Hydraulic control strategy online optimization method and system based on ai model predictive control for lifting machine

Abstract

This invention provides an online optimization method and system for the hydraulic control strategy of a lift based on AI model predictive control, relating to the field of artificial intelligence technology. Starting with the current control strategy of the lift's hydraulic system, the method generates a multi-round pre-simulation optimization strategy through multiple rounds of iterative simulation using an AI predictive model. Real-time operational feedback features are extracted and evaluated by a dynamic adaptation coupling unit to generate a strategy adaptation priority sequence. Based on this priority sequence, a strategy refinement process is triggered to generate a target optimized control strategy. This strategy is then converted into control commands and applied to the system, extracting new feedback features. The above steps are repeated using the new strategy and the new feedback features to achieve online optimization of the lift's hydraulic control strategy, improving operational performance and safety.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and more specifically, to an online optimization method and system for hydraulic control strategies of a lifting machine based on AI model predictive control. Background Technology

[0002] In industrial production, lifting platforms are crucial material handling equipment, and the control strategy of their hydraulic systems directly affects their operational performance and safety. Traditional lifting platform hydraulic control strategies typically employ fixed-parameter control, meaning the hydraulic system's operation is controlled based on pre-set parameters. This control method struggles to adjust control parameters in real-time to adapt to varying operational needs under complex and changing working conditions. For example, when the load weight on the lifting platform changes, the ambient temperature fluctuates, or the hydraulic system experiences wear, the fixed-parameter control strategy cannot respond promptly, potentially leading to unstable operation, low efficiency, or even safety hazards.

[0003] While some existing optimization methods can adjust control strategies to a certain extent, they often require manual intervention, and the adjustment process is cumbersome, inefficient, and unable to achieve real-time online optimization. Furthermore, these methods lack accurate prediction of the system's future operating state, making it difficult to identify potential problems in advance and optimize accordingly. Consequently, they fail to meet the stringent requirements of modern industrial production for the efficient, stable, and safe operation of lifting platform hydraulic systems. Summary of the Invention

[0004] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide an online optimization method for hydraulic control strategy of a lifting machine based on AI model predictive control, the method comprising: Taking the current control strategy of the lift hydraulic system as the starting point of the iteration, the AI ​​prediction model is triggered to perform a multi-round strategy pre-exercise iteration process. The AI ​​prediction model is used to adjust the parameters of the current control strategy and perform operation state pre-exercise in multiple rounds to generate a multi-round pre-exercise optimization strategy. Each round of pre-exercise optimization strategy corresponds to a set of predicted operation states of the lift hydraulic system. The multi-round pre-exercise optimization strategy includes a set of control strategies with different parameter configurations. The real-time operation feedback features of the lifting machine hydraulic system are extracted. The multi-round pre-exercise optimization strategy and the real-time operation feedback features of the lifting machine hydraulic system are input into the dynamic adaptation coupling unit. The dynamic adaptation coupling unit comprehensively evaluates the adaptability of each round of pre-exercise optimization strategy and generates a strategy adaptation priority sequence. The strategy adaptation priority sequence contains the adaptation ranking information of all pre-exercise optimization strategies and the corresponding strategy features and predicted state feature information. Based on the policy adaptation priority sequence, the policy refinement process of the AI ​​prediction model is triggered. The top K pre-optimization policies in the policy adaptation priority sequence are subjected to feature fusion, core feature enhancement and parameter convergence processing to generate the target optimization control policy. The target optimization control policy includes all optimized control parameter configuration information and parameter association logic features. The target optimization control strategy is converted into control commands that can be recognized by each actuator of the lifting hydraulic system. The control commands are then transmitted to the execution control unit of the lifting hydraulic system to drive each actuator to operate according to the requirements of the target optimization control strategy. This realizes the application of the target optimization control strategy in the lifting hydraulic system and extracts the real-time operation feedback features after the application. Taking the target-optimized control strategy after application as the new iteration starting point, and combining the real-time operation feedback characteristics after application, all steps of the multi-round strategy pre-exercise iteration process, dynamic adaptation coupling evaluation, and strategy refinement process are repeatedly executed to continuously generate optimized target-optimized control strategies and apply them to the lift hydraulic system, thereby realizing online optimization of the lift hydraulic control strategy.

[0005] In another aspect, embodiments of the present invention also provide an online optimization system for hydraulic control strategy of a lift based on AI model predictive control, including a processor and a machine-readable storage medium connected to the processor. The machine-readable storage medium is used to store programs, instructions or code, and the processor is used to run the programs, instructions or code in the machine-readable storage medium to implement the above-described method.

[0006] Based on the above, this embodiment of the invention takes the current control strategy of the lift hydraulic system as the starting point for iteration. Through a multi-round strategy pre-exercise iteration process using an AI prediction model, it can comprehensively explore the potential optimization space of the control strategy under different parameter configurations, generate multi-round pre-exercise optimization strategies containing multiple possibilities, extract real-time operation feedback features, and combine them with dynamic adaptation coupling units to accurately evaluate the adaptability of each round of pre-exercise optimization strategy with actual working conditions, generating a strategy adaptation priority sequence. Based on the strategy refinement process triggered by this strategy adaptation priority sequence, it can deeply optimize highly adaptable strategies and generate a target optimization control strategy. This target optimization control strategy not only has reasonable parameter configuration but also has clear parameter association logic features. After applying the target optimization control strategy to the lift hydraulic system, iterative optimization is continuously performed from the application results as a new starting point, realizing real-time, automatic, and accurate online optimization of the lift hydraulic control strategy, effectively improving the operating performance and safety of the lift. Attached Figure Description

[0007] Figure 1 This is a schematic diagram of the execution flow of the online optimization method for hydraulic control strategy of a lift based on AI model predictive control provided in an embodiment of the present invention.

[0008] Figure 2 This is a schematic diagram of the hardware architecture of the online optimization system for hydraulic control strategy of a lift based on AI model predictive control provided in an embodiment of the present invention. Detailed Implementation

[0009] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a flowchart illustrating an online optimization method for hydraulic control strategy of a lift based on AI model predictive control, provided in one embodiment of the present invention. The following is a detailed description of this online optimization method for hydraulic control strategy of a lift based on AI model predictive control.

[0010] Step S110: Taking the current control strategy of the lift hydraulic system as the starting point of the iteration, trigger the multi-round strategy pre-exercise iteration process of the AI ​​prediction model. The AI ​​prediction model performs multiple rounds of adjustment and operation state pre-exercise on the parameters of the current control strategy to generate a multi-round pre-exercise optimization strategy. Each round of pre-exercise optimization strategy corresponds to a set of predicted operation states of the lift hydraulic system. The multi-round pre-exercise optimization strategy includes a set of control strategies with different parameter configurations.

[0011] In this embodiment, the current control strategy of the lift hydraulic system encompasses multiple control parameters, such as the hydraulic pump output flow parameter Q, the hydraulic valve opening parameter K, and the hydraulic cylinder stroke control parameter S. Using this current control strategy as the starting point for iteration means using the aforementioned existing parameter configuration as initial input to initiate a multi-round strategy pre-simulation iteration process for the AI ​​prediction model. The AI ​​prediction model will systematically adjust the aforementioned parameters based on preset algorithms and logic, and simulate the operating state of the lift hydraulic system after each adjustment. For example, assuming the current hydraulic pump output flow parameter Q is set to a specific value, the AI ​​prediction model might first attempt to adjust this flow parameter according to preset rules, and then simulate the lifting speed of the lift, the pressure distribution of the hydraulic system, and the stress on various components under this adjustment, thereby obtaining the first round of pre-simulation optimization strategy and its corresponding predicted operating state. Next, based on the prediction results of the first round, the parameters are adjusted again, such as adjusting the hydraulic valve opening parameter K, and the operating state is simulated again. This process continues, and after multiple rounds of parameter adjustments and state pre-simulation, a multi-round pre-simulation optimization strategy containing various different parameter configurations is finally generated.

[0012] Step S111: Extract all control parameters from the current control strategy of the lifting hydraulic system, record the complete identifier of each control parameter, which includes the parameter name, the component to which it belongs, and its position in the control strategy. Perform parameter sensitivity analysis and parameter correlation analysis on each control parameter, calculate the influence weight of each control parameter on the operating state of the lifting hydraulic system, sort all control parameters in descending order of influence weight, and generate a parameter influence weight sequence.

[0013] First, a comprehensive review of the current control strategy is necessary, extracting all control parameters. For example, for the lift's hydraulic system, this might include the "output flow rate Q" parameter of the hydraulic pump, belonging to the "hydraulic pump" component, and located in the control strategy as "main control module, segment A, line B"; and the "opening degree K" parameter of the hydraulic valve, belonging to the "hydraulic valve assembly," located in "branch control module, segment C, line D," etc. During the extraction process, it is crucial to ensure that the complete identifier of each parameter is accurately recorded to avoid omissions or incorrect identification.

[0014] Step S1111: Traverse all configuration contents of the current control strategy of the lifting hydraulic system, identify and extract each control parameter in the current control strategy, and record the complete identifier of each control parameter. The complete identifier of the control parameter includes the parameter name, the component to which it belongs, and the location information in the control strategy. Check all extracted control parameters one by one to confirm that the complete identifier information of each control parameter has been recorded and there are no missing or omitted parameter information.

[0015] In practice, a dedicated control strategy parsing tool is used to traverse the file or database storing the current control strategy. This tool identifies each control parameter defined in the file according to preset format rules. For example, if a parameter is defined in the control strategy configuration file using the format "Parameter Name: Output Flow Q; Belonging Component: Hydraulic Pump; Location: Main Control Module / Hydraulic Pump Control Section / Flow Setting Item," the parsing tool will extract "Output Flow Q" as the parameter name, "Hydraulic Pump" as the belonging component, and "Main Control Module / Hydraulic Pump Control Section / Flow Setting Item" as the location information, and record this information in a temporary data table. After the initial extraction, each parameter in the data table is checked against the original configuration of the control strategy to ensure that no parameter is omitted and that the identification information of each parameter is accurate.

[0016] Step S1112: Perform a separate parameter sensitivity analysis for each control parameter, simulating the impact of the value of the control parameter changing by a preset range on the operating state of the lifting hydraulic system while other parameters remain unchanged, and generate a sensitivity coefficient for each control parameter. The sensitivity coefficient is generated based on the range of change in the system operating state caused by the change in the parameter value.

[0017] Taking the "output flow rate Q" parameter as an example, during sensitivity analysis, while keeping other parameters such as hydraulic valve opening degree K and hydraulic cylinder stroke S constant, the output flow rate Q is adjusted according to preset rules based on its current set value. Then, an AI prediction model simulates the changes in the operating state of the lift's hydraulic system under these adjustments, such as changes in lifting speed, fluctuation range of system pressure, and increase or decrease in energy consumption. Based on a comprehensive evaluation of these changes in operating state, a sensitivity coefficient for the "output flow rate Q" parameter is generated.

[0018] Step S1113: Perform parameter correlation analysis on each control parameter, identify the correlation between the control parameter and other control parameters, count the number of correlations between the control parameter and other control parameters, analyze the correlation strength between the control parameter and other control parameters, and generate the correlation coefficient for each control parameter. The correlation coefficient is generated based on the number and strength of the correlations between parameters.

[0019] Taking the parameter "hydraulic valve opening degree K" as an example, by analyzing the logical relationship of the control strategy, it can be found that it is related to the parameter "output flow rate Q" because the opening degree of the hydraulic valve directly affects the flow distribution of hydraulic oil; it is also related to the parameter "hydraulic cylinder stroke S" because different opening degrees affect the extension and retraction speed of the hydraulic cylinder. The number of correlations between this parameter and other parameters is counted. Next, the correlation strength is analyzed; for example, the correlation strength between "hydraulic valve opening degree K" and "output flow rate Q" is relatively high, while the correlation strength with "hydraulic cylinder stroke S" is relatively low. The calculation of the correlation coefficient comprehensively considers both the number and strength of parameter correlations.

[0020] Step S1114: Normalize the sensitivity coefficient and correlation coefficient respectively to make them comparable within the same dimension range; after normalization, use a weighted summation method to generate the influence weight of each control parameter on the operating state of the lifting hydraulic system. The influence weight is a comprehensive quantitative result of the sensitivity coefficient and correlation coefficient, and the magnitude of the influence weight represents the degree of influence of the parameter on the operating state of the system.

[0021] During normalization, for sensitivity coefficients, the sensitivity coefficients of each parameter are processed according to a preset normalization rule to ensure they fall within a preset dimension range. Correlation coefficients are processed using a similar normalization rule. After normalization, the sensitivity coefficients and correlation coefficients are weighted and summed according to preset weight ratios to obtain the influence weight of each control parameter.

[0022] Step S1115: Sort all control parameters in descending order of influence weight values ​​to generate an initial parameter sorting sequence. The initial parameter sorting sequence contains the complete identifiers of all control parameters and their corresponding influence weight information.

[0023] After obtaining the influence weight of each control parameter, all control parameters are arranged in descending order of influence weight. For example, assuming the influence weight of "output flow rate Q" is W1, the influence weight of "hydraulic valve opening degree K" is W2, and the influence weight of "hydraulic cylinder stroke S" is W3, and W1>W2>W3, then the initial parameter sorting sequence is [output flow rate Q (W1), hydraulic valve opening degree K (W2), hydraulic cylinder stroke S (W3), ...], where each parameter is accompanied by its complete identifier and corresponding influence weight information.

[0024] Step S1116: Classify and organize the control parameters in the initial parameter sorting sequence. Divide the control parameters into different parameter categories according to the component type to which the parameters belong, and generate a classified parameter sorting sequence. The classified parameter sorting sequence contains the sorting information of control parameters under different categories.

[0025] The control parameters in the initial parameter sorting sequence are categorized according to the component type to which they belong. For example, parameters belonging to the "hydraulic pump" component are grouped into one category, parameters belonging to the "hydraulic valve assembly" component into another, and parameters belonging to the "hydraulic cylinder" component into a third, and so on. Within each category, the control parameters maintain the order in the initial parameter sorting sequence. This forms a categorized parameter sorting sequence, such as the hydraulic pump parameter sequence [output flow Q (W1), ...], the hydraulic valve assembly parameter sequence [hydraulic valve opening degree K (W2), ...], the hydraulic cylinder parameter sequence [hydraulic cylinder stroke S (W3), ...], etc.

[0026] Step S1117: Extract the overall influence weight of each parameter category on the operating status of the lifting hydraulic system, average the influence weight values ​​of all parameters under that parameter category, and generate the category influence weight. The category influence weight is used to evaluate the overall influence of parameters of different component types on the operating status of the system.

[0027] For each parameter category, the influence weights of all control parameters under that category are summed, and then divided by the number of parameters in that category to obtain the average influence weight of that category, i.e., the category influence weight. For example, if there are m parameters in the parameter sequence of the hydraulic pump category, and their influence weights are W1, W4, W5, etc., then the category influence weight of the hydraulic pump category is (W1+W4+W5+...) / m.

[0028] Step S1118: Sort the parameter categories in the classification parameter sorting sequence in descending order of category influence weight values ​​to generate a category sorting sequence. The category sorting sequence contains the identifiers of all parameter categories and the corresponding category influence weight information, which is used to determine the adjustment priority of component types.

[0029] The category influence weights of each parameter category are compared and sorted in descending order. For example, assuming the category influence weight of the hydraulic pump category is WC1, the category influence weight of the hydraulic valve group category is WC2, and the category influence weight of the hydraulic cylinder category is WC3, and WC1>WC2>WC3, then the category sorting sequence is [hydraulic pump (WC1), hydraulic valve group (WC2), hydraulic cylinder (WC3), ...], which includes the identifier of each parameter category and the corresponding category influence weight information.

[0030] Step S1119: Associate and integrate the category sorting sequence with the classification parameter sorting sequence to generate the final parameter influence weight sequence. The parameter influence weight sequence includes the complete identifier of each control parameter, the corresponding influence weight information, the parameter category to which it belongs, and the category influence weight information.

[0031] Using the parameter category order in the category sorting sequence as the primary order, control parameters from the corresponding classification parameter sorting sequence are placed under each parameter category. In this way, the parameter influence weight sequence contains both the parameter category sorting and the sorting of control parameters within each category. Furthermore, each control parameter includes its complete identifier, influence weight information, its parameter category, and category influence weight information. For example, the structure of the parameter influence weight sequence might be: Hydraulic pump (WC1) - Output flow rate Q (W1, Category: Hydraulic pump, Category influence weight: WC1), Hydraulic pump (WC1) - ..., Hydraulic valve group (WC2) - Hydraulic valve opening degree K (W2, Category: Hydraulic valve group, Category influence weight: WC2), Hydraulic valve group (WC2) - ... etc.

[0032] Step S112: Taking the current control strategy of the lifting hydraulic system as the starting point of the iteration, input the parameter influence weight sequence into the pre-simulation initialization unit of the AI ​​prediction model to generate the initial parameter adjustment direction and adjustment step size for the first round of pre-simulation. The initial parameter adjustment direction and adjustment step size are determined based on the weight information in the parameter influence weight sequence and are used to guide the parameter adjustment of the current control strategy.

[0033] The parameter influence weight sequence reflects the degree of influence of each control parameter on the operating state of the lifting machine's hydraulic system and the priority of the parameter categories. After inputting it into the pre-simulation initialization unit of the AI ​​prediction model, the pre-simulation initialization unit will determine which parameters need to be adjusted first in the first round of pre-simulation, as well as the direction and step size of the adjustment, based on the weight information. For example, the "output flow rate Q" parameter, which has a high influence weight in the parameter influence weight sequence, may be identified as a parameter to be adjusted first. The adjustment direction may be to increase or decrease it, and the size of the adjustment step size will also be set according to its influence weight. The higher the influence weight, the smaller the initial adjustment step size may be to avoid excessive fluctuations in the system's operating state.

[0034] Step S113: Adjust the core parameters in the current control strategy of the lifting hydraulic system according to the initial parameter adjustment direction and adjustment step size of the first round of pre-simulation, generate the adjusted parameter values, substitute the adjusted parameter values ​​into the current control strategy, replace the original parameter values, and generate the first round of pre-simulation optimization strategy. The first round of pre-simulation optimization strategy includes all the adjusted control parameters and parameter association logic.

[0035] Based on the adjustment direction and step size of the initial parameters determined in step S112 for the first round of pre-simulation, the "output flow rate Q," which is identified as a core parameter in the current control strategy, is adjusted. For example, if the adjustment direction is to increase and the adjustment step size is a certain set value, then the current value of "output flow rate Q" is added to this adjustment step size to obtain the adjusted parameter value Q1. Then, Q1 is substituted into the current control strategy, replacing the original "output flow rate Q" parameter value. At the same time, since there is correlation logic between the parameters in the control strategy, it is necessary to check and ensure that the adjusted parameter value does not destroy the original parameter correlation logic, or to adjust the correlation logic accordingly based on the new parameter value, ultimately generating the first round of pre-simulation optimization strategy containing all adjusted control parameters and parameter correlation logic.

[0036] Step S114: Input the first round of pre-simulation optimization strategy into the operation state pre-simulation unit of the AI ​​prediction model to simulate the complete operation process of the lift hydraulic system under the control of the first round of pre-simulation optimization strategy, and generate a set of predicted operation states corresponding to the first round of pre-simulation optimization strategy. The predicted operation states include the operation characteristics of each component of the lift hydraulic system and the overall operation trend.

[0037] The AI ​​prediction model's operational state pre-simulation unit contains a virtual simulation model of the lift's hydraulic system. This virtual simulation model can simulate the system's operation under different control strategies. After the first round of pre-simulation optimization strategy is input into this unit, the virtual simulation model will simulate the complete operation process of the lift, from start-up, lifting, pausing, to lowering, based on the control parameters and parameter association logic in the strategy. During the simulation, the operating characteristics of each component are recorded in real time, such as the speed change of the hydraulic pump, the action sequence of the hydraulic valves, and the displacement curve of the hydraulic cylinders, as well as the overall operating trend of the system, such as the lifting time, the stability of system pressure, and the distribution of energy consumption. This information together constitutes the predicted operating state corresponding to the first round of pre-simulation optimization strategy.

[0038] Step S115: Extract key feature information from the predicted running state corresponding to the first round of pre-exercise optimization strategy, feed the key feature information back to the pre-exercise iteration unit of the AI ​​prediction model, and generate the parameter adjustment direction and adjustment step size for the second round of pre-exercise. The parameter adjustment direction and adjustment step size for the second round of pre-exercise are generated based on the feature optimization of the first round of predicted running state and are used to guide the parameter adjustment of the first round of pre-exercise optimization strategy.

[0039] From the predicted operating states corresponding to the first round of pre-simulation optimization strategies, key feature information that significantly impacts system performance is selected, such as the system's maximum pressure, average lifting speed, total energy consumption, and temperature variation range of each component. This key feature information is input into the pre-simulation iteration unit of the AI ​​prediction model. The unit compares and analyzes this information with preset ideal operating state features. If certain features do not meet the ideal state, the direction and step size of parameter adjustments in the next round are determined based on the discrepancies. For example, if the average lifting speed in the predicted operating state is lower than the ideal value, it might be determined that the "output flow rate Q" parameter needs to be increased in the second round of pre-simulation, or the "hydraulic valve opening degree K" parameter needs to be adjusted to increase the lifting speed. The adjustment step size is optimized based on the magnitude of the speed difference.

[0040] Step S116: According to the parameter adjustment direction and adjustment step size of the second round of pre-exercise, the core parameters in the first round of pre-exercise optimization strategy are readjusted again to generate readjusted parameter values. The readjusted parameter values ​​are substituted into the first round of pre-exercise optimization strategy to replace the original parameter values ​​and generate the second round of pre-exercise optimization strategy. The second round of pre-exercise optimization strategy includes all the readjusted control parameters and parameter association logic.

[0041] Based on the adjustment direction and step size of the second round of pre-simulation parameters generated in step S115, the core parameters in the first round of pre-simulation optimization strategy are adjusted again. Assuming that analysis indicates the "hydraulic valve opening degree K" needs adjustment, with the adjustment direction being to increase and the adjustment step size being a certain set value, then the parameter value of "hydraulic valve opening degree K" in the first round of pre-simulation optimization strategy is added to this adjustment step size to obtain the adjusted parameter value K1. K1 is then substituted into the first round of pre-simulation optimization strategy, replacing the original "hydraulic valve opening degree K" parameter value. The parameter association logic is checked and adjusted to ensure that the new parameter value has a reasonable association with other parameters, ultimately generating the second round of pre-simulation optimization strategy.

[0042] Step S117: Input the second round of pre-simulation optimization strategy into the operation state pre-simulation unit of the AI ​​prediction model to simulate the complete operation process of the lifting hydraulic system under the control of the second round of pre-simulation optimization strategy, and generate a set of predicted operation states corresponding to the second round of pre-simulation optimization strategy.

[0043] Similar to step S114, the second round of pre-simulation optimization strategy is input into the operation state pre-simulation unit. The virtual simulation model of the operation state pre-simulation unit will simulate the complete operation process of the lifting hydraulic system again according to the parameter configuration and correlation logic in the strategy, record the operation characteristics of each component and the overall operation trend of the system, and generate the predicted operation state corresponding to the second round of pre-simulation optimization strategy.

[0044] Step S118: Based on the predicted operating state characteristics corresponding to the second round of pre-exercise optimization strategy, generate the parameter adjustment direction and adjustment step size for the third round of pre-exercise, adjust the parameters of the second round of pre-exercise optimization strategy, generate the third round of pre-exercise optimization strategy and pre-exercise the corresponding operating state, and perform multiple iterations in this manner. The parameter adjustment direction and adjustment step size of each round are optimized and generated based on the predicted operating state characteristics of the previous round.

[0045] Step S1181: Set the termination condition for the strategy pre-exercise iteration. The termination condition is that the degree of fit between the predicted operating state corresponding to the pre-exercise optimization strategy and the ideal operating state of the lifting hydraulic system reaches a preset standard value. Record the current round of pre-exercise iteration. The initial round is the second round, and the current pre-exercise optimization strategy is the second round pre-exercise optimization strategy.

[0046] A predetermined termination condition is set: the pre-exercise iteration process stops when the degree of fit between the predicted operating state and the ideal operating state of the system corresponding to a certain round of pre-exercise optimization strategy reaches a preset standard value. For example, under the ideal operating state, the system's average lifting speed is V0, energy consumption is E0, and pressure fluctuation range is P0-P1, etc. The degree of fit is a value obtained by comprehensively considering the matching degree of these characteristics. When this value is greater than or equal to the preset standard value, the iteration terminates. Initially, the current pre-exercise iteration round is set to the second round, and the current pre-exercise optimization strategy is the second round pre-exercise optimization strategy.

[0047] Step S1182: Extract key feature information from the predicted running state corresponding to the current pre-exercise optimization strategy, and generate the predicted state feature vector for the current round. The predicted state feature vector for the current round contains all the key feature information of the predicted running state for the current round.

[0048] From the predicted operating status corresponding to the second round of pre-exercise optimization strategy, key feature information such as average lifting speed, total energy consumption, maximum pressure value, and temperature changes of each component are extracted. The above information is organized in a certain order to form a vector data structure, namely the predicted state feature vector of the current round (second round). This predicted state feature vector can comprehensively reflect the key features of the current predicted operating status.

[0049] Step S1183: Compare the predicted state feature vector of the current round with the ideal operating state feature vector of the lifting hydraulic system, calculate the degree of fit between the predicted state feature and the ideal operating state feature of the current round, and generate the degree of fit score of the current round. The degree of fit score is generated based on the degree of fit between the predicted state feature and the ideal state feature.

[0050] The ideal operating state feature vector is pre-defined based on the design goals and performance requirements of the lift's hydraulic system, and includes characteristic information such as ideal average lifting speed, energy consumption, and pressure. The predicted state feature vector for the current round (second round) is compared one by one with the corresponding features in the ideal operating state feature vector, calculating the fit degree for each feature. Then, the fit degrees of all features are combined to obtain an overall fit degree value, which is used to generate the fit degree score for the current round. For example, if the fit degrees of each feature are high, the overall fit degree value is large, and the corresponding fit degree score is high.

[0051] Step S1184: If the fit score of the current round does not reach the preset standard value in the termination condition, extract the feature dimensions in the current round's predicted state feature vector that differ from the ideal running state feature vector by more than a preset difference range, and generate a list of difference feature dimensions. The list of difference feature dimensions contains the identification information of all feature dimensions whose differences exceed the preset difference range.

[0052] If the fit score for the current round (second round) calculated in step S1183 is lower than the preset standard value, it indicates that there is still a gap between the predicted operating state and the ideal state. At this point, it is necessary to identify which specific feature dimensions have significant differences. The predicted state feature vector and the ideal operating state feature vector are compared for each corresponding feature dimension, and the difference value is calculated. If the difference value exceeds a preset difference range, the identifier information of that feature dimension is recorded, forming a list of differing feature dimensions. For example, if the difference in average lifting speed exceeds the preset range, then the identifier of the feature dimension "average lifting speed" will be added to the list.

[0053] Step S1185: Based on the list of differential feature dimensions, identify the control parameters related to the differential feature dimensions and generate a list of parameters to be adjusted. The list of parameters to be adjusted contains complete identification information of all control parameters related to the differential feature dimensions.

[0054] Based on the feature dimension identifiers in the list of differential feature dimensions, analyze which control parameters affect the feature dimension. For example, the differential feature dimension "average lifting speed" may be related to the control parameters "output flow rate Q" and "hydraulic valve opening degree K". By querying the parameter-feature dimension correlation table, find all control parameters related to the differential feature dimension, record the complete identifier information of these parameters, and generate a list of parameters to be adjusted, such as [output flow rate Q (complete identifier), hydraulic valve opening degree K (complete identifier), ...].

[0055] Step S1186: Extract the influence weight information of each parameter in the list of parameters to be adjusted from the parameter influence weight sequence, determine the adjustment priority of the parameters to be adjusted in descending order of influence weight value, and generate the parameter adjustment priority sequence.

[0056] In the parameter influence weight sequence generated in step S1119, the influence weight information corresponding to each parameter in the list of parameters to be adjusted is found. Then, the above parameters are arranged in descending order of their influence weight values ​​to obtain the parameter adjustment priority sequence. For example, if the influence weight of "output flow rate Q" in the list of parameters to be adjusted is W1, and the influence weight of "hydraulic valve opening degree K" is W2, and W1>W2, then the parameter adjustment priority sequence is [output flow rate Q (W1), hydraulic valve opening degree K (W2), ...].

[0057] Step S1187: Based on the parameter adjustment priority sequence and the predicted state feature vector of the current round, generate the parameter adjustment direction and adjustment step size for the next round of pre-exercise. The parameter adjustment direction and adjustment step size for the next round are determined based on the optimization requirements of the difference feature dimension, and the corresponding parameters are adjusted in order of adjustment priority from high to low.

[0058] Combining the parameter adjustment priority sequence and the predicted state feature vector of the current round (second round), for each parameter to be adjusted, the adjustment direction and step size are determined based on the optimization requirements of the difference feature dimension. For example, for the highest priority "output flow rate Q", if the difference feature dimension "average lift speed" is lower than the ideal value, and increasing "output flow rate Q" can improve the lift speed, then the adjustment direction is to increase it. The adjustment step size is determined comprehensively based on the degree of difference in lift speed and the sensitivity of "output flow rate Q". The adjustment direction and step size are determined for each parameter in descending order of the parameter adjustment priority sequence.

[0059] Step S1188: Adjust the parameters to be adjusted in the current round of the pre-exercise optimization strategy according to the parameter adjustment direction and adjustment step size of the next round, generate the adjusted parameter values ​​for the next round, substitute the adjusted parameter values ​​for the next round into the current round of the pre-exercise optimization strategy, replace the original parameter values, and generate the next round of the pre-exercise optimization strategy.

[0060] Based on the adjustment direction and step size of the next round (third round) of pre-exercise parameters determined in step S1187, the parameters to be adjusted in the current round (second round) of pre-exercise optimization strategy are adjusted. For example, the "output flow rate Q" is adjusted in the direction of increase and with a determined step size to obtain a new parameter value Q2, and the "hydraulic valve opening degree K" is adjusted in the corresponding direction and step size to obtain K2. The above new parameter values ​​are substituted into the second round of pre-exercise optimization strategy, replacing the original parameter values, and ensuring that the parameter association logic adapts to the new parameter values, thereby generating the third round of pre-exercise optimization strategy.

[0061] Step S1189: Input the next round of pre-simulation optimization strategy into the running state pre-simulation unit of the AI ​​prediction model, simulate the complete operation process of the lifting hydraulic system under the control of the corresponding pre-simulation optimization strategy, and generate a set of predicted running states corresponding to the next round of pre-simulation optimization strategy.

[0062] The third round of pre-simulation optimization strategy generated in step S1188 is input into the running state pre-simulation unit. The virtual simulation model of the running state pre-simulation unit will simulate the complete running process of the system under the control of the strategy, record the running characteristics of each component and the overall running trend of the system, and generate the predicted running state corresponding to the third round of pre-simulation optimization strategy.

[0063] Step S11810: Increment the current round of the pre-exercise iteration by 1 to update it to the new current round. Use the next round of pre-exercise optimization strategy as the current round's pre-exercise optimization strategy. Repeat the steps from extracting the predicted state feature vector to generating the next round's predicted running state until the fit score of the current round reaches the preset standard value in the termination condition.

[0064] The current pre-exercise iteration round is increased by 1 from the second round to the third round, and the current pre-exercise optimization strategy is updated to the third round pre-exercise optimization strategy. Then, the operations of steps S1182 to S1189 are repeated, that is, extracting the key feature information of the predicted running state of the third round to generate the predicted state feature vector, comparing it with the ideal state feature vector to calculate the fit score. If the termination condition is not met, the list of difference feature dimensions and the list of parameters to be adjusted are generated to determine the parameter adjustment direction and step size of the fourth round of pre-exercise. The parameters are adjusted to generate the fourth round pre-exercise optimization strategy and perform pre-exercise. This cycle is repeated until the fit score of a certain round reaches the preset standard value, and the iteration stops.

[0065] Step S119: Record the complete parameter configuration and corresponding predicted running status of the pre-exercise optimization strategy for each round, forming a pre-exercise iteration record set. The pre-exercise iteration record set contains the association information between the pre-exercise optimization strategy and the predicted running status for all rounds.

[0066] After each round of pre-simulation optimization strategy generation and operational state simulation is completed, the complete parameter configuration of the optimization strategy for that round (including the adjusted values ​​of all control parameters and parameter association logic) and the corresponding predicted operational state information are associated and stored to form a record. After multiple rounds of pre-simulation iterations, all these records are collected together to form a pre-simulation iteration record set, which can clearly show the correspondence between each round of strategy adjustment and the predicted operational state.

[0067] Step S1110: Integrate all pre-exercise optimization strategies in the pre-exercise iteration record set, remove duplicate strategy configurations, and generate multi-round pre-exercise optimization strategies. Each round of pre-exercise optimization strategy in the multi-round pre-exercise optimization strategy corresponds to a set of predicted operating states of the lifting hydraulic system.

[0068] All pre-exercise optimization strategies in the pre-exercise iteration record set are checked, and the parameter configurations of strategies in different rounds are compared. If there are strategies with completely identical parameter configurations, only one record is kept, and duplicate strategy configurations are removed. After deduplication, all the resulting pre-exercise optimization strategies constitute a multi-round pre-exercise optimization strategy, where each strategy has its unique parameter configuration and corresponding predicted running state.

[0069] Step S120: Extract the real-time operation feedback features of the lifting machine hydraulic system, input the multi-round pre-simulation optimization strategy and the real-time operation feedback features of the lifting machine hydraulic system into the dynamic adaptation coupling unit, and comprehensively evaluate the adaptability of each round of pre-simulation optimization strategy through the dynamic adaptation coupling unit to generate a strategy adaptation priority sequence. The strategy adaptation priority sequence includes the adaptation ranking information of all pre-simulation optimization strategies and the corresponding strategy features and predicted state feature information.

[0070] During the actual operation of the lifting platform's hydraulic system, sensors and other devices installed on various components collect real-time operating data, such as temperature, pressure, speed, and displacement of each component. Real-time operational feedback features reflecting the system's current operating status are extracted from this data. Then, the multi-round pre-simulation optimization strategy generated in step S1110, along with these real-time operational feedback features, are input into the dynamic adaptation coupling unit. The dynamic adaptation coupling unit evaluates the adaptability of each round of pre-simulation optimization strategy to the current real-time operating state of the system. The evaluation may include whether the strategy meets the current system's operational requirements, whether it is safe and reliable, and whether it is economical and efficient. Finally, based on the evaluation results, all pre-simulation optimization strategies are ranked, generating a strategy adaptation priority sequence.

[0071] Step S121: Extract the complete parameter configuration of each round of pre-simulation optimization strategy in the multi-round pre-simulation optimization strategy, generate the strategy feature vector of each round of pre-simulation optimization strategy. The strategy feature vector contains the configuration information of all control parameters of the corresponding pre-simulation optimization strategy and the parameter association logic features. Extract the key feature information in the predicted running state corresponding to each round of pre-simulation optimization strategy, and generate the predicted state feature vector of each round of pre-simulation optimization strategy.

[0072] For each round of the multi-round pre-simulation optimization strategy, its complete parameter configuration is first extracted, including the specific values ​​of all control parameters and the logical relationships between these parameters. This information is then transformed into a vector form, namely the strategy feature vector. For example, the parameter configuration of a certain round of strategy might be output flow rate Q1, hydraulic valve opening degree K1, hydraulic cylinder stroke S1, etc., and the parameter relationship logic might be how K1 changes when Q1 reaches a certain value. This information is encoded into a vector according to preset rules. Simultaneously, key feature information, such as average lifting speed, energy consumption, and pressure, is extracted from the predicted operating state corresponding to that round of strategy and also organized into a vector form, namely the predicted state feature vector.

[0073] Step S122: Associate and bind the policy feature vector of each round of pre-exercise optimization strategy with the corresponding predicted state feature vector to generate the policy state association vector of each round of pre-exercise optimization strategy. The policy state association vector contains complete association information between the policy features and the predicted state features.

[0074] The policy feature vector and the corresponding predicted state feature vector of each round of pre-exercise optimization strategy generated in step S121 are associated and bound in a certain way, such as concatenating the two vectors in sequence or adding common identification information to the two vectors, so that it can be clearly known which predicted state feature vector a certain policy feature vector corresponds to, and vice versa. The resulting policy-state association vector contains complete association information between the policy features and predicted state features of the pre-exercise optimization strategy in that round.

[0075] Step S123: Extract the real-time operation feedback features of the lifting hydraulic system and generate a real-time feedback feature vector. The real-time feedback feature vector contains the operation features of each component and the overall operation trend features of the lifting hydraulic system during the current operation process, which serve as the benchmark features for dynamic adaptation and coupling evaluation.

[0076] The system collects real-time operating data of various components of the lifting hydraulic system using sensors and other devices, such as the real-time flow rate of the hydraulic pump, the real-time opening degree of the hydraulic valve, the real-time displacement of the hydraulic cylinder, and the real-time pressure and temperature of the system. From this real-time data, the operating characteristics of each component (such as the flow fluctuation characteristics of the hydraulic pump and the response speed characteristics of the hydraulic valve) and the overall operating trend characteristics of the system (such as the overall energy consumption trend and the stability trend of the system pressure) are extracted. The above characteristic information is organized into a vector form, namely the real-time feedback feature vector, which serves as a benchmark for evaluating the adaptability of the pre-simulation optimization strategy.

[0077] Step S124: Perform coupling analysis on the policy state association vector of each round of pre-exercise optimization strategy in the multi-round pre-exercise optimization strategy and the real-time feedback feature vector respectively. Analyze the degree of fit between the policy features, predicted state features in the policy state association vector of each round and the real-time feedback features in the real-time feedback feature vector, and generate the coupling matching degree of each round of pre-exercise optimization strategy.

[0078] For each round of pre-optimization strategy, the strategy state association vector is compared with the corresponding features in the real-time feedback feature vector to analyze whether the strategy features match the current real-time operating conditions of the system. Simultaneously, the predicted state features are compared with the real-time feedback features to analyze the similarity between the predicted operating state and the current actual operating state of the system. Combining the results of these two comparisons, the degree of fit between the pre-optimization strategy and the real-time feedback features, i.e., the coupling matching degree, is evaluated. For example, if the parameter configuration in the strategy features matches the current system's hardware conditions and operating requirements, and most features in the predicted state features are similar to those in the real-time feedback features, then the coupling matching degree is high.

[0079] Step S125: Extract the predicted operating state of each round of pre-simulation optimization strategy, analyze the degree of fit between the predicted operating state of each round of pre-simulation optimization strategy and the safe operating boundary of the lifting hydraulic system, and generate the safety adaptability of each round of pre-simulation optimization strategy. The safety adaptability is used to evaluate the adaptability of each round of pre-simulation optimization strategy at the safety level.

[0080] The hydraulic system of the lifting platform has preset safe operating boundaries, such as maximum allowable pressure, maximum allowable temperature, and minimum safe distance. For each round of pre-optimization strategy, relevant features in its predicted operating state are extracted, such as the predicted maximum system pressure and the predicted temperature of each component. These features are then compared with the safe operating boundaries. If all features in the predicted operating state are within the safe operating boundaries and have a certain safety margin, then the safety fit of that round of strategy is high; if some features are close to or exceed the safe operating boundaries, the safety fit is low.

[0081] Step S126: Extract the parameter configuration of each round of pre-simulation optimization strategy, analyze the degree of fit between the parameter configuration of each round of pre-simulation optimization strategy and the hardware performance boundary of the lifting hydraulic system, and generate the hardware adaptability of each round of pre-simulation optimization strategy. The hardware adaptability is used to evaluate the adaptability of each round of pre-simulation optimization strategy at the hardware level.

[0082] The system's hardware performance boundaries include the maximum working capacity of each hardware component, such as the maximum output flow of the hydraulic pump, the maximum opening degree of the hydraulic valve, and the maximum stroke of the hydraulic cylinder. The analysis examines whether the parameter configurations of each round of pre-optimization strategy are within the hardware performance boundaries. For example, if the configured value of "output flow Q" in a certain round of strategy exceeds the maximum output flow of the hydraulic pump, then the strategy cannot be implemented at the hardware level, resulting in low hardware adaptability; conversely, if all parameter configurations are within the hardware performance boundaries, the hardware adaptability is high.

[0083] Step S127: Based on the set weight ratio, perform weighted calculations on the coupling matching degree, security adaptability and hardware adaptability of each round of pre-exercise optimization strategy to generate a comprehensive adaptability score for each round of pre-exercise optimization strategy. Sort the pre-exercise optimization strategies in descending order of comprehensive adaptability score to generate strategy adaptability ranking information. The strategy adaptability ranking information includes the identifier of each round of pre-exercise optimization strategy and its corresponding comprehensive adaptability score and ranking position.

[0084] The weighting of coupling matching, security adaptability, and hardware adaptability in the comprehensive evaluation is preset, for example, coupling matching accounts for 40%, security adaptability accounts for 40%, and hardware adaptability accounts for 20%. For each round of pre-optimization strategy, its coupling matching, security adaptability, and hardware adaptability are multiplied by their respective weightings, and then the products are added together to obtain the comprehensive adaptability score of that round of strategy. For example, if the coupling matching of a certain round of strategy is M1, the security adaptability is S1, and the hardware adaptability is H1, with weightings of Wm, Ws, and Wh respectively, then the comprehensive adaptability score = M1×Wm + S1×Ws + H1×Wh. After calculating the comprehensive adaptability scores of all rounds of strategies, the strategies are sorted in descending order of score, and the identifier, comprehensive adaptability score, and position in the ranking of each round of strategy are recorded to generate strategy adaptability ranking information.

[0085] Step S128: Associate and bind the strategy adaptation ranking information with the complete information of the multi-round pre-exercise optimization strategy. The associated and bound content includes the identifier, comprehensive adaptation score, ranking position, strategy feature vector, predicted state feature vector, and strategy state association vector of each round of pre-exercise optimization strategy.

[0086] Each record in the strategy adaptation ranking information generated in step S127 (including strategy identifier, comprehensive adaptation score, and ranking position) is associated with the complete information of the corresponding strategy in the multi-round pre-exercise optimization strategy (including strategy feature vector, predicted state feature vector, and strategy state association vector), so that the ranking information and all feature vector information of the strategy can be obtained simultaneously through the strategy identifier, forming a complete associated dataset.

[0087] Step S129: Based on the information after association and binding, generate a strategy adaptation priority sequence, which includes the adaptation ranking information of all pre-optimization strategies and the corresponding strategy features and prediction state feature information.

[0088] The information associated and bound in step S128 is organized in descending order of sorting position to form a strategy adaptation priority sequence. Each item in this strategy adaptation priority sequence contains the adaptation ranking information (identifier, comprehensive adaptation score, sorting position) of a certain round of pre-exercise optimization strategy, as well as the corresponding strategy feature vector and predicted state feature vector, which can clearly show the adaptation priority and related features of all pre-exercise optimization strategies.

[0089] Step S130: Based on the policy adaptation priority sequence, trigger the policy refinement process of the AI ​​prediction model, perform feature fusion, core feature enhancement and parameter convergence processing on the top K pre-simulation optimization policies in the policy adaptation priority sequence, and generate a target optimization control policy. The target optimization control policy includes all optimized control parameter configuration information and parameter association logic features.

[0090] Based on the policy adaptation priority sequence, the top K pre-implementation optimization policies are selected, which are considered to have high adaptability. These policies are then input into the policy refinement process of the AI ​​prediction model. First, their policy features and predicted state features are fused to integrate the advantages of each policy. Next, core features are identified and strengthened, increasing the importance of core parameters and related state features. Finally, the fused parameters undergo convergence processing, removing redundant parameters and optimizing parameter correlation logic, ultimately generating a target optimization control policy with superior overall performance.

[0091] Step S131: Parse the policy adaptation ranking information in the policy adaptation priority sequence, extract the top K ranked pre-optimization policies, and generate a core pre-optimization policy set. The core pre-optimization policy set contains complete information of all the top K ranked pre-optimization policies, including policy feature vectors, predicted state feature vectors, and policy state association vectors.

[0092] The policy adaptation priority sequence is analyzed, and the top K ranked pre-implementation optimization policies are extracted based on their ranking position. For example, if K=3, the policies ranked first, second, and third are extracted. The complete information of these policies, including their policy feature vectors, predicted state feature vectors, and policy-state correlation vectors, is then compiled to form a core pre-implementation optimization policy set.

[0093] Step S132: Extract the policy feature vector of each pre-simulation optimization strategy in the core pre-simulation optimization strategy set, input all policy feature vectors into the feature fusion unit of the AI ​​prediction model, perform cross-fusion processing of feature vectors, fuse the policy features of all core pre-simulation optimization strategies, and generate a fused policy feature vector. The fused policy feature vector contains the feature information of all core pre-simulation optimization strategies.

[0094] The feature fusion unit of the AI ​​prediction model receives the policy feature vector of each policy in the core pre-optimization policy set and processes these vectors using a cross-fusion method. For example, for each feature dimension in the vector, it may take the average, maximum, or other statistical value of all policies in that dimension, or learn the weights of each policy feature through an algorithm and then perform a weighted combination, thereby fusing multiple policy feature vectors into a new fused policy feature vector. This fused policy feature vector contains the policy feature information of all core pre-optimization policies.

[0095] Step S133: Extract the predicted state feature vector corresponding to each pre-exercise optimization strategy in the core pre-exercise optimization strategy set, input all predicted state feature vectors into the feature fusion unit of the AI ​​prediction model, perform cross-fusion processing of feature vectors, fuse the predicted state features corresponding to all core pre-exercise optimization strategies, and generate a fused predicted state feature vector.

[0096] Similar to step S132, the predicted state feature vector corresponding to each strategy in the core pre-optimization strategy set is input into the feature fusion unit. Through cross-fusion processing, such as statistical analysis or weighted combination of the corresponding feature dimensions of each vector, a fused predicted state feature vector that integrates the predicted state features of all core strategies is generated.

[0097] Step S134: Associate and bind the fusion strategy feature vector with the fusion prediction state feature vector to generate the fusion strategy state association vector. Extract the fusion strategy feature vector from the fusion strategy state association vector. Identify the control parameter features in the fusion strategy feature vector. Retrieve the parameter influence weight sequence. Set the influence weight threshold. Determine the control parameter features whose influence weight values ​​are greater than the influence weight threshold as core control parameter features. Strengthen the core control parameter features in the fusion strategy features to increase the weight ratio of the core control parameter features in the fusion strategy features and generate the strengthened strategy feature vector.

[0098] The fusion strategy feature vector and the fusion predicted state feature vector are associated and bound, for example, through a common identifier or location correspondence, to form a fusion strategy state association vector. The fusion strategy feature vector is extracted from this association vector, and the control parameter features it contains, such as output flow rate and hydraulic valve opening degree, are analyzed. The previously generated parameter influence weight sequence is retrieved, and an influence weight threshold is set. Features corresponding to control parameters whose influence weight values ​​in the parameter influence weight sequence are greater than this threshold are identified as core control parameter features. Then, the weight ratio of these core control parameter features in the fusion strategy feature vector is increased to perform enhancement processing, for example, giving these features higher consideration weight in subsequent parameter calculations or decision-making processes, thereby generating an enhanced strategy feature vector.

[0099] Step S1341: Extract the fusion strategy feature vector from the fusion strategy state association vector, identify the control parameter features in the fusion strategy feature vector, and record the identifier and corresponding feature weight of each control parameter feature. The feature weight represents the importance of the corresponding control parameter feature in the fusion strategy features.

[0100] The fusion strategy feature vector is extracted from the fusion strategy state association vector. This vector is then parsed to identify the various control parameter features it contains, such as output flow characteristics and hydraulic valve opening degree characteristics. The identifier for each control parameter feature is recorded, such as "output flow Q feature" and "hydraulic valve opening degree K feature". Simultaneously, the feature weight of each control parameter feature in the fusion strategy feature vector is determined. This feature weight reflects the importance of the feature in the current fusion strategy features and may be automatically generated during the feature fusion process.

[0101] Step S1342: Retrieve the parameter influence weight sequence, extract the influence weight information corresponding to each control parameter feature in the fusion strategy feature vector from the parameter influence weight sequence, set the influence weight threshold, and determine the control parameter features whose influence weight values ​​are greater than the influence weight threshold as core control parameter features.

[0102] The parameter influence weight sequence is retrieved from storage. Based on the identifiers of each control parameter feature in the fusion strategy feature vector, the corresponding influence weight information is searched and extracted from the parameter influence weight sequence. Then, a preset influence weight threshold is established, and control parameter features with influence weight values ​​greater than this threshold are filtered out and identified as core control parameter features. For example, if the influence weight threshold is set to a certain value, and the influence weight of the "output flow Q feature" in the parameter influence weight sequence is W1, and W1 is greater than the threshold, then the "output flow Q feature" is identified as a core control parameter feature.

[0103] Step S1343: Obtain the enhancement weight coefficient of the core control parameter feature, multiply the feature weight of each core control parameter feature by the corresponding enhancement weight coefficient to generate the enhanced feature weight. The enhanced feature weight represents the new importance of the core control parameter feature in the fusion strategy feature.

[0104] A pre-set enhancement weight coefficient is assigned to each core control parameter feature. This enhancement weight coefficient is greater than 1 and is used to increase the weight of the core control parameter feature. The feature weight of the core control parameter feature recorded in step S1341 is multiplied by its corresponding enhancement weight coefficient to obtain the enhanced feature weight. For example, if the original feature weight of the "output flow Q feature" is Wf and the enhancement weight coefficient is Wr (Wr>1), then the enhanced feature weight is Wf×Wr.

[0105] Step S1344: Keep the feature weights of the non-core control parameter features in the fusion strategy feature vector unchanged, integrate the feature weights of the enhanced core control parameter features with the feature weights of the non-core control parameter features to generate an enhanced strategy feature vector, wherein the enhanced strategy feature vector contains all enhanced control parameter features and their corresponding feature weights.

[0106] For non-core control parameter features in the fusion strategy feature vector, their feature weights remain unchanged. The feature weights of the enhanced core control parameter features obtained in step S1343 and the feature weights of the non-core control parameter features are reorganized according to the original vector structure to form an enhanced strategy feature vector, in which the weights of the core control parameter features are increased.

[0107] Step S1345: Standardize the format of the reinforcement strategy feature vector, unify the expression of each element in the reinforcement strategy feature vector, extract the core control parameter features in the reinforcement strategy feature vector, calculate the change in the weight ratio of the core control parameter features before and after reinforcement, verify the reinforcement effect of the core control parameter features, and confirm that the change in the weight ratio of the core control parameter features in the fusion strategy features reaches the preset change range.

[0108] The generated reinforcement strategy feature vector is standardized to ensure consistent representation of each element for easier subsequent processing. Then, the core control parameter features are extracted from the reinforcement strategy feature vector, and the change in their weight percentages before and after reinforcement is calculated as (weight percentage after reinforcement - weight percentage before reinforcement). This change is compared to a preset range. If the change is within the preset range, the reinforcement effect is as expected; otherwise, the reinforcement weight coefficients need to be readjusted, and steps S1343 and S1344 are repeated until the change in the weight percentages of the core control parameter features reaches the preset range.

[0109] Step S1346: If the enhancement effect of the core control parameter features does not reach the preset change range, adjust the enhancement weight coefficient of the core control parameter features, recalculate the feature weights of the core control parameter features, and generate new enhanced feature weights.

[0110] If the verification result of step S1345 shows that the enhancement effect has not reached the preset change range, for example, the increase in weight ratio is insufficient, then it is necessary to increase the enhancement weight coefficient of the core control parameter feature. Substitute the adjusted enhancement weight coefficient back into step S1343 to recalculate the feature weight of the core control parameter feature to obtain the new enhanced feature weight.

[0111] Step S1347: Integrate the feature weights of the new enhanced core control parameter features with the feature weights of the non-core control parameter features to generate a new enhanced strategy feature vector.

[0112] Using the feature weights of the new enhanced core control parameter features obtained in step S1346, combined with the feature weights of the non-core control parameter features, and re-integrated according to the method in step S1344, a new enhanced strategy feature vector is generated.

[0113] Step S1348: The final enhancement strategy feature vector is taken as the result of feature enhancement processing. The enhancement strategy feature vector contains all enhanced control parameter features and corresponding feature weights.

[0114] After verification in step S1345, when the enhancement effect of the core control parameter features reaches the preset change range, the enhancement strategy feature vector at this time is the final feature enhancement processing result. The enhancement strategy feature vector contains all control parameter features and their adjusted feature weights.

[0115] Step S135: Extract the fusion prediction state feature vector from the fusion strategy state association vector, identify the state features related to the core control parameter features in the fusion prediction state feature vector, strengthen the state features related to the core control parameter features, increase the weight ratio of the relevant state features in the fusion prediction state features, and generate the strengthened prediction state feature vector.

[0116] The fusion prediction state feature vector is extracted from the fusion strategy state association vector. State features such as average lift speed, energy consumption, and pressure are analyzed to identify state features related to the core control parameter features determined in step S134. For example, if the "output flow rate Q feature" is a core control parameter feature, then related features such as the "average lift speed state feature" and "system flow rate state feature" are state features that need to be strengthened. Strengthening is achieved by increasing the weight of these related state features in the fusion prediction state feature vector, generating a strengthened prediction state feature vector.

[0117] Step S136: Associate and bind the reinforcement strategy feature vector with the reinforcement prediction state feature vector to generate the reinforcement strategy state association vector. Input the reinforcement strategy state association vector into the parameter convergence unit of the AI ​​prediction model to perform convergence processing on the control parameters in the fused strategy features, remove redundant parameter configurations, optimize the association logic between parameters, and generate the converged strategy state association vector.

[0118] The reinforcement strategy feature vector and the reinforcement prediction state feature vector are associated and bound to form a reinforcement strategy state association vector. This association vector is input into the parameter convergence unit, which analyzes the control parameters in the reinforcement strategy feature vector, identifies and removes redundant parameter configurations, such as those with little impact on the system's operating state or those highly correlated with other parameters. Simultaneously, based on the reinforcement prediction state feature vector and the system's operating objective, the association logic between parameters is optimized to make the coordination between parameters more harmonious. Finally, a converged strategy state association vector is generated, which contains the converged control parameter configuration and the optimized parameter association logic.

[0119] Step S137: Extract the convergence strategy feature vector from the convergence strategy state association vector. The convergence strategy feature vector contains all control parameter configuration information and parameter association logic features after convergence. Input the convergence strategy feature vector into the running state pre-simulation unit of the AI ​​prediction model to simulate the complete operation process of the lifting hydraulic system under the corresponding strategy control and generate the convergence strategy verification running state.

[0120] The convergence strategy feature vector is extracted from the convergence strategy state association vector. This feature vector contains all control parameter configuration information and parameter association logic features after convergence processing. This feature vector is then input into the operational state pre-simulation unit of the AI ​​prediction model. The unit uses its internal virtual simulation model to simulate the complete operation process of the lifting platform's hydraulic system under the control of this convergence strategy, including stages such as start-up, operation, and shutdown. It then generates a corresponding convergence strategy verification operational state, which contains the predicted operational characteristics of the system under this strategy.

[0121] Step S138: Extract key feature information from the convergence strategy verification operation state, compare the key feature information with the real-time operation feedback features of the lifting hydraulic system, and perform final parameter fine-tuning on the convergence strategy feature vector based on the comparison results to optimize the parameter configuration and correlation logic of the convergence strategy.

[0122] Key feature information, such as predicted average lifting speed, energy consumption, and pressure, is extracted from the convergence strategy verification operation status. This information is then compared with the corresponding features in the real-time feedback feature vector generated in step S123 to analyze the differences. Based on these differences, minor adjustments are made to the control parameter configuration in the convergence strategy feature vector, such as fine-tuning the values ​​of parameters like output flow rate and hydraulic valve opening degree. The parameter association logic is also optimized accordingly to ensure that the adjusted strategy more closely approximates the system's real-time operating state during simulation.

[0123] Step S139: Generate the target optimization control strategy based on the convergence strategy feature vector after parameter fine-tuning.

[0124] After parameter fine-tuning in step S138, a convergence strategy feature vector with better performance is obtained. The control parameter configuration information and parameter association logic features contained in the vector are extracted and organized into the format of the lifting hydraulic system control strategy to generate the target optimization control strategy.

[0125] Step S140: Convert the target optimization control strategy into control commands that can be recognized by each actuator of the lifting hydraulic system, transmit the control commands to the actuator control unit of the lifting hydraulic system, drive each actuator to operate according to the requirements of the target optimization control strategy, realize the application of the target optimization control strategy in the lifting hydraulic system, and extract the real-time operation feedback features after the application.

[0126] The target-optimized control strategy is a high-level parameter configuration and logical description that needs to be converted into control commands that can be directly recognized and executed by each actuator (such as hydraulic pumps, hydraulic valves, and hydraulic cylinders). These control commands may include start / stop signals and speed setpoints for hydraulic pumps, on / off signals and opening degree setpoints for hydraulic valves, and extension / retraction signals and stroke setpoints for hydraulic cylinders. After conversion, the control commands are sent to the execution control unit of the lift's hydraulic system via a communication bus or other transmission method. The execution control unit controls the actions of each actuator according to the commands, enabling the system to operate according to the target-optimized control strategy. During operation, the system's operational feedback characteristics are extracted in real time as the basis for the next round of optimization.

[0127] Step S150: Taking the applied target optimization control strategy as the new iteration starting point, and combining the real-time operation feedback characteristics after application, repeatedly execute all steps of the multi-round strategy pre-show iteration process, dynamic adaptation coupling evaluation, and strategy refinement process to continuously generate the optimized target optimization control strategy and apply it to the lift hydraulic system to realize the online optimization of the lift hydraulic control strategy.

[0128] The target optimization control strategy applied to the lifting machine's hydraulic system in step S140 serves as the starting point for the next round of strategy optimization. Simultaneously, combining the real-time operational feedback features extracted after the strategy application, and following the process from steps S110 to S139, multiple rounds of strategy pre-simulation iteration, dynamic adaptation coupling evaluation, and strategy refinement are executed again to generate a new target optimization control strategy, which is then applied to the system. This process is repeated continuously, constantly optimizing the control strategy based on the system's real-time operational status, enabling the control strategy to continuously adapt to system changes and achieve online optimization.

[0129] Step S151: Using the applied target optimization control strategy as the new iteration starting point, extract the real-time operation feedback features after application, and generate the real-time feedback feature vector after application. The real-time feedback feature vector after application includes the operation features of each component of the lifting hydraulic system under the control of the target optimization control strategy and the overall operation trend features.

[0130] After the target optimization control strategy is applied to the lifting machine's hydraulic system and operates for a period of time, the strategy itself is used as the starting point for a new iteration. Simultaneously, real-time operational feedback features are extracted from the system's operation. These features include the operational characteristics of each component (such as hydraulic pumps, hydraulic valves, and hydraulic cylinders), such as the actual output flow fluctuation characteristics of the hydraulic pump and the actual response time characteristics of the hydraulic valve, as well as the overall operational trend characteristics of the system, such as the overall energy consumption change trend over time and the stability trend of system pressure. These features are organized into a vector form, i.e., the real-time feedback feature vector after application.

[0131] Step S152: Input the real-time feedback feature vector after application into the pre-simulation initialization unit of the AI ​​prediction model to generate a new pre-simulation initial parameter adjustment direction and adjustment step size. The new pre-simulation initial parameter adjustment direction and adjustment step size are optimized and generated based on the real-time feedback feature vector after application, and are used to guide the parameter adjustment of the target optimization control strategy after application.

[0132] The real-time feedback feature vector after application is input into the pre-simulation initialization unit of the AI ​​prediction model. The pre-simulation initialization unit analyzes the difference between this feature vector and the feature vector of the ideal operating state, and determines the new direction (increase or decrease) and adjustment step size of the pre-simulation initial parameters based on the difference. For example, if the average lifting speed in the real-time feedback feature vector after application is lower than the ideal value, the adjustment direction may be determined to increase the relevant parameters (such as output flow), and the adjustment step size is set according to the magnitude of the speed difference, thereby guiding the adjustment of the parameters of the target optimization control strategy after application.

[0133] Step S153: Adjust the core parameters in the target optimization control strategy after application according to the new initial parameters adjustment direction and adjustment step size, generate new adjusted parameter values, substitute the new adjusted parameter values ​​into the target optimization control strategy after application, replace the original parameter values, generate a new first-round pre-simulation optimization strategy, and perform multiple rounds of strategy pre-simulation iteration in this way to generate new multi-round pre-simulation optimization strategies.

[0134] Based on the new initial parameter adjustment direction and step size determined in step S152, the core parameters (such as output flow rate and hydraulic valve opening degree) in the applied target optimization control strategy are adjusted to obtain new parameter values. These new parameter values ​​are then substituted into the applied target optimization control strategy, replacing the original parameter values, to generate a new first-round pre-optimization strategy. Then, following a method similar to steps S114 to S118, a new second-round pre-optimization parameter adjustment direction and step size are generated based on the predicted operating state of the new first-round pre-optimization strategy. The parameters are adjusted to generate a new second-round pre-optimization strategy. This process is repeated multiple times to generate new multi-round pre-optimization strategies.

[0135] Step S154: Extract the real-time feedback features of the lifting hydraulic system after application, input the new multi-round pre-simulation optimization strategy and the real-time feedback features of the lifting hydraulic system after application into the dynamic adaptation coupling unit, comprehensively evaluate the adaptability of each round of new pre-simulation optimization strategy, and generate a new strategy adaptation priority sequence.

[0136] Extract the application real-time feedback features generated in step S151, and input the new multi-round pre-simulation optimization strategy and the features together into the dynamic adaptation coupling unit. Following the methods in steps S121 to S129, perform processing such as generating strategy feature vectors and predicted state feature vectors, calculating coupling matching degree, analyzing security adaptation degree and hardware adaptation degree, comprehensive adaptation scoring and ranking for each round of new pre-simulation optimization strategy to generate a new strategy adaptation priority sequence.

[0137] Step S155: Based on the new strategy adaptation priority sequence, extract the new pre-simulation optimization strategies in the top K positions of the sorting, generate a new core pre-simulation optimization strategy set, perform feature fusion, core feature enhancement and parameter convergence processing on the new core pre-simulation optimization strategy set, and generate a new target optimization control strategy.

[0138] Based on the new strategy adaptation priority sequence, the top K ranked pre-exercise optimization strategies are extracted to form a new set of core pre-exercise optimization strategies. Then, following steps S132 to S139, feature fusion, core feature enhancement, and parameter convergence processing are performed on the strategies in this set to generate a new target optimization control strategy.

[0139] Step S156: Convert the new target optimization control strategy into new control commands that can be recognized by each actuator of the lifting hydraulic system, transmit the new control commands to the execution control unit of the lifting hydraulic system, drive each actuator to operate according to the requirements of the new target optimization control strategy, and realize the application of the new target optimization control strategy in the lifting hydraulic system.

[0140] The newly generated target optimization control strategy is converted into control instructions that can be recognized by each execution component, transmitted to the execution control unit, and controls the operation of each execution component, so that the new target optimization control strategy can be applied in the system.

[0141] Step S157: During the application of the new target optimization control strategy, extract the real-time operation feedback features of the lifting hydraulic system and generate an updated real-time feedback feature vector after application. The updated real-time feedback feature vector after application includes the operation features of each component of the lifting hydraulic system under the control of the new target optimization control strategy and the overall operation trend features.

[0142] During the application of the new target optimization control strategy, real-time system operation data is collected, the operation characteristics of each component and the overall operation trend characteristics are extracted, and an updated real-time feedback feature vector after application is generated. This updated real-time feedback feature vector after application reflects the actual operating status of the system under the control of the new strategy.

[0143] Step S158: Use the updated post-application real-time feedback feature vector as the new post-application real-time feedback feature vector, and take the new target optimization control strategy as the new iteration starting point. Repeat the steps of generating new pre-simulation initial parameter adjustment direction and adjustment step size, multi-round strategy pre-simulation iteration, dynamic adaptation coupling evaluation, and strategy refinement to generate a new target optimization control strategy.

[0144] The updated application-after-real-time feedback feature vector generated in step S157 is assigned as the new application-after-real-time feedback feature vector. The new target optimization control strategy is used as the starting point for the next iteration. The operations from step S152 to step S157 are repeated to continuously generate and apply the new target optimization control strategy.

[0145] Step S159: Continuously generate new optimized target control strategies and apply each new target control strategy to the operation control of the lifting hydraulic system. Each applied strategy is optimized based on the real-time operation feedback characteristics after the previous application. Through continuous strategy optimization and application, the control strategy of the lifting hydraulic system can continuously adapt to the changes in the system's operating state, thereby achieving online optimization of the lifting hydraulic control strategy.

[0146] Step S1591: Convert each newly generated target optimization control strategy into control instructions that can be recognized by each actuator of the lifting hydraulic system. The control instructions include the action parameters and action timing information of each actuator, and the format of the control instructions is consistent with the instruction receiving format of each actuator.

[0147] Each time a new target optimization control strategy is generated, it is converted into control commands. These commands specify the action parameters (such as the speed of the hydraulic pump, the opening degree of the hydraulic valve, etc.) and action timing information (when to start, when to stop, action duration, etc.) of each actuator. The format of the commands conforms to the command receiving requirements of each actuator, ensuring that the actuator can correctly parse and execute the commands.

[0148] Step S1592: The control command is transmitted to the execution control unit of the lifting hydraulic system. The execution control unit drives each execution component to operate according to the requirements of the new target optimization control strategy based on the control command.

[0149] Control commands are sent to the execution control unit through a suitable transmission method. After receiving the commands, the execution control unit controls the actions of each execution component according to the requirements of the commands, so that the hydraulic system of the lifting machine can operate under the new target optimization control strategy.

[0150] Step S1593: During the application of the new target optimization control strategy, the operating characteristics of each component of the lifting hydraulic system and the overall operating trend characteristics are extracted in real time to generate real-time operating feedback characteristics after application. The real-time operating feedback characteristics after application are compared with the predicted operating state corresponding to the new target optimization control strategy. The difference between the actual application effect and the predicted effect of the new target optimization control strategy is calculated, and application effect evaluation information is generated.

[0151] During the application of the new strategy, operational data from each component is collected in real time, and operational and overall trend characteristics are extracted to form real-time operational feedback characteristics after application. These characteristics are compared with the predicted operational status characteristics generated during the pre-implementation simulation of the new strategy, and the differences between the two are calculated, such as differences in average lifting speed and energy consumption. Based on these differences, application effect evaluation information is generated to assess the actual application effect of the new strategy.

[0152] Step S1594: Based on the application effect evaluation information, determine the key direction of the next round of strategy optimization. The key direction of the next round of strategy optimization is determined based on the difference values ​​in the application effect evaluation information, and prioritizes optimizing the parameters corresponding to features whose difference values ​​exceed the preset range.

[0153] Analyze the discrepancies in the application effect evaluation information to identify features where the discrepancies exceed preset ranges. The parameters corresponding to these features will be the focus of the next round of strategy optimization. For example, if the energy consumption discrepancies exceed the preset range, then energy-related control parameters (such as the efficiency parameters of the hydraulic pump, the pressure loss parameters of the system, etc.) will become the focus of the next round of optimization.

[0154] Step S1595: Using the new target optimization control strategy as the starting point for the next iteration, and combining the real-time operation feedback characteristics after application with the key directions of the next round of strategy optimization, generate the initial parameter adjustment direction and adjustment step size for the next round of pre-exercise.

[0155] The new target optimization control strategy is used as the starting point for the next iteration. Taking into account the real-time operation feedback characteristics after application and the key optimization directions determined in step S1594, the initial adjustment direction and adjustment step size of each control parameter are generated for the next round of pre-run. The parameters involved in the key optimization directions are adjusted first.

[0156] Step S1596: Adjust the direction and adjustment step size according to the initial parameters of the next round of pre-simulation, execute the multi-round strategy pre-simulation iteration process, and generate the multi-round pre-simulation optimization strategy for the next round. Each round of pre-simulation optimization strategy corresponds to a set of predicted operating states of the lifting hydraulic system.

[0157] Based on the initial parameters of the next round of pre-simulation, adjust the direction and adjustment step size, and execute a multi-round strategy pre-simulation iteration process similar to steps S113 to S1110 to generate multiple pre-simulation optimization strategies and corresponding predicted running states for the next round.

[0158] Step S1597: Input the multi-round pre-simulation optimization strategy for the next round and the real-time operation feedback characteristics of the lifting hydraulic system after application into the dynamic adaptation coupling unit to generate the strategy adaptation priority sequence for the next round.

[0159] The next round of multi-round pre-simulation optimization strategy and the real-time running feedback features after application are input into the dynamic adaptation coupling unit and processed according to the methods of steps S121 to S129 to generate the next round of strategy adaptation priority sequence.

[0160] Step S1598: Based on the next round of policy adaptation priority sequence, execute the policy refinement process to generate the next round of target optimization control policy. The next round of target optimization control policy is generated based on the real-time operation feedback characteristics after application and the next round of policy adaptation priority sequence.

[0161] Based on the next round of strategy adaptation priority sequence, select the top K strategies and execute the strategy refinement process from steps S132 to S139 to generate the next round of target optimization control strategy. This target optimization control strategy is obtained by optimizing based on the real-time operation feedback characteristics after application and the new strategy adaptation priority sequence.

[0162] Step S1599: Repeat the steps from switching control commands to generating the next round of target optimization control strategy, continuously generate optimized target optimization control strategies and apply them to the lifting hydraulic system. Each applied strategy is optimized based on the real-time operation feedback characteristics after the previous application.

[0163] By continuously repeating steps S1591 to S1598, new target optimization control strategies are continuously generated and applied to the system. Each strategy optimization is based on the feedback from the previous application, enabling the control strategy to continuously adapt to system changes and achieve online optimization.

[0164] Throughout the process, data collection is involved in the operation of the lift's hydraulic system, which may include sensitive information such as specific operating parameters and maintenance records. To protect the privacy and security of this data, data encryption technology is used to encrypt the collected data, and secure communication protocols, such as encrypted transmission protocols, are used during data transmission to prevent data theft or tampering. Simultaneously, access control is implemented for the database storing the data, ensuring that only authorized personnel can access and process sensitive data, thus guaranteeing data privacy, security, and prevention of leakage.

[0165] Figure 2 The diagram illustrates the hardware structure of an online optimization system 100 for lifting hydraulic control strategies based on AI model predictive control, provided in an embodiment of the present invention, for implementing the above-described online optimization method for lifting hydraulic control strategies based on AI model predictive control. Figure 2 As shown, the online optimization system 100 for hydraulic control strategy of a lift based on AI model predictive control may include a processor 110, a machine-readable storage medium 120, a bus 130, and a communication unit 140.

[0166] Machine-readable storage medium 120 can store data and / or instructions. In some embodiments, machine-readable storage medium 120 can store data acquired from an external terminal. In some embodiments, machine-readable storage medium 120 can store data and / or instructions used by the AI ​​model predictive control-based online optimization system 100 for lifting hydraulic control strategies to execute or use in order to complete the exemplary methods described in this invention. In a specific implementation, one or more processors 110 execute the computer-executable instructions stored in machine-readable storage medium 120, enabling processor 110 to execute the AI ​​model predictive control-based online optimization method for lifting hydraulic control strategies as described in the above method embodiments. Processor 110, machine-readable storage medium 120, and communication unit 140 are connected via bus 130, and processor 110 can be used to control the transmission and reception actions of communication unit 140. The specific implementation process of processor 110 can be found in the various method embodiments executed by the AI ​​model predictive control-based online optimization system 100 for lifting hydraulic control strategies, and their implementation principles and technical effects are similar, so they will not be repeated here.

[0167] Furthermore, this embodiment of the invention also provides a readable storage medium containing computer-executable instructions. When the processor executes the computer-executable instructions, the above-mentioned online optimization method for the hydraulic control strategy of a lift based on AI model predictive control is implemented.

[0168] It should be noted that, in order to simplify the description of this invention and thus aid in the understanding of one or more embodiments, the foregoing description of the embodiments of this invention sometimes combines multiple features into a single embodiment, drawing, or description thereof. Similarly, it should be noted that, in order to simplify the description of this invention and thus aid in the understanding of one or more embodiments, the foregoing description of the embodiments of this invention sometimes combines multiple features into a single embodiment, drawing, or description thereof.

Claims

1. An online optimization method for hydraulic control strategy of a lifting machine based on AI model predictive control, characterized in that, The method includes: Taking the current control strategy of the lift hydraulic system as the starting point of the iteration, the AI ​​prediction model is triggered to perform a multi-round strategy pre-exercise iteration process. The AI ​​prediction model is used to adjust the parameters of the current control strategy and perform operation state pre-exercise in multiple rounds to generate a multi-round pre-exercise optimization strategy. Each round of pre-exercise optimization strategy corresponds to a set of predicted operation states of the lift hydraulic system. The multi-round pre-exercise optimization strategy includes a set of control strategies with different parameter configurations. Extract the real-time operation feedback characteristics of the lifting machine hydraulic system, and input the multi-round pre-simulation optimization strategy and the real-time operation feedback characteristics of the lifting machine hydraulic system into the dynamic adaptation coupling unit. The dynamic adaptation coupling unit comprehensively evaluates the adaptability of each round of pre-simulation optimization strategy and generates a strategy adaptation priority sequence. Based on the policy adaptation priority sequence, the policy refinement process of the AI ​​prediction model is triggered. The top K pre-optimization policies in the policy adaptation priority sequence are subjected to feature fusion, core feature enhancement and parameter convergence processing to generate the target optimization control policy. The target optimization control strategy is converted into control commands that can be recognized by each actuator of the lifting hydraulic system. The control commands are then transmitted to the execution control unit of the lifting hydraulic system to drive each actuator to operate according to the requirements of the target optimization control strategy. This realizes the application of the target optimization control strategy in the lifting hydraulic system and extracts the real-time operation feedback features after the application. Taking the target-optimized control strategy after application as the new iteration starting point, and combining the real-time operation feedback characteristics after application, all steps of the multi-round strategy pre-exercise iteration process, dynamic adaptation coupling evaluation, and strategy refinement process are repeatedly executed to continuously generate optimized target-optimized control strategies and apply them to the lift hydraulic system, thereby realizing online optimization of the lift hydraulic control strategy.

2. The online optimization method for hydraulic control strategy of a lifting machine based on AI model predictive control according to claim 1, characterized in that, The process begins with the current control strategy of the lift's hydraulic system as the starting point, triggering a multi-round strategy pre-simulation iteration process for the AI ​​prediction model. The AI ​​prediction model performs multiple rounds of parameter adjustments and operational state simulations of the current control strategy, generating a multi-round pre-simulation optimization strategy, including: Extract all control parameters from the current control strategy of the lifting hydraulic system, record the complete identifier of each control parameter, which includes the parameter name, the component to which it belongs, and its position in the control strategy. Perform parameter sensitivity analysis and parameter correlation analysis on each control parameter, calculate the influence weight of each control parameter on the operating state of the lifting hydraulic system, sort all control parameters in descending order of influence weight, and generate a parameter influence weight sequence. Taking the current control strategy of the lifting hydraulic system as the starting point of the iteration, the parameter influence weight sequence is input into the pre-simulation initialization unit of the AI ​​prediction model to generate the initial parameter adjustment direction and adjustment step size for the first round of pre-simulation. The initial parameter adjustment direction and adjustment step size are determined based on the weight information in the parameter influence weight sequence and are used to guide the parameter adjustment of the current control strategy. According to the initial parameter adjustment direction and adjustment step size of the first round of pre-rehearsal, the core parameters in the current control strategy of the lifting hydraulic system are adjusted to generate the adjusted parameter values. The adjusted parameter values ​​are then substituted into the current control strategy to replace the original parameter values, generating the first round of pre-rehearsal optimization strategy. The first round of pre-rehearsal optimization strategy includes all the adjusted control parameters and parameter association logic. The first round of pre-simulation optimization strategy is input into the operation state pre-simulation unit of the AI ​​prediction model to simulate the complete operation process of the lift hydraulic system under the control of the first round of pre-simulation optimization strategy, and generate a set of predicted operation states corresponding to the first round of pre-simulation optimization strategy. The predicted operation states include the operation characteristics of each component of the lift hydraulic system and the overall operation trend. Key feature information is extracted from the predicted running state corresponding to the first round of pre-exercise optimization strategy. The key feature information is fed back to the pre-exercise iteration unit of the AI ​​prediction model to generate the parameter adjustment direction and adjustment step size for the second round of pre-exercise. The parameter adjustment direction and adjustment step size for the second round of pre-exercise are generated based on the feature optimization of the first round of predicted running state and are used to guide the parameter adjustment of the first round of pre-exercise optimization strategy. According to the parameter adjustment direction and adjustment step size of the second round of pre-exercise, the core parameters in the optimization strategy of the first round of pre-exercise are adjusted again to generate the adjusted parameter values. The adjusted parameter values ​​are then substituted into the optimization strategy of the first round of pre-exercise to replace the original parameter values, generating the optimization strategy of the second round of pre-exercise. The optimization strategy of the second round of pre-exercise includes all the adjusted control parameters and parameter association logic. The second round of pre-simulation optimization strategy is input into the operation state pre-simulation unit of the AI ​​prediction model to simulate the complete operation process of the lift hydraulic system under the control of the second round of pre-simulation optimization strategy, and generate a set of predicted operation states corresponding to the second round of pre-simulation optimization strategy. Based on the predicted operating state characteristics corresponding to the second round of pre-exercise optimization strategy, the parameter adjustment direction and adjustment step size for the third round of pre-exercise are generated. The parameters of the second round of pre-exercise optimization strategy are adjusted, the third round of pre-exercise optimization strategy is generated, and the corresponding operating state is pre-exercised. This process is repeated for multiple rounds. The parameter adjustment direction and adjustment step size for each round are generated based on the predicted operating state characteristics of the previous round. Record the complete parameter configuration and corresponding predicted running status of the pre-exercise optimization strategy for each round, forming a pre-exercise iteration record set. The pre-exercise iteration record set contains the correlation information between the pre-exercise optimization strategy and the predicted running status for all rounds. Integrate all pre-exercise optimization strategies in the pre-exercise iteration record set, remove duplicate strategy configurations, and generate multi-round pre-exercise optimization strategies. Each round of pre-exercise optimization strategy corresponds to a set of predicted operating states of the lifting hydraulic system.

3. The online optimization method for hydraulic control strategy of a lifting machine based on AI model predictive control according to claim 1, characterized in that, The real-time operational feedback characteristics of the lift hydraulic system are extracted, and the multi-round pre-simulation optimization strategy and the real-time operational feedback characteristics of the lift hydraulic system are input into a dynamic adaptation coupling unit. The dynamic adaptation coupling unit comprehensively evaluates the adaptability of each round of pre-simulation optimization strategy and generates a strategy adaptation priority sequence, including: Extract the complete parameter configuration of each round of pre-exercise optimization strategy in the multi-round pre-exercise optimization strategy, generate the strategy feature vector of each round of pre-exercise optimization strategy. The strategy feature vector contains the configuration information of all control parameters of the corresponding pre-exercise optimization strategy and the parameter association logic features. Extract the key feature information in the predicted running state corresponding to each round of pre-exercise optimization strategy, and generate the predicted state feature vector of each round of pre-exercise optimization strategy. The strategy feature vector of each round of pre-exercise optimization strategy is associated and bound with the corresponding predicted state feature vector to generate the strategy state association vector of each round of pre-exercise optimization strategy. The strategy state association vector contains complete association information between the strategy features and the predicted state features. Extract the real-time operation feedback features of the lifting hydraulic system and generate a real-time feedback feature vector. The real-time feedback feature vector contains the operation features of each component and the overall operation trend features of the lifting hydraulic system during the current operation process, which serve as the benchmark features for dynamic adaptation and coupling evaluation. The strategy state association vector of each round of pre-exercise optimization strategy in the multi-round pre-exercise optimization strategy is coupled with the real-time feedback feature vector. The degree of fit between the strategy features, predicted state features in the strategy state association vector of each round and the real-time feedback features in the real-time feedback feature vector is analyzed to generate the coupling matching degree of each round of pre-exercise optimization strategy. Extract the predicted operating state of each round of pre-simulation optimization strategy, analyze the degree of fit between the predicted operating state of each round of pre-simulation optimization strategy and the safe operating boundary of the lifting hydraulic system, generate the safety adaptability of each round of pre-simulation optimization strategy, and use the safety adaptability to evaluate the adaptability of each round of pre-simulation optimization strategy at the safety level. Extract the parameter configuration of each round of pre-exercise optimization strategy, analyze the degree of fit between the parameter configuration of each round of pre-exercise optimization strategy and the hardware performance boundary of the lifting hydraulic system, generate the hardware adaptability of each round of pre-exercise optimization strategy, and use the hardware adaptability to evaluate the adaptability of each round of pre-exercise optimization strategy at the hardware level. Based on the set weight ratio, the coupling matching degree, security adaptability and hardware adaptability of each round of pre-exercise optimization strategy are weighted and calculated to generate a comprehensive adaptability score for each round of pre-exercise optimization strategy. The pre-exercise optimization strategies of multiple rounds are sorted in descending order of comprehensive adaptability score to generate strategy adaptability ranking information. The strategy adaptability ranking information includes the identifier of each round of pre-exercise optimization strategy, the corresponding comprehensive adaptability score and ranking position. The strategy adaptation ranking information is associated and bound with the complete information of the multi-round pre-exercise optimization strategy. The associated and bound content includes the identifier, comprehensive adaptation score, ranking position, strategy feature vector, predicted state feature vector and strategy state association vector of each round of pre-exercise optimization strategy. Based on the information after association and binding, a strategy adaptation priority sequence is generated. The strategy adaptation priority sequence includes the adaptation ranking information of all pre-optimization strategies and the corresponding strategy features and prediction state feature information.

4. The online optimization method for hydraulic control strategy of a lifting machine based on AI model predictive control according to claim 1, characterized in that, The policy refinement process of the AI ​​prediction model, triggered by the policy adaptation priority sequence, involves feature fusion, core feature enhancement, and parameter convergence processing on the top K pre-simulated optimization policies in the policy adaptation priority sequence to generate a target optimization control policy, including: The strategy adaptation priority sequence is parsed to extract the strategy adaptation ranking information, the top K ranked pre-optimization strategies are extracted, and a core pre-optimization strategy set is generated. The core pre-optimization strategy set contains complete information of all the top K ranked pre-optimization strategies, including strategy feature vectors, predicted state feature vectors, and strategy state association vectors. Extract the policy feature vector of each pre-optimization strategy in the core pre-optimization strategy set, input all policy feature vectors into the feature fusion unit of the AI ​​prediction model, perform cross-fusion processing of feature vectors, fuse the policy features of all core pre-optimization strategies, and generate a fused policy feature vector. The fused policy feature vector contains the feature information of all core pre-optimization strategies. Extract the predicted state feature vector corresponding to each pre-exercise optimization strategy in the core pre-exercise optimization strategy set, input all predicted state feature vectors into the feature fusion unit of the AI ​​prediction model, perform cross-fusion processing of feature vectors, fuse the predicted state features corresponding to all core pre-exercise optimization strategies, and generate a fused predicted state feature vector. The fusion strategy feature vector is associated and bound with the fusion prediction state feature vector to generate a fusion strategy state association vector. The fusion strategy feature vector is extracted from the fusion strategy state association vector. The control parameter features in the fusion strategy feature vector are identified. The parameter influence weight sequence is retrieved. An influence weight threshold is set. The control parameter features whose influence weight values ​​are greater than the influence weight threshold are identified as core control parameter features. The core control parameter features in the fusion strategy features are strengthened to increase the weight ratio of the core control parameter features in the fusion strategy features, and a strengthened strategy feature vector is generated. Extract the fusion prediction state feature vector from the fusion strategy state association vector, identify the state features related to the core control parameter features in the fusion prediction state feature vector, strengthen the state features related to the core control parameter features, increase the weight ratio of the relevant state features in the fusion prediction state features, and generate the strengthened prediction state feature vector. The reinforcement strategy feature vector is associated and bound with the reinforcement prediction state feature vector to generate the reinforcement strategy state association vector. The reinforcement strategy state association vector is input into the parameter convergence unit of the AI ​​prediction model to perform convergence processing on the control parameters in the fused strategy features, eliminate redundant parameter configurations, optimize the association logic between parameters, and generate the converged strategy state association vector. Extract the convergence strategy feature vector from the convergence strategy state association vector. The convergence strategy feature vector contains all control parameter configuration information and parameter association logic features after convergence. Input the convergence strategy feature vector into the running state pre-simulation unit of the AI ​​prediction model to simulate the complete running process of the lift hydraulic system under the corresponding strategy control and generate the convergence strategy verification running state. Extract key feature information from the convergence strategy verification operation status, compare the key feature information with the real-time operation feedback features of the lifting hydraulic system, and perform final parameter fine-tuning on the convergence strategy feature vector based on the comparison results to optimize the parameter configuration and correlation logic of the convergence strategy. Based on the convergence strategy feature vector after parameter fine-tuning, a target optimization control strategy is generated. The target optimization control strategy includes all optimized control parameter configuration information and parameter association logic features.

5. The online optimization method for hydraulic control strategy of a lifting machine based on AI model predictive control according to claim 1, characterized in that, The process begins with the applied target-optimized control strategy as a new iterative starting point. Combining this with real-time operational feedback characteristics, it repeatedly executes all steps of the multi-round strategy pre-simulation iteration process, dynamic adaptation coupling evaluation, and strategy refinement process. This continuously generates optimized target-optimized control strategies and applies them to the lift hydraulic system, achieving online optimization of the lift hydraulic control strategy. This includes: Using the target optimization control strategy after application as a new iteration starting point, extract the real-time operation feedback features after application, and generate the real-time feedback feature vector after application. The real-time feedback feature vector after application contains the operation features of each component of the lifting hydraulic system under the control of the target optimization control strategy and the overall operation trend features. The real-time feedback feature vector after application is input into the pre-simulation initialization unit of the AI ​​prediction model to generate a new pre-simulation initial parameter adjustment direction and adjustment step size. The new pre-simulation initial parameter adjustment direction and adjustment step size are optimized and generated based on the real-time feedback feature vector after application, and are used to guide the parameter adjustment of the target optimization control strategy after application. According to the new initial parameters of the pre-simulation, the core parameters in the target optimization control strategy after application are adjusted according to the adjustment direction and adjustment step size, and new adjusted parameter values ​​are generated. The new adjusted parameter values ​​are substituted into the target optimization control strategy after application to replace the original parameter values ​​and generate a new first-round pre-simulation optimization strategy. In this way, multiple rounds of strategy pre-simulation iteration are carried out to generate new multi-round pre-simulation optimization strategies. Extract the real-time feedback features of the lifting hydraulic system after application, and input the new multi-round pre-simulation optimization strategy and the real-time feedback features of the lifting hydraulic system after application into the dynamic adaptation coupling unit. The adaptability of each round of new pre-simulation optimization strategy is comprehensively evaluated to generate a new strategy adaptation priority sequence. Based on the new strategy adaptation priority sequence, new pre-simulation optimization strategies are extracted from the top K positions of the sorting, generating a new core pre-simulation optimization strategy set. Feature fusion, core feature enhancement, and parameter convergence processing are then performed on the new core pre-simulation optimization strategy set to generate a new target optimization control strategy. The new target optimization control strategy is converted into new control commands that can be recognized by each actuator of the lifting hydraulic system. The new control commands are then transmitted to the execution control unit of the lifting hydraulic system, driving each actuator to operate according to the requirements of the new target optimization control strategy, thereby realizing the application of the new target optimization control strategy in the lifting hydraulic system. During the application of the new target optimization control strategy, the real-time operation feedback features of the lifting hydraulic system are extracted, and an updated real-time feedback feature vector after application is generated. The updated real-time feedback feature vector after application includes the operation features of each component of the lifting hydraulic system under the control of the new target optimization control strategy and the overall operation trend features. The updated post-application real-time feedback feature vector is used as the new post-application real-time feedback feature vector. The new target optimization control strategy is used as the new iteration starting point. The steps of generating new pre-simulation initial parameter adjustment direction and adjustment step size, multi-round strategy pre-simulation iteration, dynamic adaptation coupling evaluation, and strategy refinement to generate a new target optimization control strategy are repeatedly executed. The system continuously generates new optimized target control strategies and applies each new strategy to the operation control of the lift hydraulic system. Each applied strategy is optimized based on the real-time operation feedback characteristics after the previous application. Through continuous strategy optimization and application, the control strategy of the lift hydraulic system adapts to the changes in the system's operating state, thus achieving online optimization of the lift hydraulic control strategy.

6. The online optimization method for hydraulic control strategy of a lifting machine based on AI model predictive control according to claim 2, characterized in that, The process involves extracting all control parameters from the current control strategy of the lifting platform's hydraulic system, recording the complete identifier of each control parameter (including its name, the component it belongs to, and its location within the control strategy), performing sensitivity and correlation analysis on each parameter, calculating the influence weight of each parameter on the operating state of the lifting platform's hydraulic system, and sorting all control parameters in descending order of influence weight to generate a parameter influence weight sequence, including: Iterate through all the configurations of the current control strategy of the lifting hydraulic system, identify and extract each control parameter in the current control strategy, and record the complete identifier of each control parameter. The complete identifier of the control parameter includes the parameter name, the component to which it belongs, and the location information in the control strategy. Check all the extracted control parameters one by one to confirm that the complete identifier information of each control parameter has been recorded and there are no missing or omitted parameter information. Perform individual parameter sensitivity analysis on each control parameter to simulate the impact of the value of the control parameter changing by a preset range on the operating state of the lifting hydraulic system while other parameters remain unchanged. Generate a sensitivity coefficient for each control parameter, which is based on the magnitude of the change in system operating state caused by the change in parameter value. For each control parameter, perform parameter correlation analysis to identify the correlation between the control parameter and other control parameters, count the number of correlations between the control parameter and other control parameters, analyze the correlation strength between the control parameter and other control parameters, and generate the correlation coefficient for each control parameter. The correlation coefficient is generated based on the number and strength of the correlations between parameters. The sensitivity coefficient and correlation coefficient are normalized to make them comparable within the same dimension range. After normalization, a weighted summation method is used to generate the influence weight of each control parameter on the operating state of the lifting hydraulic system. The influence weight is a comprehensive quantitative result of the sensitivity coefficient and correlation coefficient, and the magnitude of the influence weight represents the degree of influence of the parameter on the operating state of the system. All control parameters are sorted in descending order of their influence weight values ​​to generate an initial parameter sorting sequence. The initial parameter sorting sequence contains the complete identifiers of all control parameters and their corresponding influence weight information. The control parameters in the initial parameter sorting sequence are classified and organized. The control parameters are divided into different parameter categories according to the component type to which the parameters belong, and a classified parameter sorting sequence is generated. The classified parameter sorting sequence contains the sorting information of control parameters under different categories. Extract the overall impact weight of each parameter category on the operating status of the lifting hydraulic system, and average the impact weight values ​​of all parameters under that parameter category to generate the category impact weight. The category impact weight is used to evaluate the overall impact of parameters of different component types on the operating status of the system. The parameter categories in the classification parameter sorting sequence are sorted in descending order of their category influence weight values ​​to generate a category sorting sequence. The category sorting sequence contains the identifiers of all parameter categories and their corresponding category influence weight information, which is used to determine the adjustment priority of component types. The category sorting sequence and the classification parameter sorting sequence are associated and integrated to generate the final parameter influence weight sequence. The parameter influence weight sequence includes the complete identifier of each control parameter, the corresponding influence weight information, the parameter category to which it belongs, and the category influence weight information.

7. The online optimization method for hydraulic control strategy of a lifting machine based on AI model predictive control according to claim 2, characterized in that, Based on the predicted operating state characteristics corresponding to the second round of pre-simulation optimization strategy, the parameter adjustment direction and adjustment step size for the third round of pre-simulation are generated. The parameters of the second round of pre-simulation optimization strategy are adjusted, and the third round of pre-simulation optimization strategy is generated and its corresponding operating state is pre-simulated. This process is repeated multiple times, with the parameter adjustment direction and adjustment step size for each round being optimized and generated based on the predicted operating state characteristics of the previous round. This includes: Set the termination condition for the strategy pre-exercise iteration. The termination condition is that the degree of fit between the predicted operating state corresponding to the pre-exercise optimization strategy and the ideal operating state of the lifting hydraulic system reaches a preset standard value. Record the current round of pre-exercise iteration. The initial round is the second round. The current pre-exercise optimization strategy is the second round pre-exercise optimization strategy. Extract key feature information from the predicted running state corresponding to the current pre-exercise optimization strategy, and generate the predicted state feature vector for the current round. The predicted state feature vector for the current round contains all the key feature information of the predicted running state for the current round. The predicted state feature vector of the current round is compared with the ideal operating state feature vector of the lifting hydraulic system. The fit value between the predicted state feature and the ideal operating state feature of the current round is calculated, and the fit score of the current round is generated. The fit score is generated based on the fit value between the predicted state feature and the ideal state feature. If the fit score of the current round does not reach the preset standard value in the termination condition, extract the feature dimensions in the current round's predicted state feature vector that differ from the ideal running state feature vector by more than a preset difference range, and generate a list of difference feature dimensions. The list of difference feature dimensions contains the identification information of all feature dimensions whose differences exceed the preset difference range. Based on the list of differential feature dimensions, control parameters related to the differential feature dimensions are identified, and a list of parameters to be adjusted is generated. The list of parameters to be adjusted contains complete identification information of all control parameters related to the differential feature dimensions. Extract the influence weight information of each parameter in the list of parameters to be adjusted from the parameter influence weight sequence, determine the adjustment priority of the parameters to be adjusted in descending order of influence weight value, and generate a parameter adjustment priority sequence; Based on the parameter adjustment priority sequence and the predicted state feature vector of the current round, the parameter adjustment direction and adjustment step size for the next round of pre-training are generated. The parameter adjustment direction and adjustment step size for the next round are determined based on the optimization requirements of the difference feature dimension, and the corresponding parameters are adjusted in order of adjustment priority from high to low. According to the parameter adjustment direction and adjustment step size of the next round of pre-exercise, adjust the parameters to be adjusted in the current round of pre-exercise optimization strategy, generate the adjusted parameter values ​​for the next round, substitute the adjusted parameter values ​​for the next round into the current round of pre-exercise optimization strategy, replace the original parameter values, and generate the next round of pre-exercise optimization strategy. The next round of pre-simulation optimization strategy is input into the operation state pre-simulation unit of the AI ​​prediction model to simulate the complete operation process of the lift hydraulic system under the control of the corresponding pre-simulation optimization strategy, and generate a set of predicted operation states corresponding to the next round of pre-simulation optimization strategy; Increment the current round of the pre-exercise by 1 to update it to the new current round. Use the optimization strategy for the next round of pre-exercise as the optimization strategy for the current round. Repeat the steps from extracting the feature vector of the predicted state to generating the predicted running state of the next round until the fit score of the current round reaches the preset standard value in the termination condition.

8. The online optimization method for hydraulic control strategy of a lifting machine based on AI model predictive control according to claim 4, characterized in that, The process involves extracting the fusion strategy feature vector from the fusion strategy state association vector, identifying control parameter features within the fusion strategy feature vector, retrieving the parameter influence weight sequence, setting an influence weight threshold, identifying control parameter features with influence weight values ​​greater than the influence weight threshold as core control parameter features, strengthening the core control parameter features in the fusion strategy features to increase their weight proportion, and generating a strengthened strategy feature vector, including: Extract the fusion strategy feature vector from the fusion strategy state association vector, identify all control parameter features in the fusion strategy feature vector, and record the identifier and corresponding feature weight of each control parameter feature. The feature weight represents the importance of the corresponding control parameter feature in the fusion strategy features. Retrieve the parameter influence weight sequence, extract the influence weight information corresponding to each control parameter feature in the fusion strategy feature vector from the parameter influence weight sequence, set the influence weight threshold, and determine the control parameter features whose influence weight values ​​are greater than the influence weight threshold as core control parameter features; Obtain the enhancement weight coefficient of the core control parameter feature, multiply the feature weight of each core control parameter feature by the corresponding enhancement weight coefficient to generate the enhanced feature weight. The enhanced feature weight represents the new importance of the core control parameter feature in the fusion strategy feature. Keeping the feature weights of non-core control parameter features in the fusion strategy feature vector unchanged, the feature weights of the enhanced core control parameter features and the feature weights of the non-core control parameter features are integrated to generate an enhanced strategy feature vector, which contains all enhanced control parameter features and their corresponding feature weights. The format of the reinforcement strategy feature vector is standardized to unify the expression of each element in the reinforcement strategy feature vector. The core control parameter features in the reinforcement strategy feature vector are extracted. The change in the weight ratio of the core control parameter features before and after reinforcement is calculated to verify the reinforcement effect of the core control parameter features and confirm that the change in the weight ratio of the core control parameter features in the fusion strategy features reaches the preset change range. If the enhancement effect of the core control parameter features does not reach the preset change range, adjust the enhancement weight coefficient of the core control parameter features, recalculate the feature weight of the core control parameter features, and generate new enhanced feature weights. The feature weights of the new enhanced core control parameter features are integrated with the feature weights of the non-core control parameter features to generate a new enhanced strategy feature vector. The final enhanced strategy feature vector is taken as the result of feature enhancement processing. The enhanced strategy feature vector contains all enhanced control parameter features and their corresponding feature weights.

9. The online optimization method for hydraulic control strategy of a lifting machine based on AI model predictive control according to claim 5, characterized in that, The continuously generated and optimized new target optimization control strategies are sequentially applied to the operation control of the lift hydraulic system. Each applied strategy is optimized based on the real-time operation feedback characteristics after the previous application. Through continuous strategy optimization and application, the control strategy of the lift hydraulic system continuously adapts to changes in the system's operating state, achieving online optimization of the lift hydraulic control strategy, including: Each newly generated target optimization control strategy is converted into control instructions that can be recognized by each actuator of the lifting hydraulic system. The control instructions include the action parameters and action timing information of each actuator, and the format of the control instructions is consistent with the instruction receiving format of each actuator. The control commands are transmitted to the execution control unit of the lift hydraulic system. The execution control unit drives each execution component to operate according to the requirements of the new target optimized control strategy based on the control commands. During the application of the new target optimization control strategy, the operating characteristics of each component and the overall operating trend characteristics of the hydraulic system of the lifting machine are extracted in real time, and the real-time operating feedback characteristics after application are generated. The real-time operating feedback characteristics after application are compared with the predicted operating state corresponding to the new target optimization control strategy, and the difference between the actual application effect and the predicted effect of the new target optimization control strategy is calculated to generate application effect evaluation information. Based on the application effect evaluation information, the key directions for the next round of strategy optimization are determined. The key directions for the next round of strategy optimization are determined based on the difference values ​​in the application effect evaluation information, and the parameters corresponding to the features whose difference values ​​exceed the preset range are optimized first. Using the new target optimization control strategy as the starting point for the next iteration, and combining the real-time operation feedback characteristics after application with the key directions of the next round of strategy optimization, the initial parameter adjustment direction and adjustment step size for the next round of simulation are generated. Adjust the direction and step size according to the initial parameters of the next round of pre-simulation, execute the multi-round strategy pre-simulation iteration process, generate the multi-round pre-simulation optimization strategy for the next round, and each round of pre-simulation optimization strategy corresponds to a set of predicted operating states of the lifting hydraulic system; The next round of multi-round pre-simulation optimization strategy and the real-time operation feedback characteristics of the lifting hydraulic system after application are input together into the dynamic adaptation coupling unit to generate the next round of strategy adaptation priority sequence. Based on the next round of strategy adaptation priority sequence, execute the strategy refinement process to generate the next round of target optimization control strategy. The next round of target optimization control strategy is generated based on the real-time running feedback characteristics after application and the next round of strategy adaptation priority sequence. The process of repeatedly executing the steps from switching control commands to generating the next round of target optimization control strategies is continuously performed, generating optimized target optimization control strategies and applying them to the lifting machine hydraulic system. Each applied strategy is optimized based on the real-time operational feedback characteristics after the previous application.

10. An online optimization system for hydraulic control strategy of a lifting machine based on AI model predictive control, characterized in that, The online optimization system for the hydraulic control strategy of a lift based on AI model predictive control includes a processor and a memory. The memory and the processor are connected. The memory is used to store programs, instructions, or code. The processor is used to run the programs, instructions, or code in the memory to implement the online optimization method for the hydraulic control strategy of a lift based on AI model predictive control as described in any one of claims 1-9.

Images