO according to claim 5, characterized in that: The S4 judgment of the optimization result of S3 is based on the standard low-pressure threshold. Battery capacity degradation rate Determine the optimization result of S3 ,Right now like ,and The optimal result is... It is valid and outputs the optimization results. ; like ,or The optimal result is... If invalid, dynamically adjust the weights and then return to S3; in, Allowable deviation for low-pressure early warning threshold For reference capacity decay rate, This is the allowable deviation for capacity decay rate.

7. The forklift low-pressure management optimization method based on PSO according to claim 6, characterized in that: The dynamically adjusted weights are based on a ternary mapping model or a quantitative relationship model to dynamically adjust the weights of the multi-objective fitness function. like ,and Then, the weights of the multi-objective fitness function are dynamically adjusted based on the quantitative relationship model; like ,and Then, the weights of the multi-objective fitness function are dynamically adjusted based on the ternary mapping model; like ,and First, the weights of the multi-objective fitness function are dynamically adjusted based on the ternary mapping model, and then the weights of the multi-objective fitness function are dynamically adjusted based on the quantitative relationship model.

8. The forklift low-pressure management optimization method based on PSO according to claim 7, characterized in that: The dynamic adjustment of multi-objective fitness function weights based on a quantitative relationship model includes calculating the weighting factor, i.e. ； The dynamic adjustment of multi-objective fitness function weights based on the ternary mapping model includes calculating the weighting factor, i.e. ； in, This represents the average remaining battery capacity relative to the rated capacity after temperature and voltage correction for the entire sample. This represents the mean deviation of the environmental temperature correction factor for the entire sample. This represents the mean deviation of the battery voltage correction factor across the entire sample. This represents the average deviation of the battery capacity decay rate from the reference value across the entire sample. The mean relative deviation of the fitting function for the full sample load and job duration; Calculation based on the adjustment factor After adjusting the weights and normalizing them, we obtain... , , ; in, , , To adjust the weights of the previous multi-objective fitness function, , , To adjust the weights of the multi-objective fitness function, ,and .

9. The forklift low-pressure management optimization method based on PSO according to claim 5, characterized in that: The optimization calculation process in S3 also includes synchronously correcting the temperature compensation coefficient, which is defined as the ambient temperature correction factor in the quantitative relationship model. Temperature sensitivity coefficient Temperature sensitivity coefficients were obtained by fitting R to a clean dataset. The initial value, and satisfying ; The WOA algorithm employs a global optimal solution-based approach during the global large-scale search phase. The incremental coarse correction is calculated using the following formula: ； in, For global correction step size, This represents the mean deviation of the environmental temperature correction factor for the entire sample. This represents the temperature-adaptive objective function value corresponding to the global optimal solution. The mean of the temperature fitness objective function for all solutions; The PSO algorithm employs a small-step fine correction based on congestion during the local optimization phase. The correction formula is as follows: ； in, For local correction step size, The crowding degree of the global optimal solution. The average crowding degree of the non-dominated solution set. For random disturbance factors; Revised Real-time temperature correction factor updates And in turn, it guides the optimization direction of WOA-PSO.

Images