Energy-saving control system of automobile generator

Abstract

The invention discloses an energy-saving control system of an automobile generator, which relates to the technical field of intelligent control, and comprises the following steps: defining a state space of the automobile generator, and defining an action space of the automobile generator based on an adjustment identification result; introducing a reward function, carrying out reinforcement learning training, constructing an operation control model, carrying out interactive feedback with the automobile generator, generating a control feedback parameter set, carrying out operation dynamic prediction on the automobile generator based on the control feedback parameter set, and obtaining an operation prediction dynamic trend; respectively mapping the operation dynamic trend to a state space and an action space, and formulating an operation constraint condition; and performing operation control on the automobile generator according to the operation constraint condition, determining an operation control strategy, and performing automatic energy-saving control on the automobile generator. The technical problems that in the prior art, an energy-saving control strategy is not intelligent enough, energy consumption is high, and operation efficiency is low are solved, and the technical effects of improving the operation efficiency of the generator and reducing energy consumption are achieved.

Description

Technical Field

[0001] The invention relates to the technical field of intelligent control, and in particular to an energy-saving control system of an automobile generator. Background Art

[0002] Existing automotive generator energy-saving control technologies are usually based on fixed control strategies and lack intelligent dynamic adjustment capabilities. In actual operation, the generator needs to flexibly adjust the power output and charging mode according to the real-time operating status of the vehicle, such as battery power, power demand, vehicle acceleration or deceleration, etc. However, traditional technologies often lead to problems such as excessive energy consumption, low efficiency, and waste of resources due to the failure to fully utilize operating data for real-time optimization. Summary of the invention

[0003] The present application provides an energy-saving control system for an automobile generator, which is used to solve the technical problems of low efficiency and low accuracy in the prior art when classifying and processing construction solid waste.

[0004] In view of the above problems, the present application provides an energy-saving control system for an automobile generator.

[0005] The present application provides an energy-saving control system for an automobile generator, the system comprising: An adjustment identification module, wherein the adjustment identification module defines the state space of the automobile generator according to an operation data set of the automobile generator, performs adjustment identification according to the operation data set, and defines the action space of the automobile generator based on the adjustment identification result; an operation control model construction module, wherein the operation control model construction module introduces a reward function, performs reinforcement learning training in combination with the state space and the action space, and constructs an operation control model of the automobile generator; an operation dynamic prediction module, wherein the operation dynamic prediction module interacts with the automobile generator through the operation control model to generate a control feedback parameter set, performs operation dynamic prediction on the automobile generator based on the control feedback parameter set, and obtains an operation prediction dynamic trend; an operation constraint condition formulation module, wherein the operation constraint condition formulation module maps the operation dynamic trend to the state space and the action space respectively, and formulates operation constraints; an energy-saving control module, wherein the energy-saving control module performs operation control on the automobile generator according to the operation constraints, determines an operation control strategy, and executes the operation control strategy to perform automatic energy-saving control on the automobile generator.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: The present application defines the state space of the automobile generator according to the operation data set of the automobile generator, adjusts the mark according to the operation data set, and defines the action space of the automobile generator based on the adjustment mark result; introduces a reward function, combines the state space and the action space for reinforcement learning training, and constructs an operation control model of the automobile generator; through the operation control model and the automobile generator for interactive feedback, a control feedback parameter set is generated, and the operation dynamic prediction of the automobile generator is performed based on the control feedback parameter set to obtain the operation prediction dynamic trend; the operation dynamic trend is mapped to the state space and the action space respectively, and the operation constraint conditions are formulated; the automobile generator is operated according to the operation constraint conditions, the operation control strategy is determined, and the operation control strategy is executed to automatically control the automobile generator for energy saving. The present invention solves the technical problems that the energy-saving control strategy in the prior art is not intelligent enough, the energy consumption is high, and the operation efficiency is low. The state space and the action space are constructed by the operation data set, and the reward function is introduced for reinforcement learning training, the operation control model is generated, the model interactive feedback and dynamic prediction are combined, the operation constraint conditions are formulated, the operation control strategy is optimized based on the constraint conditions, and the automatic energy-saving control is executed, so as to achieve the technical effect of improving the operation efficiency of the generator and reducing energy consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0008] Figure 1 A schematic diagram of the structure of an energy-saving control system for an automobile generator provided in an embodiment of the present application.

[0009] Figure 2 A schematic flow chart of an energy-saving control method for an automobile generator provided in an embodiment of the present application.

[0010] Explanation of the reference numerals: adjustment identification module 11 , operation control model building module 12 , operation dynamic prediction module 13 , operation constraint condition formulation module 14 , energy-saving control module 15 . DETAILED DESCRIPTION

[0011] The present application provides an energy-saving control system for an automobile generator, aiming to solve the technical problems in the prior art that the energy-saving control strategy is not intelligent enough, the energy consumption is high and the operating efficiency is low. The application constructs the state space and action space through the operation data set, introduces the reward function for reinforcement learning training, generates the operation control model, combines the model's interactive feedback and dynamic prediction, formulates the operation constraints, optimizes the operation control strategy based on the constraints, and performs automated energy-saving control, so as to achieve the technical effect of improving the operating efficiency of the generator and reducing energy consumption.

[0012] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0013] It should be noted that any variations of the terms "include" and "have" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0014] Examples, such as Figure 1 As shown, the embodiment of the present application provides an energy-saving control system for an automobile generator, which is used to perform the following Figure 2 An energy-saving control method for an automobile generator is shown, the system comprising: The adjustment identification module 11 defines the state space of the automobile generator according to the operation data set of the automobile generator, performs adjustment identification according to the operation data set, and defines the action space of the automobile generator based on the adjustment identification result.

[0015] In the embodiment of the present application, the adjustment identification module first collects and extracts the operating data set in real time and synchronously, performs correlation analysis in combination with the operating status information of the generator and the vehicle, extracts multiple key state variables and associates and combines them to construct the state space of the generator. On this basis, the operating mode of the generator is further analyzed, the adjustment requirements are identified, and the identification combination is generated through the adjustment identification. Finally, the operating state of the generator is identified according to the identification combination, the action variables are extracted and associated, and the action space of the generator is defined.

[0016] Furthermore, in the system provided in the embodiment of the application, the adjustment identification module 11 is also used for: The operation data set is obtained by performing real-time sensing and recording of the operation of the automobile generator; the operation status information of the automobile generator and the operation status information of the automobile are synchronously extracted according to the data acquisition timing; the operation status of the automobile generator and the operation status of the automobile are correlated based on the operation data set to obtain multiple correlation coefficients; the operation data set is traversed according to the multiple correlation coefficients to extract multiple state variables; the multiple state variables are associated and combined to determine multiple state combinations, and the state space of the automobile generator is constructed.

[0017] In the embodiment of the present application, firstly, the sensor array deployed on the automobile generator and the vehicle records the operation data set including the battery power, generator output power, engine load and vehicle acceleration / deceleration status in real time. Among them, the battery power is detected by the voltage sensor, the range is 10V-15V, and it is divided into high battery power (≥13V), medium battery power (10.5V-13V), and low battery power (<10.5V); the generator output power is calculated by the power sensor, the range is 0kW-5kW, and it is divided into low power (0-1.5kW), medium power (1.5kW-3.5kW), and high power (3.5kW-5kW); the vehicle acceleration / deceleration state is measured by the speed sensor, and it is divided into acceleration (acceleration>0m / s²), deceleration (acceleration<0m / s²) and uniform speed (acceleration=0m / s²) based on the speed change rate. All the collected data are synchronized and sorted according to the timestamp to form a time-series operation data set.

[0018] Next, according to the data acquisition sequence, the operating status information of the automobile generator and the operating status information of the vehicle are synchronously extracted. Then, the operating data set is subjected to correlation analysis to quantify the degree of correlation between the operating status of the automobile generator and the operating status of the vehicle. Specifically, the Pearson correlation coefficient calculation method is adopted, and the range of the correlation coefficient is [-1,1], which measures the linear correlation between the two variables. When r>0, it indicates positive correlation; when r<0, it indicates negative correlation; when r=0, it indicates no correlation. When performing the calculation, the mean of each variable (such as battery power and generator power) is first calculated; secondly, the product of the deviation of the two variables from the mean at each time point is calculated, and all time points are accumulated to obtain the deviation product sum; then, the square sum of the deviation of each variable from the mean is calculated separately, and the square root is calculated to obtain the standard deviation; finally, the deviation product sum is divided by the product of the standard deviation to obtain the correlation coefficient between the two variables. For example, the analysis results may show that the correlation coefficient between generator power and vehicle acceleration is 0.85, indicating that the two have a strong positive correlation; the correlation coefficient between engine load and battery power is -0.65, indicating that the battery power tends to decrease significantly when the engine load is high. Through this calculation process, multiple correlation coefficients are obtained.

[0019] Then, the operation data set is traversed according to multiple correlation coefficients to extract multiple state variables. That is, the operation data with correlation coefficients greater than a preset threshold is selected, where the preset threshold is pre-set by technical experts. The multiple state variables include battery power status, generator power status, engine load status, and vehicle acceleration / deceleration status.

[0020] After the state variables are extracted, multiple state variables are associated and combined through the Cartesian product method to generate all possible state combinations. Each state combination corresponds to the state of the vehicle generator under a specific operating condition. For example, combination 1 is high battery power (≥13V), medium generator power (1.5kW-3.5kW), and vehicle acceleration (acceleration>0m / s²); combination 2 is low battery power (<10.5V), high generator power (3.5kW-5kW), high engine load (70%-100%), and vehicle uniform speed (acceleration=0m / s²). All state combinations form a complete state space.

[0021] Furthermore, in the system provided in the embodiment of the application, the adjustment identification module 11 is also used for: Based on the operation data set, the operation mode of the automobile generator is analyzed to determine multiple adjustment requirements; based on the multiple state combinations, the multiple state variables are adjusted according to the multiple adjustment requirements to generate multiple identification combinations; according to the multiple identification combinations, the operation state of the automobile generator is traversed to perform action recognition to determine multiple action variables; the multiple action variables are associated and combined to construct the action space of the automobile generator.

[0022] In an embodiment of the present application, the operation mode of the automobile generator is first analyzed based on the operation data set to clarify the regulation requirements of the automobile generator under different operating conditions. Specifically, the data of the battery power, generator output power and vehicle acceleration / deceleration status are first timestamped and synchronized to ensure that each variable corresponds to the same time point. The sliding window technology is then used to analyze the changing trend of the data. For example, when the battery power is lower than 10.5V, the regulation requirement is identified as "need to be charged"; when the vehicle is in an accelerating state (acceleration>0m / s²), the regulation requirement is identified as "need high power support"; and when the vehicle is traveling at a constant speed and the battery power is higher than 13V, the regulation requirement is identified as "stop generator output". Through this dynamic analysis, multiple regulation requirements of the generator under various operating conditions are clarified.

[0023] Next, an adjustment mark is generated for each state variable based on the results of the operation mode analysis, and the specific adjustment methods under different states are further clarified. Specifically, the identification rule of the battery power is that when the battery power is lower than 10.5V, it is marked as "need to be charged"; when the battery power is higher than or equal to 13V, it is marked as "reduce charging or stop charging". The identification rule of the generator output power is based on the vehicle operation status. When the vehicle accelerates, it is marked as "need high power output"; when the vehicle is traveling at a constant speed and the battery power is high, it is marked as "stop output". The identification of the vehicle status is carried out according to the acceleration change. When the acceleration is greater than 0m / s², it is marked as "high power support"; when the acceleration is less than 0m / s², it is marked as "reduce power output"; when the acceleration is equal to 0m / s², it is marked as "maintain or stop output". Through these rules, an adjustment mark is generated for each state variable.

[0024] After the adjustment flag is generated, the Cartesian product method is used to perform Cartesian product combination on the adjustment flags of all state variables to generate multiple flag combinations. The flag set of the battery state includes "need to charge", "reduce charging" and "stop charging"; the flag set of the generator output power includes "high power output", "low power output" and "stop output"; the flag set of the vehicle state includes "acceleration", "deceleration" and "constant speed". All flag combinations are generated by traversing the values ​​of each flag set one by one. For example, combination 1 is "need to charge + high power output + acceleration", and combination 2 is "reduce charging + stop output + constant speed".

[0025] After generating the identification combination, the action variable is identified for each identification combination through the rule mapping method to clarify the executable actions of the generator in the specific state. For example, the action variable corresponding to the identification combination "need to charge + high power output + acceleration" is "high power output"; the action variable corresponding to the identification combination "reduce charging + stop output + uniform speed" is "stop generator operation". The definition of action variables includes high power output (for low battery), low power output (for high battery), and stop generator (for high battery and constant speed). By traversing all identification combinations and mapping them one by one to the corresponding action variables, multiple action variables are determined.

[0026] Finally, multiple action variables are associated and combined to construct the action space of the automobile generator, which contains all possible action variables and is associated with the corresponding identification combinations one by one. When constructing the action space, all action variables are collected, each identification combination is associated with the corresponding action variable, and the result is stored as a multidimensional matrix. For example, combination 1 is mapped to the action "high power output" and combination 2 is mapped to the action "stop the generator operation".

[0027] In summary, through operating mode analysis, multiple adjustment requirements of the automobile generator are determined; through adjustment identification generation, the adjustment methods of different state variables are clarified; through identification combination generation, identification combinations covering all operating conditions are generated; through action variable identification, the specific actions corresponding to each identification combination are clarified; finally, through action space construction, a comprehensive logical definition of the automobile generator under different operating conditions is achieved, providing complete data support and execution basis for energy-saving optimization.

[0028] The operation control model construction module 12 introduces a reward function, combines the state space and the action space to perform reinforcement learning training, and constructs an operation control model of the automobile generator.

[0029] In the embodiment of the present application, the operation control model construction module introduces a reward function, where the function expression is: , where R represents the reward obtained by performing action a in state s; The target energy consumption value is pre-set by technical experts based on the current state, such as battery power, vehicle operation state, etc. For example, the target energy consumption is higher in the acceleration state and lower in the constant speed state; is the actual energy consumption value, which is obtained through real-time monitoring by sensors; It is a penalty value used to measure the degree of deviation between the action and the state requirement, taking into account the power output, charging state and action matching.

[0030] Penalty value The power output deviation, charging state deviation and state matching are comprehensively considered. The first is the power output deviation penalty, which is used to measure the deviation between the actual power output and the required power. When the penalty is 0, When the penalty is 2, when the deviation The penalty is 5. The second penalty is the state of charge penalty, which is used to evaluate the deviation between the actual battery capacity and the target battery capacity. When , the penalty is 0; deviation When , the penalty is 2; deviation The penalty is 5. Finally, the state matching penalty is used to evaluate the compatibility of the current state and the action. For example, when "Stop Generator" is selected in the "High Battery Power + Constant Speed" state, the penalty is 0; if "High Power Output" is selected, it does not meet the operating requirements and a penalty value of 5 is directly given.

[0031] During the reinforcement learning training process, an initial state s is randomly selected from the state space, and an action a is selected in the action space. After executing action a in state s, the immediate reward R is calculated, and the current state, action, reward, and new state transferred after the action are recorded and stored as training samples. By analyzing these sample data, the model optimizes the action selection strategy and gradually improves the cumulative reward value to ensure that the actions selected in the future under the same or similar states are more in line with energy saving and performance goals.

[0032] Through multiple rounds of reinforcement learning training, the operation control model gradually optimizes the action selection strategy to achieve optimal control of the automobile generator in all states. Finally, the operation control model dynamically outputs control actions according to the real-time state of the input, completes the construction of the operation control model of the automobile generator, achieves the dual goals of energy saving and performance optimization, and provides reliable technical support for intelligent operation.

[0033] The operation dynamic prediction module 13 performs interactive feedback with the automobile generator through the operation control model to generate a control feedback parameter set, performs operation dynamic prediction on the automobile generator based on the control feedback parameter set, and obtains an operation prediction dynamic trend.

[0034] In an embodiment of the present application, the operation dynamic prediction module interacts with the automobile generator through the operation control model to generate a control feedback parameter set. Specifically, the operation data set is synchronized to the operation control model for fitting analysis to obtain an initial operation control scheme. After that, the scheme is executed on the automobile generator to generate an interactive parameter set, and energy-saving judgment is made based on the interactive parameter set, the energy-saving deviation value is calculated, and the control feedback parameter set is obtained by weighted calculation. Subsequently, the expected energy-saving operation threshold is set, and a positive feedback or negative feedback response signal is generated based on the comparison between the interactive parameter set and the threshold, and the feedback result is optimized in combination with abnormal analysis and deviation calculation. Finally, the control feedback parameter set is time-serialized and feature-analyzed, the prediction time window is determined by regression analysis, the operation change trend is generated and integrated into the operation prediction dynamic trend.

[0035] Furthermore, in the system provided in the embodiment of the application, the running dynamic prediction module 13 is also used for: The operation data set is synchronized to the operation control model for fitting analysis to obtain an initial operation control scheme; the initial operation control scheme is executed on the automobile generator for control interaction to generate an interaction parameter set; an operation energy-saving judgment is performed based on the interaction parameter set to generate an energy-saving judgment result, and an operation energy-saving deviation is calculated according to the energy-saving judgment result to generate an energy-saving deviation value; and a weighted calculation is performed on the interaction parameter set according to the energy-saving deviation value to determine the control feedback parameter set.

[0036] In the embodiment of the present application, the operation data set is first synchronized to the operation control model, and the initial operation control scheme is generated through fitting analysis. Specifically, this process uses the decision logic of the operation control model to match the real-time collected operation data (such as battery power, generator power output, vehicle operation status, etc.) with the state space and action space in the model to generate a control strategy suitable for the current operation state, that is, the initial operation control scheme.

[0037] The initial operation control scheme is then executed on the vehicle generator, and the key feedback data during the execution process is collected in real time through sensors to generate an interactive parameter set. The interactive parameter set includes information such as actual power output, battery charging status, and vehicle operation status response.

[0038] Then, based on the generated interactive parameter set, an operation energy-saving judgment is made. The actual operation data is compared with the target energy-saving level through the preset energy-saving operation threshold to evaluate the energy-saving effect of the control scheme. If the operation data contained in the interactive parameter set meets or exceeds the energy-saving threshold, the current operation is judged to meet the energy-saving target, and the energy-saving judgment result is generated as energy-saving. If the energy-saving target is not met, the data anomalies in the interactive parameter set are further analyzed, and the operation energy-saving deviation is calculated.

[0039] Then, according to the energy-saving judgment result, positive feedback or negative feedback response signals are generated respectively. If the operation data meets the energy-saving threshold, a positive feedback signal is generated, and the initial operation control plan is retained and recorded as a valid plan. At the same time, the positive feedback signal is added to the energy-saving judgment result and output; if the operation data does not meet the standard, a negative feedback signal is generated, and multiple abnormal values ​​in the interaction parameters are extracted in combination with abnormal analysis.

[0040] Finally, the interactive parameter set is weighted and calculated according to the energy-saving deviation value, and the operating parameters are optimized by comprehensively considering factors such as energy-saving deviation, battery power deviation and vehicle state response to generate the final control feedback parameter set. Specifically, the energy-saving deviation value is first extracted from the interactive parameter set, including the excess of actual power output, the insufficient percentage of actual charging efficiency and the excess time of vehicle response delay. Then, the extracted deviation values ​​are normalized one by one, and each deviation value is proportionally converted to its corresponding maximum allowable deviation, so as to eliminate the influence of different physical units and standardize all deviation values ​​into a dimensionless unified dimension. After the normalization process is completed, the weight of each parameter is calculated according to the normalized result of the deviation value, that is, the weight of each parameter is determined by comparing the normalized results, and the sum of the weights is 1. Then, the various operating parameters in the interactive parameter set are weighted and corrected according to the calculated weights. For the power output parameter, the actual power value is corrected according to the weight of the power output, and it is adjusted to a value closer to the target power output, thereby reducing the deviation of the power output. For charging efficiency parameters, the charging rate is adjusted or the target charging state is optimized according to the charging efficiency weight to improve the actual charging efficiency, ensure that the minimum charging efficiency requirements are met, and avoid overcharging or undercharging. For vehicle response parameters, the actual response time is optimized according to the response weight, the response delay is reduced, and the accuracy of action execution is improved, thereby reducing the impact of delays on overall operating efficiency.

[0041] Finally, the corrected power output, charging efficiency, and vehicle response parameters are integrated into a control feedback parameter set.

[0042] Furthermore, in the system provided in the embodiment of the application, the running dynamic prediction module 13 is also used for: An expected energy-saving operation threshold is set to determine whether the interaction parameter set is greater than or equal to the expected energy-saving operation threshold; if the interaction parameter set is greater than or equal to the expected energy-saving operation threshold, a positive feedback response signal is generated, and the initial operation control plan is retained and recorded according to the positive feedback response signal in combination with the interaction parameter set, and the positive feedback response signal is added to the energy-saving determination result for output; if the interaction parameter set is less than the expected energy-saving operation threshold, a negative feedback response signal is generated, and an abnormal analysis is performed on the interaction parameter set according to the negative feedback response signal, and multiple abnormal interaction parameters are extracted, and the negative feedback response signal is added to the energy-saving determination result for output; a deviation is calculated based on the multiple abnormal interaction parameters and the expected energy-saving operation threshold to obtain the energy-saving deviation value.

[0043] In the embodiment of the present application, the energy-saving operation threshold is first set by technical experts based on historical operation data and system performance goals. The threshold includes multiple key indicators, such as the maximum allowable power consumption, the minimum charging efficiency, and the upper limit of the delay of the vehicle operation response. It is used to determine whether the interactive parameter set is greater than or equal to the expected energy-saving operation threshold.

[0044] Next, each data in the interactive parameter set is compared with the energy-saving operation threshold one by one to determine whether it meets the set requirements. For the operation data that meets the threshold, the current state is determined to be up to standard; for the data that does not meet the standard, it is determined that there is an abnormality. If all the data in the interactive parameter set is greater than or equal to the energy-saving operation threshold, a positive feedback response signal is generated, indicating that the current operating state meets the energy-saving target. The positive feedback signal is associated with the interactive parameter set, and the current initial operation control scheme is saved as a valid scheme for reuse under similar operating conditions. At the same time, the positive feedback signal is added to the energy-saving judgment result, and the output indicates that the system energy-saving effect meets the standard.

[0045] If some data in the interactive parameter set is less than the energy-saving operation threshold, a negative feedback response signal is generated, and the interactive parameter set is analyzed for abnormalities. Extract multiple abnormal interactive parameters that do not meet the standards during operation, such as the deviation of the actual power output above the target value, the percentage difference of the charging efficiency below the target value, or the number of milliseconds that the vehicle response delay exceeds. The extraction process of abnormal parameters clarifies the specific source of deviation by comparing them with the threshold item by item, and associates the abnormal interactive parameters with the negative feedback signal and adds them to the energy-saving judgment result.

[0046] Finally, the energy-saving deviation value is calculated based on the difference between the extracted abnormal interaction parameters and the expected energy-saving operation threshold. For each abnormal parameter, the specific difference between the actual value and the target value is quantified, such as recording the specific value of the actual power exceeded, the percentage of the actual charging efficiency insufficient, and the time of the response delay exceeded. All deviation values ​​are summarized to obtain the energy-saving deviation value.

[0047] Furthermore, in the system provided in the embodiment of the application, the running dynamic prediction module 13 is also used for: The control feedback parameter set is time-serialized to determine a feedback parameter time series; a feature analysis is performed on the control feedback parameter set according to the feedback parameter time series to extract a plurality of control features; a regression analysis is performed on the automobile generator based on the plurality of control features to determine a plurality of operation prediction time windows; a change prediction is performed on the control feedback parameter set according to the plurality of operation prediction time windows to construct a plurality of operation change trends; the plurality of operation change trends are integrated according to the feedback parameter time series to generate the operation prediction dynamic trend.

[0048] In an embodiment of the present application, the control feedback parameter set is first processed to determine the feedback parameter time series, that is, the parameters such as power output, battery power, vehicle operating status, etc. are time-sorted according to the timestamps of parameter collection to form continuous time series data.

[0049] Then, the time series of feedback parameters is analyzed to extract multiple key control features. For example, by analyzing the time series of power output, its change trend, fluctuation amplitude and stability characteristics are extracted; by analyzing the charge and discharge curve of battery power, the rate of change, fluctuation cycle and extreme value characteristics of power are extracted; by analyzing the time series of vehicle operation status, the power demand mode during acceleration, deceleration or constant speed driving is extracted.

[0050] After feature extraction is completed, the random forest regression model is used to predict and analyze the feedback parameter time series. The training of the random forest regression model is based on historical operation data, which includes the actual values ​​of the feedback parameters in the past operation process and their time series changes. During the model training process, multiple groups of samples are first randomly selected from the historical data, and each group of samples is used to train a decision tree. Each decision tree learns the mapping relationship between feature values ​​and target values ​​by recursively splitting nodes to fit the nonlinear change law of the feedback parameters. After training, the random forest integrates the output results of all decision trees and generates an overall prediction result by averaging the predicted values.

[0051] The trained random forest regression model is then used to predict the feedback parameter time series and determine multiple operation prediction time windows. These time windows define the range of changes in the operating status within a specific time period in the future, such as predicting the change trend of the battery power in the next hour or the fluctuation range of the power output in the next ten minutes. Based on these time windows, the model further predicts the changes in the feedback parameter time series and constructs multiple operation change trends, including battery power change trends and power output change trends. The battery power change trend reflects the charging and discharging status of the battery in the future, helping to identify potential low-power risks; the power output change trend predicts the dynamic changes in the generator power output to evaluate whether it can meet the power needs of the vehicle.

[0052] Finally, the constructed multiple operation change trends are integrated according to the feedback parameter time series to generate a complete operation prediction dynamic trend.

[0053] The operation constraint condition formulation module 14 maps the operation dynamic trend to the state space and the action space respectively, and formulates the operation constraint condition.

[0054] In an embodiment of the present application, the operation constraint formulation module formulates comprehensive operation constraints by mapping the operation dynamic trend to the state space and the action space. Specifically, firstly, the operation state trend is extracted from the operation dynamic trend with the state variable as the index, and the state variable is predicted and marked to generate the state prediction label, and then mapped to the state space to formulate the operation state constraint; at the same time, the operation action trend is extracted from the operation dynamic trend with the action variable as the index, and the action variable is predicted and marked to generate the action prediction label, and then mapped to the action space to formulate the operation action constraint. Finally, the operation state constraint and the operation action constraint are synergistically integrated to form a complete operation constraint.

[0055] Furthermore, in the system provided in the embodiment of the application, the operation constraint condition formulation module 14 is also used for: The multiple state variables are used as the first index data, and the running dynamic trends are traversed to search and extract multiple running state trends; the multiple action variables are used as the second index data, and the running dynamic trends are traversed to search and extract multiple running action trends; the multiple state variables are predicted and marked according to the multiple running state trends to generate state prediction labels, and the multiple running state trends are mapped to the state space according to the state prediction labels to formulate running state constraints; the multiple action variables are predicted and marked according to the multiple running action trends to generate action prediction labels, and the multiple running action trends are mapped to the action space according to the action prediction labels to formulate running action constraints; the running state constraints are collaboratively transformed with the running action constraints to determine the running constraints.

[0056] In an embodiment of the present application, firstly, a plurality of state variables are used as the first index data, and the operation dynamic trend is traversed for retrieval to extract a plurality of operation state trends. The operation dynamic trend is a prediction of future operation state changes generated according to a time series, and a time interval matching method is used to analyze the changes of state variables in the operation dynamic trend section by section. For example, for the battery power, through the segmented trend analysis of the future prediction curve, key change nodes are extracted, such as the time point when the power gradually decreases from higher than 13V to 10.5V; for power output, a threshold detection method is used to mark the time period higher than 5kW to form a power fluctuation trend. Through this process, the specific change trend of each state variable in the future time period is obtained.

[0057] Then, multiple action variables are used as the second index data to traverse the operation dynamic trends for retrieval, and multiple operation action trends are extracted. Action variables describe specific behaviors that the system can perform, such as high power output, low power output, stop output, etc. Through the behavior demand matching method, the system requirements reflected in the dynamic trend are compared with the action variables one by one. For example, if the operation dynamic trend shows that the vehicle continues to accelerate during a certain period of time and the battery is insufficient, the action trend of "high power output + charging" is matched; if the vehicle is traveling at a constant speed and the battery is sufficient, the action trend of "low power output" or "stop output" is matched. Through step-by-step retrieval, the operation action trends that may need to be executed in the future are extracted.

[0058] Then, the state variables are mapped with labeling rules according to the extracted operating state trends to generate state prediction labels. For example, according to the specific range of battery power, a segmented labeling method is used to label it as "low power" (lower than 10.5V), "medium power" (10.5V~13V) or "high power" (higher than 13V); according to the fluctuation amplitude of power output, it is labeled as "high power fluctuation" (more than 5kW) or "low power fluctuation" (less than 2kW). The corresponding relationship between the state prediction label and the operating state trend is mapped to the state space to form operating state constraints, such as "in the low power state, charging must be started first".

[0059] Similarly, action variables are tagged with specific requirements based on the extracted operation action trends to generate action prediction labels. For example, action variables are tagged as "high power output", "low power output" or "stop output", and are further tagged in combination with the vehicle's operating status and power requirements. The action prediction labels and operation action trends are mapped to the action space, and operation action constraints are formulated, such as "maintaining high power output and ensuring charging during the vehicle acceleration phase".

[0060] Finally, the operating state constraints and operating action constraints are matched according to rules to form the final operating constraints. By comparing the condition ranges of states and actions one by one, comprehensive rules are formulated, such as "when the battery power is less than 10.5V and the vehicle is in the acceleration state, high power output must be started and charging must begin". The collaborative matching process ensures the logical consistency between state variables and action variables, so that the final operating constraints can fully adapt to the needs of future operating scenarios.

[0061] Through the above steps, using methods such as time interval matching, threshold detection, behavior requirement matching and rule collaborative matching, key state and action information are gradually extracted from the operation dynamic trend, prediction labels are generated and mapped to the state space and action space, and finally clear operation constraints are formed.

[0062] The energy-saving control module 15 performs operation control on the automobile generator according to the operation constraint conditions, determines an operation control strategy, and executes the operation control strategy to perform automatic energy-saving control on the automobile generator.

[0063] In the embodiment of the present application, the energy-saving control module controls the operation of the automobile generator according to the operation constraints, determines multiple operation state classes by analyzing the operation state constraints, and determines multiple operation action classes by analyzing the operation action constraints. After that, a closed-loop simulation control is performed in combination with the operation state class and the operation action class to generate a simulation control effect, and an energy-saving analysis is performed based on the effect. The operation control is corrected according to the analysis results, and energy-saving scheduling recommendation information is generated. Finally, by dynamically adjusting the operation state class and the operation action class, the operation control strategy is formulated and executed to realize the automatic energy-saving control of the automobile generator.

[0064] Furthermore, in the system provided in the embodiment of the application, the energy-saving control module 15 is also used for: According to the operating state constraint conditions, the automobile generator is operated and analyzed to determine a plurality of operating state classes; according to the operating action constraint conditions, the automobile generator is operated and analyzed to determine a plurality of operating action classes; based on the plurality of operating state classes and the plurality of operating action classes, closed-loop simulation control is performed to generate a simulation control effect; based on the simulation control effect, energy-saving analysis is performed, and operation control correction is performed according to the analysis results to generate energy-saving scheduling recommendation information; according to the energy-saving scheduling recommendation information, the plurality of operating state classes and the plurality of operating action classes are dynamically adjusted to formulate the operation control strategy.

[0065] In the embodiment of the present application, the real-time operating state of the generator is first analyzed according to the operating state constraints, and the parameters such as battery power, power output, and vehicle operating state are compared with the classification rules in the operating state constraints using a multi-condition matching method. For example, the "low power" category is matched according to the real-time collected battery power data (such as 11V), the "high power" category is matched according to the power output data (such as 5.5kW), and the "acceleration" category is matched according to the vehicle acceleration data (such as 0.3m / s²). By combining these classification results, multiple operating state categories are determined, such as "low power-high power-acceleration" or "medium power-low power-constant speed".

[0066] Next, the possible control actions of the generator are analyzed according to the operating action constraints, and the control requirements of the current operating state are matched one by one using the action requirement matching method and the logical decomposition method. For example, in the "low battery-high power-acceleration" operating state, the "high power output + start charging" action is determined according to the operating action constraints; in the "medium battery-low power-constant speed" operating state, the action requirement of "low power output + stop charging" is matched. By comparing the operating state class and the control requirements one by one, the specific control actions that can be executed in each operating state are clarified, thereby dividing multiple operating action classes, such as "high power output + start charging" and "low power output + keep charging".

[0067] Then, closed-loop simulation control is performed based on the operating state class and the operating action class. Through simulation modeling technology, the operating effects of different state and action combinations are simulated. For example, "high power output + start charging" is executed in the "low battery-high power-acceleration" state to test power consumption, battery charging efficiency and operating stability; "low power output + stop charging" is executed in the "medium battery-low power-constant speed" state to test energy saving effects and vehicle response. In the simulation test, key data such as energy consumption, power fluctuation amplitude and battery recovery speed are recorded to generate simulated control effects.

[0068] Next, based on the simulated control effect, the energy-saving analysis of the combination of the operating state class and the operating action class is carried out, and the energy-saving effect of each combination is evaluated by comparing the energy consumption, efficiency and stability indicators. For example, the high energy consumption problem of the "high power output + unnecessary charging" combination is identified, which may be due to excessive power output or excessive charging; while the "low power output + maintenance charging" combination shows high energy-saving characteristics and is marked as a priority recommendation solution. In the analysis, the data decomposition analysis method is further used to identify specific links with high energy consumption, such as excessive output power or unnecessary charging time at a specific stage.

[0069] According to the results of energy-saving analysis, operation control correction is performed to optimize the combination of operation status and operation action. For example, in the "low battery - acceleration" state, "high power output + start charging" is optimized to "high power output + precision charging", and unnecessary energy consumption is reduced by providing charging support only within the range with the highest charging efficiency; in the "medium battery - uniform speed" state, "low power output + stop charging" is adjusted to "stop output" to further reduce energy consumption.

[0070] Then, energy-saving scheduling recommendation information is generated to describe the correspondence between the optimized operating state class and the operating action class. For example, it is recommended to prioritize "high power output + precise charging" in the "low battery-acceleration" state, and prioritize "stop output" in the "medium battery-constant speed" state. The module converts the recommendation information into specific operations for dynamic adjustment, optimizes and adjusts the operating state class and the operating action class, and ensures that the optimal control combination is used in subsequent operations.

[0071] Finally, the optimized operation state class and operation action class are combined to formulate the operation control strategy and clarify the corresponding rules of each state and action. For example, formulate the strategy of "in the low-power acceleration state, execute high-power output and start precise charging"; in the "medium-power uniform speed state, execute the strategy of stopping output".

[0072] Through the above series of steps, from state analysis to simulation control, from energy-saving analysis to correction optimization, we finally completed the formulation of an efficient and energy-saving operation control strategy, providing a scientific control solution for the subsequent operation of the generator.

[0073] In the embodiments of the present application, in summary, the embodiments of the present application have at least the following technical effects: The present application defines the state space of the automobile generator according to the operation data set of the automobile generator, adjusts the mark according to the operation data set, and defines the action space of the automobile generator based on the adjustment mark result; introduces a reward function, combines the state space and the action space for reinforcement learning training, and constructs an operation control model of the automobile generator; through the operation control model and the automobile generator for interactive feedback, a control feedback parameter set is generated, and the operation dynamic prediction of the automobile generator is performed based on the control feedback parameter set to obtain the operation prediction dynamic trend; the operation dynamic trend is mapped to the state space and the action space respectively, and the operation constraint conditions are formulated; the automobile generator is operated according to the operation constraint conditions, the operation control strategy is determined, and the operation control strategy is executed to automatically control the automobile generator for energy saving. The present invention solves the technical problems that the energy-saving control strategy in the prior art is not intelligent enough, the energy consumption is high, and the operation efficiency is low. The state space and the action space are constructed by the operation data set, and the reward function is introduced for reinforcement learning training, the operation control model is generated, the model interactive feedback and dynamic prediction are combined, the operation constraint conditions are formulated, the operation control strategy is optimized based on the constraint conditions, and the automatic energy-saving control is executed, so as to achieve the technical effect of improving the operation efficiency of the generator and reducing energy consumption.

[0074] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0075] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0076] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. An energy-saving control method for an automobile generator, characterized in that: The method comprises: An adjustment identification module, wherein the adjustment identification module defines a state space of the automobile generator according to an operation data set of the automobile generator, performs adjustment identification according to the operation data set, and defines an action space of the automobile generator based on the adjustment identification result; An operation control model construction module, wherein the operation control model construction module introduces a reward function, combines the state space and the action space to perform reinforcement learning training, and constructs an operation control model of the automobile generator; An operation dynamic prediction module, wherein the operation dynamic prediction module interacts with the automobile generator through the operation control model to generate a control feedback parameter set, performs operation dynamic prediction on the automobile generator based on the control feedback parameter set, and obtains an operation prediction dynamic trend; An operation constraint condition formulation module, wherein the operation constraint condition formulation module maps the operation dynamic trend to the state space and the action space respectively, and formulates operation constraint conditions; An energy-saving control module controls the operation of the automobile generator according to the operation constraint conditions, determines an operation control strategy, and executes the operation control strategy to perform automatic energy-saving control on the automobile generator.

2. The energy-saving control method of an automobile generator according to claim 1, characterized in that: Defining a state space of the automobile generator according to the operating data set of the automobile generator, the method includes: The operation data set is obtained by performing real-time sensing and recording of the operation of the automobile generator; According to the data acquisition sequence, the operating status information of the automobile generator and the operating status information of the automobile are synchronously extracted; Based on the operation data set, a correlation analysis is performed between the operation state of the automobile generator and the operation state of the automobile to obtain a plurality of correlation coefficients; Traversing the operation data set according to the multiple correlation coefficients to extract multiple state variables; The multiple state variables are associated and combined to determine multiple state combinations, and the state space of the automobile generator is constructed.

3. The energy-saving control method of an automobile generator according to claim 2, characterized in that: Performing adjustment identification according to the operating data set, and defining the action space of the automobile generator based on the adjustment identification result, the method includes: Analyzing the operation mode of the vehicle generator based on the operation data set to determine a plurality of adjustment requirements; According to the multiple state combinations, the multiple state variables are adjusted according to the multiple adjustment requirements to generate multiple identification combinations; Traversing the operating state of the automobile generator according to the plurality of identification combinations to perform action recognition and determine a plurality of action variables; The plurality of action variables are associated and combined to construct the action space of the automobile generator.

4. The energy-saving control method of an automobile generator according to claim 1, characterized in that: By performing interactive feedback between the operation control model and the automobile generator, a control feedback parameter set is generated, and the method includes: Synchronizing the operation data set to the operation control model for fitting analysis to obtain an initial operation control solution; Executing the initial operation control scheme on the automobile generator to perform control interaction and generate an interaction parameter set; Performing an operation energy-saving determination based on the interactive parameter set to generate an energy-saving determination result, calculating an operation energy-saving deviation according to the energy-saving determination result, and generating an energy-saving deviation value; The interaction parameter set is weightedly calculated according to the energy-saving deviation value to determine the control feedback parameter set.

5. The energy-saving control method of an automobile generator as claimed in claim 4, characterized in that: Performing an operation energy-saving determination based on the interactive parameter set, generating an energy-saving determination result, calculating an operation energy-saving deviation according to the energy-saving determination result, and generating an energy-saving deviation value, the method comprising: Setting an expected energy-saving operation threshold, and determining whether the interaction parameter set is greater than or equal to the expected energy-saving operation threshold; If the interactive parameter set is greater than or equal to the expected energy-saving operation threshold, a positive feedback response signal is generated, and according to the positive feedback response signal combined with the interactive parameter set, the initial operation control scheme is retained and recorded, and the positive feedback response signal is added to the energy-saving determination result for output; If the interaction parameter set is less than the expected energy-saving operation threshold, a negative feedback response signal is generated, an abnormal analysis is performed on the interaction parameter set according to the negative feedback response signal, a plurality of abnormal interaction parameters are extracted, and the negative feedback response signal is added to the energy-saving determination result for output; Deviation calculation is performed based on the multiple abnormal interaction parameters and the expected energy-saving operation threshold to obtain the energy-saving deviation value.

6. The energy-saving control method of an automobile generator according to claim 1, characterized in that: Based on the control feedback parameter set, the automobile generator is dynamically predicted to obtain a dynamic trend of the predicted operation, and the method includes: Performing time serialization on the control feedback parameter set to determine a feedback parameter time series; Performing feature analysis on the control feedback parameter set according to the feedback parameter time series to extract multiple control features; Performing regression analysis on the automobile generator based on the multiple control characteristics to determine multiple operation prediction time windows; Predicting changes in the control feedback parameter set according to the multiple operation prediction time windows to construct multiple operation change trends; The multiple operation change trends are integrated according to the feedback parameter time series to generate the operation prediction dynamic trend.

7. The energy-saving control method of an automobile generator as claimed in claim 3, characterized in that: The operation dynamic trend is mapped to the state space and the action space respectively, and operation constraint conditions are formulated, the method comprising: Using the multiple state variables as first index data, traversing the operation dynamic trend for retrieval, and extracting multiple operation state trends; Using the multiple action variables as second index data, traversing the operation dynamic trend for retrieval, and extracting multiple operation action trends; Predictively mark the multiple state variables according to the multiple operating state trends, generate state prediction labels, map the multiple operating state trends to the state space according to the state prediction labels, and formulate operating state constraints; Predictively label the multiple action variables according to the multiple operation action trends, generate action prediction labels, map the multiple operation action trends to the action space according to the action prediction labels, and formulate operation action constraint conditions; The operating state constraint condition and the operating action constraint condition are collaboratively transformed to determine the operating constraint condition.

8. The energy-saving control method of an automobile generator as claimed in claim 7, characterized in that: According to the operation constraint conditions, the automobile generator is controlled to operate and an operation control strategy is determined, the method comprising: Performing operation analysis on the automobile generator according to the operation state constraint conditions to determine a plurality of operation state classes; Performing operation analysis on the automobile generator according to the operation action constraint conditions to determine a plurality of operation action classes; Perform closed-loop simulation control based on the multiple operating state classes and the multiple operating action classes to generate a simulation control effect; Energy-saving analysis is performed based on the simulated control effect, operation control correction is performed according to the analysis result, energy-saving scheduling recommendation information is generated, the multiple operation status classes and the multiple operation action classes are dynamically adjusted according to the energy-saving scheduling recommendation information, and the operation control strategy is formulated.

Images