Charging pile power regulation and control method and device, electronic equipment and storage medium

Abstract

The invention relates to the technical field of electric vehicle charging control, and discloses a charging pile power regulation and control method and device, electronic equipment and a storage medium. The method comprises the following steps of: monitoring multi-dimensional time sequence data such as charging voltage, current and battery temperature in real time, performing state segmentation, and calculating the similarity between a data sequence of each stage and a standard charging template so as to accurately screen out a stable charging state stage. And extracting a power adjustment amount based on the time sequence characteristic and the fluctuation mode of the stage, fusing the adjustment amount of the environment temperature, the equipment load parameter and the charging power, and dynamically generating a power switching rule by using a fuzzy logic controller. And finally, smooth switching and cyclic adjustment of the power levels of the charging piles are realized according to the rule. According to the method, the accuracy and the anti-interference capability of charging state identification are improved, and the adaptability and the collaborative optimization capability of power regulation to a complex environment and a power grid condition are enhanced, so that the charging efficiency and the power grid stability are both considered on the premise of ensuring the safety of the battery.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric vehicle charging control, in particular to a charging pile power regulation method and device, an electronic device and a storage medium. BACKGROUND

[0002] The current power regulation strategy of electric vehicle charging piles generally relies on threshold monitoring of single or a small number of parameters such as battery voltage and current. The judgment basis for the stage conversion of the common constant current constant voltage charging method is usually a preset voltage threshold. When the charging voltage reaches a certain fixed threshold, the system switches from the constant current stage of large current to the constant voltage stage. This judgment mechanism based on instantaneous threshold is sensitive to sensor noise and instantaneous fluctuations of the battery, which may lead to inaccurate judgment of the stage switching time. Early entry into the constant voltage stage will prolong the charging time, and late entry may affect the battery life. The judgment of the charging state relies too much on the instantaneous parameters of the battery itself, and lacks evaluation of the overall dynamic mode of the charging process.

[0003] The existing power adjustment scheme mostly adopts a preset fixed rule or a lookup table method, and the control logic is relatively rigid. These methods usually do not systematically analyze key external factors such as the ambient temperature of the charging pile, the dynamic changes of the total load on the grid side or inside the charging station, and the charging state of the battery. In a low-temperature environment, the battery resistance increases, and if the standard mode is still used for high-power charging, there is a safety hazard; during the peak electricity consumption period, if the charging pile cluster simultaneously operates at high power, it will cause an impact on the local power grid. The fixed power switching rule is difficult to adapt to such complex scenarios with multiple variables and ambiguous conditions, and cannot achieve dynamic balance among safety, efficiency and the power grid. SUMMARY

[0004] The purpose of the present application is to provide a charging pile power regulation method and device, an electronic device and a storage medium to solve the problems raised in the background.

[0005] To achieve the above purpose, the present application provides a charging pile power regulation method, which comprises: real-time monitoring of the charging process of the electric vehicle battery, collecting multi-dimensional time series data of charging voltage, charging current and battery temperature; state segmentation processing of the multi-dimensional time series data, identifying different stages of the charging process, and calculating the similarity of the data sequence of each stage with the standard charging template, thereby screening out the stage sequence representing the stable charging state; extracting the adjustment amount of the charging power according to the time sequence characteristics and fluctuation mode of the data points in the stage sequence; fusing the ambient temperature, equipment load parameters and adjustment amount of the charging power, and using a fuzzy logic controller to generate a switching rule for the charging power; The switching rule is executed to make the charging pile smoothly switch between preset power levels and cyclically adjust the output power.

[0006] Preferably, the state segmentation processing of the multi-dimensional time series data identifies different stages of the charging process and calculates the similarity of the data sequence of each stage to a standard charging template, thereby screening out a stage sequence representing a stable charging state, which includes: The multi-dimensional time series data is input into a pre-constructed hidden Markov model, and the most likely charging state sequence is decoded by a Viterbi algorithm, including a constant current stage, a constant voltage stage and a floating charge stage. The dynamic time warping algorithm is applied to the data sequence in each charging state stage to calculate the minimum path distance between it and the pre-stored standard charging template, and the stage similarity score is obtained. Based on the stage similarity score, an adaptive threshold is set, and stages with scores higher than the threshold are marked as candidate stable stages. The continuity of the candidate stable stages is checked, and if the adjacent stages are continuous in time axis and the similarity score changes smoothly, they are merged into the stage sequence.

[0007] Preferably, the adjustment amount of the charging power is extracted according to the time sequence characteristics and fluctuation patterns of the data points in the stage sequence, which includes: The Kalman filter is applied to the charging voltage and current data in the stage sequence for state estimation to obtain a smoothed sequence after removing noise. The first-order difference value of the smoothed sequence is calculated to capture the time sequence fluctuation characteristics, the fluctuation energy spectrum is constructed based on the first-order difference value, and the significant fluctuation points are identified by a peak detection algorithm. According to the amplitude and interval of the significant fluctuation points, combined with the historical adjustment record of the charging power, the adjustment amount of the charging power is derived.

[0008] Preferably, the switching rule of the charging power is generated by fusing the environmental temperature, device load parameters and adjustment amount of the charging power using a fuzzy logic controller, which includes: The environmental temperature, device load current and adjustment amount of the charging power are taken as input variables of the fuzzy logic controller, and the membership functions of the input variables are set. Inference is performed through a pre-configured fuzzy rule base, which contains if-then rules defined based on expert experience. The defuzzification output is obtained using the gravity method to obtain the switching time and target power value of the charging power, and the switching time and target power value are integrated to form the switching rule.

[0009] Preferably, the inference through the pre-configured fuzzy rule base includes: Language variables defining input variables, including low, medium, and high three levels; For each input variable combination, match the corresponding fuzzy rule, if the ambient temperature is high and the device load is high, the switching opportunity is early; Weighted average of all activated rules to get preliminary reasoning results, consistency check of preliminary reasoning results to ensure priority strategy when rule conflict.

[0010] Preferably, the use of the barycenter method to defuzzify the output includes: Calculate the area barycenter of the membership function curve of the output fuzzy set, and take the corresponding numerical value of the area barycenter as the accurate output value, that is, the switching time of the charging power. Quantify the target power value according to the output value and map it to the preset power level.

[0011] Preferably, the execution of the switching rule makes the charging pile smoothly switch between the preset power levels and cyclically adjust the output power. According to the switching time in the switching rule, trigger the power conversion instruction, use the ramp function to control the power change process, ensure the linear transition of the output power between levels, monitor the response data of the battery after each switching, and feedback to the state segmentation processing step for closed-loop adjustment. When a cycle is completed, reinitialize the power sequence and start the next round of cycle adjustment.

[0012] The power sequence is an ordered set of multiple target power values sorted by switching time.

[0013] Preferably, the application further includes a charging pile power regulation device, the device includes: A data acquisition module for real-time monitoring of the charging process of an electric vehicle battery and collecting multi-dimensional time series data of charging voltage, charging current and battery temperature; A state processing module connected to the data acquisition module for state segmentation processing of the multi-dimensional time series data, identifying different stages of the charging process, and calculating the similarity of data sequences of each stage with the standard charging template, thereby screening out stage sequences representing stable charging states; A feature extraction module connected to the state processing module for extracting the adjustment amount of charging power according to the time sequence characteristics and fluctuation patterns of data points in the stage sequence; A rule generation module connected to the feature extraction module for fusing environmental temperature and device load parameters and generating switching rules of charging power using a fuzzy logic controller; A power execution module connected to the rule generation module for executing the switching rule to make the charging pile smoothly switch between the preset power levels and cyclically adjust the output power.

[0014] Preferably, the present application further includes an electronic device, comprising: a processor and a memory; wherein the memory is configured to store a computer program executable on the processor; and the processor is configured to execute the program stored on the memory to implement the steps of the charging pile power regulation method as described above.

[0015] Preferably, the present application further includes a computer readable storage medium storing one or more programs, which when executed by an electronic device comprising a plurality of application programs, cause the electronic device to perform the steps of the charging pile power regulation method as described above.

[0016] Compared with the prior art, the present application has the following beneficial effects: The stable state phase is screened by calculating the similarity of the data sequence of each stage of the charging process and the standard charging template. This method improves the judgment basis from an isolated instantaneous threshold to the overall pattern matching of continuous time series data. This analysis based on data sequence similarity can effectively filter out abnormal data points caused by measurement noise or transient disturbance in the charging process, thereby more essentially identifying the real stable state charging process of the battery. Based on the evaluation of the overall fluctuation pattern rather than the triggering of a single point threshold, the division of the charging stage is more accurate and robust, providing a reliable time sequence judgment basis for subsequent power adjustment and avoiding frequent or unnecessary power switching caused by misjudgment.

[0017] The environmental temperature and device load parameters are fused, and the power switching rules are dynamically generated using a fuzzy logic controller, realizing the collaborative decision of multi-source heterogeneous parameters. The ability of fuzzy logic to handle imprecise and nonlinear problems enables the system to convert information in different dimensions such as battery internal state, external environmental constraints, and system operation restrictions into a unified decision basis. The power switching instructions generated by this mechanism are no longer fixed, but adaptively adjust the power curve according to the environmental temperature to ensure battery safety, and intelligently adjust the output according to the device load to smooth the power grid impact. The power regulation behavior therefore has dynamic adaptability, achieving multi-objective optimization such as safety, efficiency, and power grid stability in complex real-world application scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 The working principle diagram of the charging pile power regulation method described in the present application; Figure 2 The flowchart for extracting the charging power adjustment amount; Figure 3 The flowchart for generating switching rules for the fuzzy logic controller; Figure 4 The relationship diagram of charging power switching timing and control parameters; Figure 5A schematic diagram of the defuzzification process for fuzzy inference. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0020] Please refer to Figure 1 The present application provides a charging pile power regulation method, which comprises the following steps: real-time monitoring of the charging process of an electric vehicle battery, collection of multi-dimensional time series data of charging voltage, charging current and battery temperature. Subsequently, the multi-dimensional time series data is subjected to state segmentation processing, different stages of the charging process are identified, and the similarity of the data sequence of each stage to the standard charging template is calculated, thereby screening out a stage sequence representing a stable charging state. Based on the stage sequence, the adjustment amount of the charging power is further extracted according to the time sequence characteristics and fluctuation patterns of the data points therein. In the decision-making link, the method fuses environmental temperature and equipment load parameters, and uses a fuzzy logic controller to generate switching rules for the charging power. The generated switching rules are executed to enable the charging pile to smoothly switch between preset power levels and cyclically adjust the output power, so as to achieve optimized control of the charging process.

[0021] In a specific implementation, the multi-dimensional time series data is subjected to state segmentation processing, different stages of the charging process are identified, and the similarity of the data sequence of each stage to the standard charging template is calculated, thereby screening out a stage sequence representing a stable charging state, which is the basis for constructing a reliable power regulation strategy. In implementation, the multi-dimensional time series data stream composed of the real-time collected charging voltage, charging current and battery temperature is input into a pre-constructed and trained hidden Markov model. The hidden Markov model can effectively model the time series data, and describe the evolution law and observation characteristics between different states in the charging process through its internal state transition probability matrix and observation probability matrix. In a specific implementation, the hidden Markov model is decoded using the Viterbi algorithm, which is a dynamic programming algorithm that recursively calculates the maximum probability path of each time point in each hidden state and backtracks to find the global optimal path, thereby decoding the most likely charging state sequence in the entire charging process. Typical charging states include constant current stage, constant voltage stage and floating charge stage. The application of the Viterbi algorithm ensures the global optimality of state division and avoids the influence of local optimal solution on the segmentation result.

[0022] After identifying each charging state phase by using the Hidden Markov Model and the Viterbi algorithm, the data sequence in each phase needs to be evaluated for similarity. In a specific implementation, for each identified charging state phase, such as a constant current phase or a constant voltage phase, the corresponding charging voltage and charging current data sequence in the phase is extracted, and then a dynamic time warping algorithm is applied to calculate the minimum path distance between the data sequence and a standard charging template pre-stored in a database. The standard charging template is an ideal charging curve obtained based on statistical analysis of a large amount of historical normal charging data, and represents the expected charging behavior under standard working conditions. The dynamic time warping algorithm can overcome the stretching and shifting effects of different charging instances on the time axis, and by finding a non-linear alignment path with the smallest cumulative distance between the two sequences, a phase similarity score is obtained, which is a quantitative indicator, and the smaller the value, the more similar the actual sequence is to the standard template.

[0023] Based on the phase similarity score calculated by the dynamic time warping algorithm, a threshold needs to be set to screen stable phases. In a specific implementation, an adaptive threshold method is used, which is not a fixed value, but is dynamically adjusted according to the overall quality of the current charging process; for example, the moving average and standard deviation of the phase similarity scores in recent periods can be calculated, and the threshold is set to the moving average plus several times the standard deviation, which can adapt to the fluctuations caused by individual differences of different batteries and different initial charging conditions. Phases with a phase similarity score higher than the adaptive threshold are marked as candidate stable phases, which means that the charging behavior in the phase is highly consistent with the expected behavior, the data quality is high, and it is suitable for fine power adjustment analysis.

[0024] After obtaining a series of candidate stable phases, continuity checking is needed to form the final phase sequence for power adjustment. In a specific implementation, the continuity checking includes time continuity and score stationarity checks. The time continuity check confirms whether the candidate stable phases appear continuously on the time axis or are isolated fragments, and the score stationarity check analyzes whether the similarity scores between adjacent candidate stable phases change smoothly without dramatic jumps. If adjacent candidate stable phases are continuous on the time axis and the similarity scores change smoothly, these phases are merged into a longer, continuous phase sequence representing a stable charging state. It can be understood that the merged phase sequence contains stable charging data over a longer time span, providing a richer and more reliable information base for subsequent extraction of charging power adjustment amounts. It can be understood that through the series of steps of state segmentation, similarity calculation, threshold screening, and continuity checking, the final screened phase sequence can effectively represent the stable state in the charging process, providing a solid data support for accurate power regulation.

[0025] In some embodiments, the training process of the hidden Markov model needs to be done independently. The parameters of the hidden Markov model include the initial state probability distribution, the state transition probability matrix, and the observation probability matrix, which need to be trained and learned through a large amount of historical charging data. The training process usually uses the Baum-Welch algorithm, which is an expectation maximization algorithm that can automatically learn the optimal parameters of the hidden Markov model using unlabeled time series data. In specific implementation, a large amount of complete charging process data covering different battery types, different environmental temperatures, and different initial state of charge are collected as a training set, and then the Baum-Welch algorithm is applied for iterative optimization until the model parameters converge, thereby obtaining a hidden Markov model that can be generalized for new charging process state recognition.

[0026] In some embodiments, the implementation of the dynamic time warping algorithm can adopt different local path constraints and weight settings. Local path constraints specify the allowed movement directions when calculating the cumulative distance path, for example, horizontal, vertical, and diagonal movements can be allowed, but different weights can be assigned to different directions to control the shape of the path. In specific implementation, a symmetric weight setting is usually selected, that is, the weights of horizontal and vertical movements are the same, while the weight of diagonal movement may be different, which helps to obtain reasonable sequence alignment effect. Optionally, in order to speed up the calculation, the dynamic time warping algorithm can be optimized, for example, using windowing constraints, searching for paths only within a strip-shaped area, thereby significantly reducing the amount of calculation while ensuring accuracy. Optionally, the original minimum path distance calculated can be normalized, for example, divided by the path length, to eliminate the influence of different sequence lengths on the similarity score, so that the similarity between stages of different lengths is comparable.

[0027] In specific implementation, the setting strategy of the adaptive threshold can be adjusted according to actual application requirements. In addition to using the moving average and standard deviation method, a weighted average based on stage duration can also be considered, or prior knowledge about the battery type can be introduced to fine-tune the threshold. The "change is stable" in the continuity test also needs a quantitative judgment standard, for example, the difference between the similarity scores of adjacent stages can be calculated, and an upper limit of the difference is set, when the score difference of consecutive stages is below the upper limit, it is considered that the change is stable. The merging operation of the stage sequence usually extends the time boundary of the continuous stages that meet the conditions, merges multiple short stages into one long stage, and treats all data points in the merged stage as a whole data sequence for subsequent processing.

[0028] Embodiment 2: see Figure 2In a specific implementation, the adjustment amount of the charging power is extracted according to the time sequence characteristics and fluctuation patterns of the data points in the stage sequence, and this process relies on in-depth signal analysis and feature extraction on the screened stable stage sequence. In implementation, first, the Kalman filter, a recursive optimal estimation algorithm, is applied to the charging voltage and charging current data contained in the stage sequence representing the stable charging state for state estimation. The Kalman filter performs optimal estimation of the true state of the system in the sense of minimum mean square error through iteration between the prediction and update steps. For the charging process, the state variables of the system can be defined as the true values of the charging voltage and the charging current, and the observation variables are the charging voltage and charging current readings collected by the sensor containing noise. The prediction step of the Kalman filter predicts the state at the next time point based on the state space model of the system, and the update step corrects the predicted value using new observation data. By continuously adjusting the Kalman gain matrix, the Kalman filter can effectively suppress random noise in the observation data and output a smoothed sequence that removes noise. This smoothed sequence more truly reflects the trend of changes in the battery charging parameters.

[0029] After obtaining the smoothed sequence that removes noise, time sequence characteristics and fluctuation patterns that can guide power adjustment need to be extracted from it. Calculating the first-order difference values of the smoothed sequence is a basic step to achieve this goal. The first-order difference values reflect the change between adjacent data points in the sequence, and the calculation formula can be expressed as:

[0030] wherein: represents the first-order difference value at time point represents the smoothed sequence value (which can be the charging voltage or the charging current) at time point represents the smoothed sequence value at time point

[0031] Based on the calculated first-order difference value sequence, a fluctuation energy spectrum is constructed. The fluctuation energy spectrum is a tool for analyzing the time sequence fluctuation energy according to the frequency distribution; in a specific implementation, short-time Fourier transform or time-frequency analysis methods more suitable for non-stationary signals can be applied to the first-order difference value sequence to obtain a distribution diagram of signal energy in the two-dimensional plane of time and frequency, i.e., the fluctuation energy spectrum. The fluctuation energy spectrum can reveal the appearance and disappearance of fluctuations of different frequency components in the time axis during the charging process. By analyzing the fluctuation energy spectrum through a peak detection algorithm, the peak detection algorithm finds points with significantly higher amplitudes in local regions by scanning the amplitude values of the fluctuation energy spectrum. These points are identified as significant fluctuation points, which usually correspond to characteristic events in the charging process caused by load changes, internal electrochemical reaction mutations, or external disturbances.​​​

[0032] According to the characteristics of the identified significant fluctuation points, combined with historical information, the adjustment amount of the charging power is finally derived. The amplitude of the significant fluctuation points directly reflects the intensity of the fluctuation, and a larger amplitude usually means a more intense disturbance; the interval between the significant fluctuation points reflects the frequency of the fluctuation event, and the shorter the interval, the more frequent the fluctuation. In specific implementation, it is necessary to analyze the fluctuation characteristics in association with the historical adjustment records of the charging power, which are stored in the memory unit of the system and contain successful power adjustment strategies and their effect evaluation in the face of similar fluctuation patterns in the past. By analyzing the amplitude and interval pattern of the current significant fluctuation points and matching similar cases in the historical adjustment records, a specific and quantitative charging power adjustment amount can be derived, which indicates the direction and magnitude of the power change. It can be understood that this data-driven method makes the power adjustment decision not only dependent on the current instantaneous state, but also incorporates historical experience, thereby making the adjustment more accurate and robust.

[0033] In some embodiments, the implementation of the Kalman filter requires the design of a specific charging process model. The state equation and the observation equation of the system need to be defined, for example, the state variables can include the charging current and its rate of change, and the observation variables are the measured charging current values. The two key parameters, process noise covariance matrix and observation noise covariance matrix, need to be reasonably set or estimated online according to the sensor characteristics and system dynamics. In specific implementation, adaptive Kalman filtering technology can be used, which can dynamically adjust the process noise covariance matrix or the observation noise covariance matrix according to the statistical characteristics of the innovation sequence (the difference between the observation value and the predicted value), so as to better cope with system model uncertainty or time-varying noise characteristics, and further improve the accuracy of state estimation.

[0034] In some embodiments, the construction of the fluctuation energy spectrum can use a variety of time-frequency analysis methods. In addition to the short-time Fourier transform, wavelet transform is also a commonly used and powerful tool. Wavelet transform analyzes signals by using a scalable and shiftable wavelet basis function, which can provide higher frequency resolution and lower time resolution in the low-frequency part, and higher time resolution and lower frequency resolution in the high-frequency part. This characteristic is very suitable for analyzing transient high-frequency fluctuations and slowly varying low-frequency trends in non-stationary signals. In specific implementation, a suitable mother wavelet function can be selected, and appropriate scale and shift parameters can be set to perform continuous wavelet transform or discrete wavelet transform on the first-order difference numerical sequence, and the corresponding wavelet coefficient matrix is obtained. The modulus square of this wavelet coefficient matrix can be used as a form of fluctuation energy spectrum.

[0035] In some embodiments, the implementation strategy of the peak detection algorithm can be adjusted. The algorithm needs to set a threshold to determine whether a point is "significant" enough to be considered as a peak, which can be fixed or adaptive, for example, based on the average and standard deviation of the local region of the fluctuation energy spectrum. In addition, in order to avoid mis-detecting a very wide fluctuation region as multiple dense peaks, a minimum peak separation constraint is usually needed, that is, after a peak is detected, no other peaks are detected within a set time-frequency window around it. Optionally, the detected peaks can also be filtered, for example, only keeping the peaks with energy higher than a certain multiple of the global average, to focus on the most important fluctuation events.

[0036] Embodiment 3: refer to Figure 3 In specific implementation, the switching rule of charging power is generated by using a fuzzy logic controller to fuse the ambient temperature, the device load parameter and the adjustment amount of charging power. This process is the key link of converting precise numerical input into fuzzy logic consistent with human decision-making thinking, and finally outputting clear control instructions. In implementation, the real-time value read by the ambient temperature sensor, the current value obtained by the device load current monitoring unit and the charging power adjustment amount obtained from the stage sequence feature extraction are taken as the three input variables of the fuzzy logic controller. For each input variable, its corresponding membership function is defined. The membership function is the core concept of fuzzy logic, which describes the degree to which a precise input value belongs to a certain fuzzy language variable. The function value is between 0 and 1, indicating the strength of the membership. For example, for the ambient temperature input variable, three fuzzy sets can be defined: "low temperature", "medium temperature" and "high temperature". Each fuzzy set corresponds to a membership function, which can be triangular, trapezoidal or Gaussian, etc. By setting the parameters of the function, the boundary range of each fuzzy language is determined, realizing the softening of the grading from precise value to fuzzy concept.

[0037] In specific implementation, the inference by the pre-configured fuzzy rule base is the core operation step of the fuzzy logic controller. The fuzzy rule base contains a set of if-then rules defined in advance based on the experience of domain experts. Each rule describes a fuzzy relationship between input conditions and output conclusions. The "if" part (antecedent) of the rule is a combination of fuzzy languages of input variables, for example, "if the ambient temperature is high, the device load current is high, and the charging power adjustment amount is positive large"; the "then" part (consequent) of the rule is the fuzzy language of the output variable, for example, "then the switching time is early". The inference process is to activate all fuzzy rules related to the current precise input value by calculating its membership degree under different membership functions, and calculate the contribution degree of each activated rule to the output fuzzy set according to the fuzzy logic operation rules (such as "and" operation usually takes the minimum value, "or" operation usually takes the maximum value).

[0038] In specific implementations, a defuzzification method is used to process the fuzzy output generated by the inference to obtain the precise charging power switching timing and target power value that can be used for execution. The defuzzification method, also known as the centroid method, is a commonly used method for defuzzification. It calculates the area centroid of the region enclosed by the membership function curve of the output fuzzy set and the horizontal coordinate axis, and takes the precise numerical value corresponding to this centroid point on the horizontal coordinate axis as the clear output of the system. For the charging power switching rule, the output variables can include "switching timing" and "target power level offset". The calculation of the centroid method can be represented as:

[0039] wherein: represents the precise output value obtained after defuzzification, represents the domain of the output variable, represents the membership function value of the synthesized output fuzzy set at point after fuzzy inference, and the integral operation is performed on the entire output domain. The precise numerical value obtained is quantized to a specific switching time point (and target power value) through a pre-defined mapping relationship, and these information is finally integrated to form a complete and executable charging power switching rule.

[0040] In some embodiments, the design of the input variable membership function has an important influence on the control performance. The shape, overlapping area, and coverage range of the membership function need to be carefully designed. Triangular membership functions are commonly used due to their simple calculation, trapezoidal membership functions can provide a wider smooth region, and Gaussian-shaped membership function curves are smooth. The overlap between membership functions is necessary, as it ensures that changes in input values can smoothly cause changes in output, avoiding dramatic jumps in control output. In specific implementations, the initial values of the parameters of the membership functions can be set based on expert experience, and then fine-tuned and optimized through simulation or actual running data to make the behavior of the controller more consistent with expectations.

[0041] In some embodiments, the construction of the fuzzy rule base is a systematic process. The rule base should be as complete as possible, covering all possible typical input situation combinations. The number of rules increases exponentially with the increase in input variables and the number of fuzzy partitions for each variable, so a balance needs to be struck between control accuracy and rule base complexity. For a system with three input variables, each divided into three levels (low, medium, high), theoretically up to 27 rules are needed. In specific implementations, by analyzing the input-output characteristics of the system, some rule combinations that are almost impossible to occur in practice or have little impact on output can be omitted, thereby simplifying the rule base. The consequent (output) of each rule also needs to be carefully defined according to the control target, for example, whether to be conservative and safe or aggressive and efficient.

[0042] In some embodiments, the fuzzy inference process can involve different types of internal fuzzy implication operators and composition methods. Commonly used implication operators include the Mamdani minimum operation implication and the Larsen product operation implication, and commonly used composition methods include the max-min composition and the max-product composition. Different combinations of operators will produce slightly different inference results. In specific implementations, the Mamdani-type fuzzy system is widely used due to its intuitive nature, which usually uses the minimum operation as the implication operator and the maximum operation for composition. It can be understood that the selection of these mathematical operations will affect the aggressiveness and smoothness of the inference. It can be understood that the advantage of the fuzzy logic controller is that it does not rely on the precise mathematical model of the controlled object, but uses language-based expert knowledge for inference, which is particularly suitable for handling systems such as charging processes that have nonlinearity, time-varying and uncertainty.

[0043] Referring to Figure 4 , it demonstrates the complex relationship between the charging power switching timing and the control parameters based on the fuzzy logic controller. The figure clearly presents the intelligent adjustment strategy of the system for the charging power switching timing under different combinations of environmental temperature and device load current conditions through the form of contour lines. In the chart, the environmental temperature is the vertical axis variable, the device load current is the horizontal axis variable, and the color depth represents the adjustment degree of the switching timing. From the figure, it can be observed that in the high temperature and high load area, the system tends to adopt a more aggressive early switching strategy, which reflects sufficient consideration for device thermal management and safe operation. In the low temperature and low load area, the switching timing is relatively conservative, ensuring the stability of the charging process. The distribution pattern of the contour lines reflects the comprehensive effect of the membership functions and inference rules in the fuzzy logic controller. The change in the spacing between the curves shows the nonlinear characteristics of the system response, with higher control sensitivity in the key parameter interval. This design enables the charging pile to dynamically optimize the power switching strategy according to the real-time working conditions, ensuring the charging efficiency while ensuring the safe and reliable operation of the system. The overall chart presents the superiority of fuzzy logic control in handling multivariate coupling problems, and through the combination of expert experience and mathematical models, it realizes the effective mapping of human decision-making thinking in automatic control systems.

[0044] In a specific implementation, the inference by the pre-configured fuzzy rule base and the defuzzification using the center of gravity method to generate crisp control instructions are the core steps of the fuzzy logic controller. In the inference stage by the pre-configured fuzzy rule base, the linguistic variables of the input variables need to be defined first. For the three input variables of the ambient temperature, the device load current, and the charging power adjustment amount, each variable is divided into discrete linguistic levels for fuzzy description. Typical divisions include low, medium, and high. For example, the linguistic variable of the ambient temperature can be defined as "temperature low", "temperature medium", and "temperature high"; the linguistic variable of the device load current is defined as "load low", "load medium", and "load high"; and the linguistic variable of the charging power adjustment amount is defined as "adjustment amount negative large", "adjustment amount zero", and "adjustment amount positive large". Each linguistic level corresponds to a membership function, which maps the precise input value to a membership degree in the interval [0, 1], indicating the degree to which the value belongs to a certain linguistic level.

[0045] In a specific implementation, for each possible combination of input variable linguistic variables, the corresponding if-then rule needs to be matched in the fuzzy rule base. The fuzzy rule base is a collection of conditional statements, each of which describes the fuzzy output strategy to be taken under specific input conditions. For example, a typical rule can be: "if the ambient temperature is high, the device load is high, and the charging power adjustment amount is positive large, then the switching timing is significantly advanced". The rule matching process is parallel, and the system will traverse the entire fuzzy rule base to find all rules whose antecedent conditions have a non-zero intersection with the current input fuzzy set. The fuzzy inference engine then calculates the activation strength of each activated rule. For antecedents using "and" connections, the minimum operation is usually used as the activation strength or trigger weight of the rule.

[0046] In practice, a weighted average of all activated rules is taken to obtain a preliminary inference result. This process involves scaling the consequent of each activated rule according to its activation strength, and then using a union operation to merge all scaled output fuzzy sets into a total, comprehensive output fuzzy set. This preliminary inference result is still a fuzzy set, representing the fuzzy distribution of possible values ​​for the output variable under the combined effect of all activated rules. Subsequently, a consistency check is required on the preliminary inference result. The purpose of the consistency check is to handle potential rule conflicts in the rule base. A rule conflict arises when two or more rules are activated under similar input conditions but give completely different or even opposite output suggestions. In practice, a priority strategy is used to resolve conflicts. Each rule can be assigned a static priority weight, or the priority can be dynamically determined based on the matching degree of the rule's antecedent conditions. When a conflict occurs, the output suggestion of the rule with higher priority will dominate when synthesizing the total output fuzzy set, thereby ensuring the consistency and rationality of the control decision.

[0047] In practical implementation, referring to Table 1, using the centroid method to defuzzify the output is a crucial step in transforming fuzzy inference results into precise control quantities. The specific operation of the centroid method for defuzzifying the output involves calculating the area centroid under the membership function curve of the aforementioned comprehensive output fuzzy set. For a discrete universe of discourse, the centroid method calculation can be represented as multiplying the value of each point in the universe by its corresponding membership degree, summing all products, and then dividing by the sum of the membership degrees of all points, thus obtaining a precise and clear numerical output. This numerical output represents a quantified value of the charging power switching timing, such as a scalar representing the "advancement" or "delay." Based on this precise output value, it needs to be further quantized and mapped to a preset power level to determine the target power value. For example, the system may preset several discrete power levels. The "target power change" value obtained through defuzzification will be normalized to the closest actually available power level through a predefined mapping table or function, thus forming the final executable power switching command.

[0048] Table 1: Fragments of the Fuzzy Rule Base

[0049] In some embodiments, the construction of the fuzzy rule base can be more refined. The linguistic level division of input variables is not limited to three levels: "low, medium, and high." For example, it can be extended to five levels: "very low, low, medium, high, and very high," or even more levels, to provide more nuanced discrimination capabilities. Correspondingly, the size of the fuzzy rule base will also increase. Table 1 shows a simplified example of rule representation. The rule base in a real system will be more detailed, striving to cover all important operating conditions. It can be understood that a well-designed fuzzy rule base is a decisive factor in the performance of the fuzzy logic controller.

[0050] In practical implementation, the precise numerical value output by the defuzzification process needs to be adjusted by an output scaling factor to map to the actual system's time scale or power level. The output scaling factor is an adjustable parameter used to scale up or down the normalized value obtained from defuzzification to the dimensions required for practical engineering applications. For example, the defuzzified output value might be in the range of [-1, 1]. By multiplying by a scaling factor, the actual amount of time that needs to be advanced or delayed can be obtained. Optionally, for mapping the target power value, in addition to simply rounding to the nearest preset level, more complex strategies can be adopted, such as considering the minimum switching step size limit between the current power level and the target power level, or fine-tuning based on the battery's real-time acceptance capability. Optionally, to ensure control stability, the defuzzified output value obtained over several consecutive control cycles can be smoothed using a filter, such as a first-order low-pass filter, to suppress high-frequency jitter in the output. It is understandable that the centroid method of defuzzification, due to its intuitive calculation and ability to comprehensively reflect the information of the entire output fuzzy set, is one of the most widely used methods. It can produce smooth control output, contributing to the smooth switching of charging pile power.

[0051] See Figure 5This diagram illustrates the complete reasoning and defuzzification process in a fuzzy logic controller, highlighting the key steps in transforming fuzzy reasoning results into precise control commands. The diagram shows three main fuzzy sets: the delayed set, the normal set, and the advanced set, represented by different line types. These sets represent linguistic variables adjusted for switching timing, mapping precise values ​​to fuzzy concepts through membership functions. Activated areas are filled in purple, representing the combined effect of each rule being triggered under the current input conditions. Black vertical lines mark the precise output value calculated using the centroid method, the core result of the defuzzification process. The application of the centroid method ensures the smoothness and continuity of the control output, avoiding the abrupt changes caused by traditional switching control. This method calculates the area centroid of the output fuzzy set to obtain a precise control quantity that comprehensively reflects the contributions of all activated rules. The diagram clearly demonstrates the transformation process of fuzzy logic control from multi-rule parallel reasoning to a single precise value output. This approach enables the system to exhibit good adaptability and robustness when dealing with nonlinear and time-varying systems based on expert-experienced linguistic rules. The effective implementation of the defuzzification process provides a reliable technical guarantee for the smooth switching of charging pile power.

[0052] Example 5: In specific implementation, executing the switching rules to smoothly switch the charging pile between preset power levels and cyclically adjust the output power is the final step in translating control decisions into actual actions and achieving closed-loop adaptive regulation. Based on the switching timing explicitly specified in the charging power switching rules generated by the fuzzy logic controller, the system triggers the corresponding power conversion command. The switching timing may be an absolute point in time, such as executing at a specific moment, or it may be a relative time offset relative to a certain charging state characteristic point.

[0053] In practical implementation, a ramp function is used to control the power change process to ensure a linear transition of output power between levels. A ramp function is a control function that makes the output change linearly with time. Its core parameter is the slope of the switching ramp, i.e., the rate of power change (kW / s). For example, assuming the current power level is 22 kW and the target power level is 11 kW, with a preset power change rate of 2.2 kW / s, the entire power decrease process will take (22-11) / 2.2 = 5 seconds to complete. At the start of the switching, the power execution unit calculates the desired power setpoint at the current moment based on the ramp function at a fixed time interval and controls the output of the power electronics to ensure that the actual output power closely follows the setpoint, thus forming a smooth and continuous power trajectory from the source power level to the target power level. This ramp control method effectively avoids step changes in power, which can cause inrush current and harmonic interference to the battery, on-board charger, and power grid due to rapid changes in current. The use of the ramp function ensures the safety and power quality of the charging process.

[0054] After each power switching operation is completed, the system immediately initiates a monitoring program to continuously collect battery response data. The monitored physical quantities include charging voltage, charging current, and battery temperature. This data is recorded in real-time at a high frequency to assess the battery's operating status after the power switch. This monitoring data is fed back in real-time to the state segmentation processing step, i.e., re-entering the process described in Example 1. Specifically, the newly collected multi-dimensional time-series data containing post-switching response data is again input into a pre-built Hidden Markov Model for state segmentation, identifying the current charging stage, and calculating the similarity to a standard charging template to filter stable stage sequences. Subsequently, the feature extraction module analyzes the time-series characteristics and fluctuation patterns based on the new stage sequences, while the fuzzy logic controller combines the latest ambient temperature and equipment load parameters to generate new charging power switching rules adapted to the current battery state. This process constitutes a complete closed-loop adjustment system, enabling the power regulation strategy to be dynamically optimized based on the battery's actual response to the previous switching operation, achieving feedback-based adaptive control.

[0055] After a complete power adjustment cycle is completed during the charging process—for example, from initial high-power charging through several rule-based power increase or decrease switches until the charging termination condition is met—the system reinitializes the power control sequence. This reinitialization includes clearing temporary variables stored in the cache, historical power adjustment records, and intermediate calculation results accumulated in the previous cycle. The power control sequence is reset to its initial state, ready to begin the next round of cyclic adjustment for the new charging process. This cyclic adjustment mechanism allows the charging station to continuously adapt to the differences in characteristics of different batteries and the constantly changing internal and external conditions during a single charge, achieving full-cycle, adaptive optimized charging control. In some embodiments, the specific form of the ramp function can be adjusted. Besides a linear ramp, nonlinear transition functions such as an S-curve can also be used. An S-curve has a smaller rate of change at the beginning and end of the switching phases and a larger rate of change in the middle phase, enabling smoother start-up and shutdown, further reducing the impact on the system. The slope of the ramp can be a fixed parameter or it can be designed to be configurable and dynamically adjusted according to the size of the target power level difference. For example, a gentler slope can be used when switching power over a large range, and a steeper slope can be used when adjusting within a small range, in order to strike a balance between safety and fast response.

[0056] In some embodiments, the conditions for triggering the power conversion command can be more complex. In addition to relying on the absolute or relative switching timing of the fuzzy logic controller output, additional safety constraints can be introduced. For example, a watchdog timer can be set; if the switching timing point is not reached within a specified time, but the battery voltage or temperature is detected to exceed a safety threshold, the system will immediately trigger an emergency power reduction command. This command has higher priority than normal fuzzy logic control commands, thereby ensuring the inherent safety of the system.

[0057] In practical implementation, the closed-loop cycle of monitoring battery response data and feeding it back to the state segmentation processing step needs careful design. A feedback cycle that is too short can cause the system to be overly sensitive, potentially overreacting to measurement noise; a feedback cycle that is too long can lead to a sluggish system response, failing to track rapid changes in battery state in a timely manner. Optionally, the feedback cycle can be consistent with the power acquisition cycle or an integer multiple thereof, for example, performing state segmentation and rule recalculation every 10 power data acquisition cycles. Optionally, the system can be designed as event-triggered rather than time-periodic, meaning that a feedback adjustment process is only initiated when certain key characteristic quantities exceed predetermined thresholds; this approach offers higher computational efficiency. The power sequence is an ordered set of multiple target power values ​​sorted according to switching timing. Optionally, when reinitializing the power sequence, some long-learned parameters related to specific battery characteristics can be retained to improve control performance in the initial stage of the next cycle, achieving a certain degree of personalized adaptation. This continuous iteration of execution and feedback constitutes the core mechanism for the self-optimization and self-adaptation of the power regulation method.

[0058] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0059] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for regulating the power of a charging pile, characterized in that, The charging pile power regulation method includes: Real-time monitoring of the charging process of electric vehicle batteries, collecting multi-dimensional time-series data on charging voltage, charging current and battery temperature; The multi-dimensional time-series data is processed by state segmentation to identify different stages of the charging process, and the similarity between the data sequence of each stage and the standard charging template is calculated to select the stage sequence that represents the stable charging state. Based on the temporal characteristics and fluctuation patterns of the data points in the stage sequence, the adjustment amount of the charging power is extracted; By integrating ambient temperature, equipment load parameters, and charging power adjustment, a fuzzy logic controller is used to generate charging power switching rules. The switching rules are executed to smoothly switch the charging pile between preset power levels and cyclically adjust the output power.

2. The charging pile power regulation method according to claim 1, characterized in that, The process of performing state segmentation on the multi-dimensional time-series data to identify different stages of the charging process and calculating the similarity between the data sequence of each stage and the standard charging template, thereby selecting stage sequences that characterize a stable charging state, includes: The multi-dimensional time-series data is input into a pre-constructed hidden Markov model, and the most likely charging state sequence is decoded using the Viterbi algorithm. The charging state includes a constant current stage, a constant voltage stage, and a float charging stage. The dynamic time warping algorithm is applied to the data sequence in each charging state stage to calculate the minimum path distance between it and the pre-stored standard charging template, and the stage similarity score is obtained. An adaptive threshold is set based on the stage similarity score, and stages with scores higher than the threshold are marked as candidate stable stages. A continuity test is performed on the candidate stable stages. If adjacent stages are continuous on the time axis and the similarity score changes steadily, they are merged into the stage sequence.

3. The charging pile power regulation method according to claim 2, characterized in that, The step of extracting the adjustment amount of charging power based on the temporal characteristics and fluctuation patterns of data points in the stage sequence includes: A Kalman filter is applied to the charging voltage and charging current data within the stage sequence to perform state estimation, resulting in a smoothed sequence after noise removal. The first-order difference value of the smoothed sequence is calculated to capture the temporal fluctuation characteristics. A fluctuation energy spectrum is constructed based on the first-order difference value, and significant fluctuation points are identified by a peak detection algorithm. Based on the amplitude and interval of significant fluctuation points, combined with historical adjustment records of charging power, the adjustment amount of charging power is derived.

4. The charging pile power regulation method according to claim 3, characterized in that, The method of integrating ambient temperature, equipment load parameters, and charging power adjustment amounts, and using a fuzzy logic controller to generate charging power switching rules includes: The ambient temperature, equipment load current, and the adjustment amount are used as input variables of the fuzzy logic controller, and the membership function of the input variables is set. Reasoning is performed using a pre-configured fuzzy rule base, which contains if-then rules defined based on expert experience; The output is defuzzified using the center of gravity method to obtain the switching timing and target power value of the charging power. The switching timing and target power value are then integrated to form the switching rule.

5. The charging pile power regulation method according to claim 4, characterized in that, The reasoning using a pre-configured fuzzy rule base includes: Define the language variables for the input variables, including three levels: low, medium, and high; For each combination of input variables, the corresponding fuzzy rule is matched. If the ambient temperature is high and the device load is high, the switching timing is advanced. A weighted average of all activated rules is taken to obtain a preliminary inference result. The preliminary inference result is then checked for consistency to ensure that a priority strategy is used when rules conflict.

6. The charging pile power regulation method according to claim 4, characterized in that, The deblurring output using the centroid method includes: Calculate the centroid of the area under the membership function curve of the output fuzzy set, and use the value corresponding to the centroid of the area as the precise output value, that is, the timing of the switching of charging power. Quantize the target power value according to the output value and map it to the preset power level.

7. The charging pile power regulation method according to claim 1, characterized in that, The execution of the switching rules, enabling the charging pile to smoothly switch between preset power levels and cyclically adjust the output power, includes: According to the switching timing in the switching rules, the power conversion command is triggered, and the power change process is controlled by a ramp function to ensure that the output power transitions linearly between levels. After each switching, the response data of the battery is monitored and fed back to the state segmentation processing step for closed-loop adjustment. After one cycle is completed, the power sequence is reinitialized and the next round of cycle adjustment begins. The power sequence is an ordered set of multiple target power values ​​sorted according to the switching timing.

8. A power regulation device for a charging pile, characterized in that, The device includes: The data acquisition module is used to monitor the charging process of electric vehicle batteries in real time and collect multi-dimensional time-series data on charging voltage, charging current and battery temperature. The state processing module, connected to the data acquisition module, is used to perform state segmentation processing on the multi-dimensional time series data, identify different stages of the charging process, and calculate the similarity between the data sequence of each stage and the standard charging template, thereby filtering out the stage sequence that represents the stable charging state. The feature extraction module, connected to the state processing module, is used to extract the adjustment amount of charging power based on the temporal characteristics and fluctuation patterns of data points in the stage sequence. The rule generation module, connected to the feature extraction module, is used to integrate ambient temperature, equipment load parameters and charging power adjustment amounts, and to generate charging power switching rules using a fuzzy logic controller. The power execution module, connected to the rule generation module, is used to execute the switching rules, enabling the charging pile to smoothly switch between preset power levels and cyclically adjust the output power.

9. An electronic device, characterized in that, include: A processor and a memory; wherein the memory is used to store a computer program that can run on the processor; the processor is used to execute the program stored in the memory to implement the steps of the charging pile power regulation method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores one or more programs, which, when executed by an electronic device including multiple applications, cause the electronic device to perform the steps of the charging pile power control method as described in any one of claims 1-7.

Images