Controller optimization processing method of engine servo system

Abstract

The invention relates to the technical field of engine control, in particular to a controller optimization processing method of an engine servo system. The method comprises the following steps: acquiring historical environmental data when an engine works, monitoring data of a to-be-analyzed type, an initial cluster and an initial cluster center; obtaining a monitoring data segment of monitoring data of one to-be-analyzed type in one initial cluster, and then solving the average influence degree of each type of environmental data on each to-be-analyzed type; then the influence degree of each to-be-analyzed type on the initial clustering center at each moment is obtained, the initial clustering center is updated, and a final cluster is obtained; obtaining an average adjustment effect of the final cluster according to the adjustment effect after each instruction is sent in the final cluster; and adjusting an integral gain parameter of the PID algorithm according to the average adjustment effect of the final cluster to which the current moment belongs. According to the invention, the control stability and accuracy of the engine servo system controller can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of engine control, in particular to a controller optimization processing method of an engine servo system. BACKGROUND

[0002] The engine servo system is a feedback control system for precisely controlling the operation of various components of the engine. By combining sensor feedback and controller algorithms, the engine operation is ensured to be efficient, stable and fast-responding. The controller of the servo system receives target instructions and current engine-related data, generates corresponding output signals using control algorithms to control the engine speed, torque, etc., so that the engine reaches the target state.

[0003] The PID control algorithm is a commonly used control algorithm for the current engine servo system controller. According to the difference between the current data and the target value, combined with the proportional, integral and derivative parameters in the PID algorithm, the required control amount is obtained, and the engine is controlled to make a responsive action. Among them, the selection of parameters in the PID algorithm is the key to determine the control effect. The existing method mainly obtains the PID parameter value through multiple experiments. However, because the control of the engine under different working conditions will be different, the fixed PID parameter value is difficult to adapt to different situations, resulting in poor stability of the servo system control effect, which affects the working efficiency of the engine. SUMMARY

[0004] In order to solve the above technical problems, the purpose of the present application is to provide a controller optimization processing method of an engine servo system, and the technical solution adopted is as follows: An embodiment of the present application provides a controller optimization processing method of an engine servo system, which comprises: Obtaining different kinds of monitoring data and environmental data of the engine during operation in the history; screening the types of monitoring data to obtain a type to be analyzed, and clustering the monitoring data of the type to be analyzed to obtain an initial cluster and an initial cluster center; Segmenting the monitoring data according to the change degree of each monitoring data of a type to be analyzed in an initial cluster to obtain a monitoring data segment; obtaining the influence degree of each environmental data corresponding to each monitoring data segment in time sequence on the type to be analyzed; Obtaining the average influence degree of each environmental data on a type to be analyzed in each initial cluster to obtain the average influence degree of the environmental data on the type to be analyzed; The influence degree of each kind of environment data on the to-be-analyzed species is obtained based on the average influence degree of each kind of environment data on the to-be-analyzed species, the various environment data at a historical moment and the current moment; and the initial cluster center is updated based on the influence degree of each to-be-analyzed species on the initial cluster center at each moment in each initial cluster, to obtain a final cluster. The adjustment effect after each instruction is issued in history is obtained; the average adjustment effect of a final cluster is obtained according to the adjustment effect after each instruction is issued in the final cluster; and the integral gain parameter is adjusted according to the average adjustment effect of the final cluster to which the current moment belongs.

[0005] Preferably, the to-be-analyzed species is obtained by screening the kinds of monitoring data, including: The principal component analysis method is used to process the monitoring data of different kinds in history when the engine is working, to obtain the characteristic value of each kind of monitoring data, and a set number of kinds with the largest characteristic value are taken as the to-be-analyzed species.

[0006] Preferably, the monitoring data segments are obtained by segmenting according to the change degree of each monitoring data of a to-be-analyzed species in an initial cluster, including: The monitoring data of a to-be-analyzed species in an initial cluster are arranged in ascending order to obtain an arrangement sequence; the difference between the latter monitoring data and the former monitoring data of two adjacent monitoring data in the arrangement sequence is obtained and normalized to obtain the change degree of the latter monitoring data; the monitoring data with a change degree greater than a first threshold value is taken as a demarcation point, and the arrangement sequence is segmented by using the demarcation point to obtain the monitoring data segments.

[0007] Preferably, the influence degree of each kind of environment data on the to-be-analyzed species is obtained based on the average influence degree of each kind of environment data on the to-be-analyzed species, the various environment data at a historical moment and the current moment; and the initial cluster center is updated based on the influence degree of each to-be-analyzed species on the initial cluster center at each moment in each initial cluster, to obtain a final cluster. The mean value of one kind of environment data corresponding to a monitoring data segment in time sequence is obtained as the environment mean value corresponding to the monitoring data segment; the absolute value of the sum of the difference between the environment mean values corresponding to each two adjacent monitoring data segments is calculated, and the mean value is taken to obtain the average change amount of the environment data; the reciprocal of the mean value of the standard deviation of the kind of environment data corresponding to each monitoring data segment in time sequence is calculated, and is recorded as the environment data stability amount; the influence degree of the kind of environment data on the to-be-analyzed species is obtained by multiplying the average change amount of the environment data and the environment data stability amount and normalizing.

[0008] Preferably, the influence degree of each kind of environment data on the to-be-analyzed species is obtained based on the average influence degree of each kind of environment data on the to-be-analyzed species, the various environment data at a historical moment and the current moment; and the initial cluster center is updated based on the influence degree of each to-be-analyzed species on the initial cluster center at each moment in each initial cluster, to obtain a final cluster. An absolute value of a difference between an environment data of a time in history and an environment data of a current time is obtained, and an environment similarity is obtained by adding a hyperparameter to the absolute value and taking an inverse; a first eigenvalue corresponding to the environment data of the time is obtained by multiplying the environment similarity corresponding to the time in history and an average influence degree of the environment data on a to-be-analyzed species; and an influence degree of the to-be-analyzed species on an initial cluster center at the time is obtained by summing the first eigenvalues corresponding to the environment data of the time.

[0009] Preferably, the initial cluster centers are updated based on the influence degrees of the to-be-analyzed species on the initial cluster centers in each initial cluster at each time to obtain final clusters, including: An influence weight of a to-be-analyzed species on an initial cluster center of an initial cluster at a time is obtained by comparing an influence degree of the to-be-analyzed species on the initial cluster center of the initial cluster at the time and a sum of influence degrees of the to-be-analyzed species on the initial cluster center of the initial cluster at all times; An average monitoring data value corresponding to a to-be-analyzed species is obtained by weighting and averaging monitoring data of the to-be-analyzed species at each time using influence weights of the to-be-analyzed species on an initial cluster center of an initial cluster at each time; similarly, average monitoring data values corresponding to other to-be-analyzed species are obtained, and data points composed of the average monitoring data values of all to-be-analyzed species in the initial cluster form a first updated initial cluster center corresponding to the initial cluster center of the initial cluster; a first clustering result is obtained by clustering the first updated initial cluster centers corresponding to the initial cluster centers of all initial clusters, and the first clustering result is updated again until data points in a cluster in the clustering result no longer change, and the updating is stopped to obtain final clusters.

[0010] Preferably, the adjustment effect after each instruction is issued in history is obtained, including: A parameter corresponding to an engine and needing to be controlled is denoted as a control parameter, a difference between a stable control parameter of the engine after each instruction is issued in history and a set target control parameter is obtained as an adjustment effect after each instruction is issued.

[0011] Preferably, an average adjustment effect of a final cluster is obtained according to adjustment effects after each instruction is issued in the final cluster, including: The time of issuing each instruction in each final cluster is obtained according to the corresponding time in history when the instruction is issued; the reciprocal of the distance between the data point corresponding to the time of issuing each instruction in one final cluster and the cluster center of the final cluster is obtained, and is recorded as the weight of the data point corresponding to the time of issuing each instruction; and the adjustment effect after issuing each instruction in the final cluster is weighted and averaged by using the weight of the data point corresponding to the time of issuing each instruction, so that a weighted average value is obtained, and is recorded as the average adjustment effect of the final cluster.

[0012] Preferably, the integral gain parameter is adjusted according to the average adjustment effect of the final cluster to which the current time belongs, and the adjustment includes: The data point composed of the monitoring data of each to-be-analyzed category at the current time is obtained, and is recorded as a target data point; the distances between the target data point and the cluster centers of the final clusters are respectively calculated, the final cluster with the minimum distance is the final cluster to which the current time belongs; the proportional coefficient is obtained by comparing the average adjustment effect of the final cluster to which the current time belongs with the maximum value of the absolute values of the average adjustment effects of the final clusters; and the adjusted gain integral parameter at the current time is obtained by adding the proportional coefficient and the first preset value and multiplying the result by the original integral gain parameter.

[0013] The embodiment of the application has at least the following beneficial effects: the application collects different kinds of monitoring data and environmental data of the engine in history, then screens the kinds of monitoring data, screens out the kinds which are more important for the engine, obtains the monitoring data of the to-be-analyzed categories, and then reduces the complexity of analysis while ensuring the accuracy of analysis; further, in one initial cluster, the average influence degree of one kind of environmental data on one to-be-analyzed category is obtained by combining the influence of various environmental data on each to-be-analyzed category, then the influence degree of one to-be-analyzed category on the center of the initial cluster at one time in the initial cluster is obtained, and the center of the initial cluster is updated to obtain the final cluster, so that the clustering result can accurately reflect the real working condition of the engine, and the accuracy of subsequent analysis is improved; then the adjustment effect after issuing each instruction in history is obtained, the average adjustment effect of each final cluster is obtained, and the integral gain parameter of the PID control algorithm is adjusted based on the average adjustment effect, so that the problem that the fixed PID parameter is not applicable to different engine working conditions can be effectively solved, and the control stability and accuracy of the engine servo system controller are improved. BRIEF DESCRIPTION OF DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, and the advantages thereof, the following will briefly introduce the drawings needed in the embodiments or the prior art description. Obviously, the drawings in the following description are only some embodiments of the application, and for those skilled in the art, other drawings can be obtained from these drawings without creative labor.

[0015] Figure 1 A method flowchart of an engine servo system controller optimization processing method provided by an embodiment of the present application. DETAILED DESCRIPTION

[0016] In order to further clarify the technical means and effects taken by the present application to achieve the predetermined object, the following describes in detail the specific implementation, structure, features and effects of the engine servo system controller optimization processing method according to the present application, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0018] The following describes in detail the specific scheme of the engine servo system controller optimization processing method provided by the present application.

[0019] Embodiment: The main application scenario of the present application is that the control situation of the engine servo system controller is greatly affected by the working condition of the engine and the environmental situation, resulting in poor stability of the control effect of the fixed PID control algorithm parameters in the face of different engine situations, which affects the engine operation efficiency, and therefore the PID control algorithm parameters are adaptively adjusted.

[0020] Please refer to Figure 1 which shows a method flowchart of the engine servo system controller optimization processing method provided by an embodiment of the present application, which includes the following steps: Step S1, obtaining different kinds of monitoring data and environmental data of the engine during operation; screening the types of monitoring data to obtain the types to be analyzed, and clustering the monitoring data of the types to be analyzed to obtain initial clusters and initial cluster centers.

[0021] In order to realize accurate control and real-time response, the engine servo system needs to collect and monitor various data of the engine in real time, so relevant sensors are installed at corresponding positions of the engine for real-time monitoring. Different kinds of monitoring data of the engine during operation are collected, such as monitoring the engine speed by using a Hall sensor, monitoring the intake pressure by using a pressure sensor, monitoring the air mass flow into the engine by using a mass air flow sensor, monitoring the throttle opening by using a throttle position sensor, etc., and recording the data such as the injection pulse width and the ignition advance angle controlled by the control system.

[0022] Because the change of external environmental factors will change the operating state and power response characteristics of the engine, thereby affecting the PID control effect, it is also necessary to collect different types of environmental data when the engine is working, among which temperature, humidity and air pressure are the main environmental factors affecting PID control, thereby using a thermistor to monitor the intake temperature, using a humidity sensor to monitor the air humidity, and using an air pressure sensor to detect the atmospheric pressure.

[0023] In order to facilitate analysis according to multi-dimensional engine monitoring data, the engine monitoring data needs to be normalized first. The engine monitoring data in this application is all normalized data.

[0024] Because the degree of adjustment of the controller needs to be different when controlling the engine under different working conditions, that is, the control effect of the original PID parameters is different under different engine working conditions, in order to obtain PID parameters suitable for different engine working conditions, the engine working conditions need to be classified first.

[0025] The classification of engine working conditions is mainly based on the comprehensive analysis of multi-dimensional data such as engine speed, intake pressure and torque, but different types of monitoring data reflect the engine working conditions to different degrees, and some monitoring data have weak correlation with engine working conditions. Therefore, for the engine monitoring data, principal component analysis is used for processing, and the monitoring data with larger eigenvalues are the monitoring data that change more obviously under different engine states, that is, the monitoring data that can better reflect the engine working conditions. Specifically, the principal component analysis is used to process different types of monitoring data when the engine is working in history, the eigenvalues of each type of monitoring data are obtained, and the set number of types with the largest eigenvalues are taken as the types to be analyzed. The value of the preset number is 3, which is a reference value. The implementer can adjust the value of the preset number according to the actual situation, appropriately increase the accuracy of subsequent analysis, but need to pay attention to the increase of the complexity of calculation.

[0026] According to the operating state of the engine, the engine working conditions mainly include idle working condition, acceleration working condition, deceleration working condition, constant speed cruising working condition, heavy load working condition, etc. Because the engine characteristics are different under different working conditions, the required PID parameters are also different. Further, the monitoring data of each type to be analyzed at a time is combined into a data point; then according to the common working condition state of the engine, k=6 is set, and the monitoring data of the type to be analyzed is clustered to obtain the initial cluster center. Specifically, in a three-dimensional space, six initial cluster centers are selected, the distances of other data points to each initial cluster center are calculated, and they are distributed to the cluster represented by the nearest cluster center, thereby obtaining the preliminary clustering result, obtaining the initial cluster and the initial cluster center. The clustering algorithm is K-means clustering algorithm.

[0027] Step S2, segmenting the monitoring data according to the variation degree of each monitoring data of a to-be-analyzed category in an initial cluster to obtain a monitoring data segment; and obtaining the influence degree of each environment data on the to-be-analyzed category based on the environment data corresponding to each monitoring data segment in time sequence.

[0028] The performance, combustion efficiency and the like of the engine under different environmental conditions will change, that is, the relevant monitoring data of the engine under similar working conditions in different environments can also have great differences. Therefore, in order to make the K-means clustering result more reflect the real working condition of the current engine, the proximity of each data point to the cluster center needs to be adjusted according to the similarity of the environment condition of the historical data to the current environment, so that the characteristic value of the cluster center of the final clustering result is closer to the characteristics of different working conditions under the current environment.

[0029] Different environmental factors have different influences on the operation of the engine, so different environmental factors have different influences on the above-mentioned three to-be-analyzed categories. Therefore, the influence degree of each environmental factor on the above-mentioned to-be-analyzed categories is first calculated.

[0030] Because the data points in each initial cluster in the preliminary clustering result obtained in step S1 are data points with relatively high proximity of working condition characteristics, the influence of the environmental factor on each to-be-analyzed category is analyzed in each initial cluster of the preliminary clustering result. Because the analysis processes of different to-be-analyzed categories are similar, only one to-be-analyzed category and one environmental factor are taken as an example for analysis here.

[0031] Firstly, because the performance and the like of the engine have certain adaptability to the change of the environmental factor, that is, the air humidity will not cause obvious change of the data characteristics of the engine within a certain range, but when the air humidity changes beyond a certain range, it will have a relatively obvious influence. Therefore, the influence degree of the air humidity on the to-be-analyzed category A cannot be judged only by its linear relationship.

[0032] The monitoring data segment is obtained according to the variation degree of each monitoring data of a to-be-analyzed category in an initial cluster. Specifically, the monitoring data of a to-be-analyzed category in an initial cluster is arranged in order from small to large to obtain an arrangement sequence; the difference between the latter monitoring data and the former data in two adjacent monitoring data in the arrangement sequence is obtained and normalized to obtain the variation degree of the latter monitoring data.

[0033] The calculation model of the variation degree is specifically: , wherein, represents the change degree of the i+1th monitoring data in the permutation sequence corresponding to each monitoring data of the to-be-analyzed species A in the u-th initial cluster; norm represents a normalization function; and respectively represent the i+1th and the ith monitoring data in the permutation sequence corresponding to each monitoring data of the to-be-analyzed species A in the u-th initial cluster.

[0034] Because the characteristic values under the same external condition and under similar working conditions should have a high degree of similarity, that is, the change degree of the to-be-analyzed species A is small, if the to-be-analyzed species A changes significantly, it indicates that the environmental conditions corresponding to the front and back monitoring data may have changed greatly; thus, the monitoring data with a change degree greater than the first threshold value is taken as a dividing point, and the monitoring data segment is obtained by segmenting the permutation sequence using the dividing point. The environmental conditions of different monitoring data segments are greatly different. The value of the first threshold value is 0.5 because the middle value is taken as the threshold value.

[0035] Finally, the influence degree of air humidity on the to-be-analyzed species A is calculated. If the air humidity has a great influence on the to-be-analyzed species A, the air humidity values corresponding to each monitoring data segment in time sequence should be relatively close, that is, the air humidity values corresponding to the time corresponding to each monitoring data in the monitoring data segment should be relatively close, that is, the standard deviation of the air humidity values corresponding to each monitoring data segment in time sequence is small, and the air humidity values corresponding to different monitoring data segments in time sequence should have great differences and present a certain change trend. Therefore, the average value of the air humidity corresponding to each monitoring data segment in time sequence is calculated as the overall level of the air humidity corresponding to each monitoring data segment in time sequence, the influence degree of air humidity on the to-be-analyzed species A is calculated according to the difference and change trend of the air humidity between adjacent monitoring data segments, and the standard deviation of the air humidity corresponding to each monitoring data segment in time sequence.

[0036] The influence degree of each kind of environmental data on the to-be-analyzed species is calculated according to the kind of environmental data corresponding to each monitoring data segment in time sequence. Specifically, the mean value of the kind of environmental data corresponding to each monitoring data segment in time sequence is obtained as the environmental mean value corresponding to the monitoring data segment; the absolute value of the sum of the difference between the environmental mean values corresponding to each two adjacent monitoring data segments is calculated, and the mean value is obtained as the average environmental change amount; the reciprocal of the mean value of the standard deviation of the kind of environmental data corresponding to each monitoring data segment in time sequence is calculated, and is recorded as the environmental data stability amount; the average environmental change amount and the environmental data stability amount are multiplied and normalized to obtain the influence degree of the kind of environmental data on the to-be-analyzed species. The kind of environmental data here can be taken as air humidity for analysis, and the analysis is the influence degree of one kind of environmental data on one to-be-analyzed species, that is, the influence degree on the monitoring data under the to-be-analyzed species.

[0037] A specific calculation model of the influence degree of an environmental data on a to-be-analyzed species is as follows: , wherein, represents the influence degree of the air humidity (an environmental data) in the u-th initial cluster on the to-be-analyzed species A; represents the number of monitoring data segments corresponding to the monitoring data of the to-be-analyzed species A in the u-th initial cluster; and respectively represent the average values of the air humidity values corresponding to the p+1-th and p-th monitoring data segments in the monitoring data segments corresponding to the monitoring data of the to-be-analyzed species A in the u-th initial cluster, that is, the environmental average values corresponding to the p+1-th and p-th monitoring data segments; represents the difference between the environmental average values of adjacent monitoring data segments, the absolute value of the difference represents the humidity change degree, and the positive or negative of the difference represents the change direction of the corresponding air humidity when the to-be-analyzed species A increases; represents the sum of the differences between the environmental average values of adjacent monitoring data segments in the monitoring data segments corresponding to the monitoring data of the to-be-analyzed species A in the u-th initial cluster, that is, the greater the difference between the air humidity of different monitoring data segments and the more consistent the change direction of the air humidity between adjacent monitoring data segments, the greater the influence degree of the air humidity on the to-be-analyzed species A, is the average change amount of the environmental data, and the average value is used to avoid the influence of the different numbers of monitoring data segments corresponding to the monitoring data of the to-be-analyzed species A obtained by different initial clusters on the results; represents the standard deviation of the air humidity values corresponding to the p-th monitoring data segment in the monitoring data segments corresponding to the monitoring data of the to-be-analyzed species A in the u-th initial cluster, is the inverse of the average value of the standard deviation of the air humidity values corresponding to the p-th monitoring data segment in the monitoring data segments corresponding to the monitoring data of the to-be-analyzed species A in the u-th initial cluster, that is, the environmental data stability, and the smaller the standard deviation, the closer the air humidity values in the same monitoring data segment, that is, the greater the influence degree of the air humidity on the to-be-analyzed species A.

[0038] Thus, in each initial cluster, the influence degree of each environmental data on each to-be-analyzed species can be obtained.

[0039] Step S3: obtaining the average influence degree of an environmental data on a to-be-analyzed species by calculating the average value of the influence degree of the environmental data on the to-be-analyzed species in each initial cluster.

[0040] The influence degree of an environmental data on a to-be-analyzed species in an initial cluster is obtained in the above step, and thus the average influence degree of an environmental data on the monitoring data of a to-be-analyzed species in all initial clusters can be calculated, and then the average influence degree of each environmental data on each to-be-analyzed species in all initial clusters is obtained.

[0041] The calculation model of the average influence degree of an environmental data on a to-be-analyzed species is specifically: Among them, represents the average influence degree of air humidity (an environmental data) on the to-be-analyzed species A of the engine, represents the number of initial clusters (n), represents the influence degree of air humidity on the to-be-analyzed species A in the u-th initial cluster, represents the average influence degree of air humidity on the to-be-analyzed species A obtained by all initial clusters, that is, the average influence degree of air humidity on the to-be-analyzed species A of the engine. Thus, the average influence degree of each environmental data on each to-be-analyzed species can be obtained

[0042] represents the average influence degree of the s-th environmental data on the t-th to-be-analyzed species.

[0043] Step S4: obtaining the influence degree of the to-be-analyzed species on the initial cluster center at a time based on the average influence degree of each environmental data on the to-be-analyzed species, the environmental data at a time in the history and the current time; updating the initial cluster center based on the influence degree of each to-be-analyzed species on the initial cluster center at each time in each initial cluster to obtain a final cluster.

[0044] After obtaining the average influence degree of each environmental data on each to-be-analyzed species by the above method, the influence degree of each to-be-analyzed species on the initial cluster center at each time can be calculated according to the similarity of the environment at each time in the history and the current time, and in combination with the average influence degree of the environmental factor on the to-be-analyzed species of the engine.

[0045] obtaining the influence degree of the to-be-analyzed species on the initial cluster center at a time based on the average influence degree of each environmental data on the to-be-analyzed species, the environmental data at a time in the history and the current time.

[0046] ​​​Specifically, the absolute value of the difference between the environmental data of a time in history and the environmental data of the current time is obtained, added to a hyperparameter, and then the reciprocal is obtained to obtain the environmental similarity; the environmental similarity corresponding to the time in history is multiplied by the average influence degree of the environmental data on a to-be-analyzed category to obtain a first eigenvalue corresponding to the environmental data of the time on the initial cluster center of the to-be-analyzed category; and the first eigenvalues corresponding to each environmental data of the time are summed to obtain the influence degree of the to-be-analyzed category on the initial cluster center at the time.

[0047] The specific calculation model is as follows: , wherein, represents the influence degree of the i th to-be-analyzed category at the i th time in history on the initial cluster center, represents the number of categories of environmental data, , represents the value of the s th environmental data at the i th time in history, represents the value of the s th environmental data at the current time, represents the similarity between the historical environmental factor and the current environmental factor, that is, the environmental similarity, the greater the value, the closer the historical environmental condition and the current environmental condition, that is, the greater the ability to reflect the relationship between the current engine working condition and the PID control parameter, and ε is a hyperparameter for ensuring that the fraction is meaningful, and herein ε = 0.01 is specified. represents the average influence degree of the s th environmental data on the t th to-be-analyzed category, represents the first eigenvalue corresponding to the s th environmental data at the i th time, and the sum is obtained to obtain the influence degree of the t th to-be-analyzed category at the i th time on the initial cluster center.

[0048] Further, the initial cluster center needs to be updated based on the influence degree of each to-be-analyzed category at each time on the initial cluster center in each initial cluster to obtain a final cluster.

[0049] Specifically, first, the influence weight of each to-be-analyzed category on the initial cluster center of an initial cluster at each time (each data point) needs to be obtained in the initial cluster.

[0050] Specifically, the influence degree of a to-be-analyzed category in an initial cluster on the initial cluster center of the initial cluster at a time is compared with the sum of the influence degrees of the to-be-analyzed category on the initial cluster center of the initial cluster at all times in the initial cluster to obtain the influence weight of the to-be-analyzed category on the initial cluster center of the initial cluster at the time.

[0051] The specific calculation model of the influence weight is as follows: , wherein, represents the influence weight of the tth analyzed species at the ith time in the u initial cluster on the initial cluster center of the initial cluster, and represents the influence degree of the tth analyzed species at the ith time on the initial cluster center in history (at the ith time in the u initial cluster), represents the number of data points contained in the u cluster, which can also be said to be the number of times contained, represents the sum of the influence degree of the tth analyzed species at each time in the u initial cluster on the initial cluster center.

[0052] Finally, the monitoring data of each time of the analyzed species is weighted and averaged by using the influence weight of each time of the analyzed species on the initial cluster center of the initial cluster, to obtain the average monitoring data value corresponding to the analyzed species; similarly, the average monitoring data value corresponding to other analyzed species is obtained, and the data points corresponding to the average monitoring data values of all analyzed species in the initial cluster form a primary updated initial cluster center corresponding to the initial cluster center of the initial cluster; clustering is performed according to the primary updated initial cluster center corresponding to each initial cluster center of each initial cluster to obtain a primary clustering result, and the process is repeated to update each primary updated initial cluster center in the primary clustering result until the data points in the cluster in the clustering result no longer change, and the updating is stopped to obtain a final cluster. The engine operating conditions of the data points at each time in the final cluster obtained are similar. Each final cluster obtained represents an operating condition of the engine.

[0053] Step S5, obtaining the adjustment effect after each instruction is issued in history; obtaining the average adjustment effect of a final cluster according to the adjustment effect after each instruction is issued in the final cluster; adjusting the integral gain parameter according to the average adjustment effect of the final cluster to which the current time belongs.

[0054] The degree of closeness between the stable value and the target value of the engine corresponding data when the engine corresponding data reaches stability after the engine servo system controller receives each instruction reflects the effect of PID control. Therefore, the difference between the stable value and the set target value of the engine corresponding parameter (such as output power, torque and speed, etc.) that needs to be controlled after reaching stability is obtained, which represents the adjustment effect after each instruction is issued in history. Specifically, the parameter that needs to be controlled of the engine is denoted as a control parameter, and the difference between the stable control parameter of the engine after each instruction is issued in history and the set target control parameter is obtained as the adjustment effect after each instruction is issued.

[0055] The specific calculation model of the adjustment effect is: , wherein, represents the adjustment effect after the cth instruction is issued in history, represents the value of the corresponding control parameter of the engine after the cth instruction is issued in history, that is, the stable control parameter, represents the target value set for the corresponding control parameter of the engine, that is, the target control parameter, and the closer the difference is to 0, the better the adjustment effect of the control parameter of the engine after the cth instruction is issued.

[0056] Further, the adjustment effect after the corresponding instruction issuing time in each final cluster needs to be analyzed. Further, the instruction issuing time in each final cluster is obtained according to the corresponding time when the instruction is issued in history; the instruction issuing time is also the time when the data points in a final cluster correspond to the time when the instruction is issued in history.

[0057] Thus, the PID control effect under each engine operating condition is obtained according to the average value of the adjustment effect after each instruction is issued in each final cluster. Since the cluster center of each final cluster corresponds to the engine data most consistent with the engine operating condition characteristics of the current engine environment, the data points closer to the cluster center reflect the PID control effect under the condition closer to the current time, that is, the weight in the calculation of the average value is greater, and here the weight is represented by represents the reciprocal of the distance between the data point corresponding to the cth instruction issuing time in the uth final cluster and the cluster center of the uth final cluster.

[0058] Specifically, the reciprocal of the distance between the data point corresponding to the instruction issuing time in a final cluster and the cluster center of the final cluster is obtained, denoted as the weight of the data point corresponding to the instruction issuing time; the weighted average of the adjustment effect after each instruction is issued in the final cluster is obtained by using the weight of the data point corresponding to each instruction issuing time, and the weighted average value is denoted as the average adjustment effect of the final cluster. The adjustment effect after each instruction can also be regarded as the adjustment effect corresponding to each instruction issuing time.

[0059] The specific calculation model of the average adjustment effect of a final cluster is specifically: , wherein, represents the average adjustment effect of the uth final cluster, that is, the weighted average value of the difference between the actual value after the instruction adjustment in the uth final cluster and the set target value, represents the number of instruction issuing times in the uth final cluster, represents the adjustment effect after the cth instruction is issued in the uth final cluster, represents the reciprocal of the distance between the data point corresponding to the cth instruction issuing moment in the u th final cluster and the cluster center of the final cluster, that is, the weight of the data point corresponding to the instruction issuing moment. The greater the value, the closer the environmental condition corresponding to the cth instruction issuing moment is to the environmental condition at the current moment, so the greater the weight. represents the sum of the weights of the data points corresponding to the instruction issuing moment in the final cluster. Thus, the difference between the PID control result and the set target value after the instruction issuing, that is, the adjustment effect, under each working condition can be obtained.

[0060] Further, the integral gain parameter of the PID algorithm is adjusted according to the average adjustment effect of the final cluster to which the current moment belongs. Specifically, a data point composed of the monitoring data of each to-be-analyzed category at the current moment is obtained, denoted as a target data point; the distances between the target data point and the cluster centers of the final clusters are calculated respectively to obtain the final cluster with the smallest distance as the cluster of the target data point, that is, the final cluster to which the current moment belongs.

[0061] After obtaining the final cluster to which the current moment belongs, the average adjustment effect of the final cluster to which the current moment belongs under the current working condition is obtained . Because the integral gain parameter in the PID algorithm is the size of the stable value reached after the engine servo system responds, the integral gain parameter value of the PID algorithm at the current moment is adjusted according to the difference between the PID control result and the set target value under the current working condition .

[0062] The proportional coefficient is obtained by comparing the average adjustment effect of the final cluster to which the current moment belongs with the maximum value of the absolute values of the average adjustment effects of the final clusters; the proportional coefficient is added to the first preset value and multiplied by the original integral gain parameter to obtain the adjusted gain integral parameter at the current moment.

[0063] The calculation model of the adjusted gain integral parameter is specifically: , wherein, represents the adjusted gain integral parameter at the current moment, represents the average adjustment effect of the final cluster to which the current moment belongs, represents the maximum value of the absolute values of the average adjustment effects of the final clusters, and max represents a maximum value operation, represents the proportional coefficient, which is used to control the adjustment degree of the integral gain parameter to avoid system instability caused by too large adjustment degree, represents the original PID integral gain parameter value. Thus, the adjusted gain integral parameter at the current moment can be obtained.

[0064] To sum up, the application analyzes the influence relationship between PID control algorithm parameters and control effects under different working conditions according to engine historical data, so as to realize self-adaptive adjustment of PID parameters according to engine working conditions, and improve the stability and efficiency of engine servo system control.

[0065] It should be noted that the above-mentioned sequence of the embodiments of the application is only for description, and does not represent the advantages and disadvantages of the embodiments. The above describes specific embodiments of the present application. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are possible or can be advantageous.

[0066] Each embodiment in the specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other. Each embodiment focuses on the difference from other embodiments.

[0067] The above only describes the preferred embodiments of the application, and does not limit the application. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the application shall be included in the protection scope of the application.

Claims

1. A controller optimization method for an engine servo system, characterized in that, The method includes: Acquire different types of monitoring data and environmental data from historical engine operation; filter the types of monitoring data to obtain the types to be analyzed, and cluster the monitoring data of the types to be analyzed to obtain the initial clusters and initial cluster centers; The monitoring data segments are obtained by dividing the data into segments based on the degree of change of each monitoring data of a species to be analyzed in an initial cluster; the degree of influence of the environmental data corresponding to each monitoring data segment in time series is obtained based on the environmental data of that species to be analyzed. The average influence of a particular type of environmental data on a particular species can be obtained by calculating the mean value of the influence of a particular type of environmental data on a species to be analyzed in each initial cluster. Based on the average impact of each type of environmental data on a given category, and various environmental data from a historical moment and the current moment, the impact of that category on the initial cluster center at that moment is obtained; the initial cluster centers are updated based on the impact of each category on the initial cluster center at each moment in each initial cluster, resulting in the final cluster; Obtain the adjustment effect after each command was issued in history; obtain the average adjustment effect of a final cluster based on the adjustment effect after each command was issued in the final cluster; adjust the integral gain parameter based on the average adjustment effect of the final cluster to which the current moment belongs.

2. The controller optimization method for an engine servo system according to claim 1, characterized in that, The filtering of monitoring data types to obtain the types to be analyzed includes: Principal component analysis was used to process different types of monitoring data from engine operation in history, and the characteristic values ​​of each type of monitoring data were obtained. A set number of types with the largest characteristic values ​​were selected as the types to be analyzed.

3. The controller optimization processing method for an engine servo system according to claim 1, characterized in that, The process of segmenting and obtaining monitoring data segments based on the degree of change in each monitoring data of a species to be analyzed within an initial cluster includes: Arrange the monitoring data of a certain category to be analyzed in an initial cluster in ascending order to obtain the permutation sequence; obtain the difference between the next monitoring data and the previous data in two adjacent monitoring data in the permutation sequence and normalize it to obtain the degree of change of the next monitoring data; take the monitoring data with the degree of change greater than the first threshold as the dividing point, and use the dividing point to segment the permutation sequence to obtain the monitoring data segment.

4. The controller optimization processing method for an engine servo system according to claim 1, characterized in that, The method of obtaining the degree of influence of each type of environmental data on the category to be analyzed based on the time-series corresponding environmental data for each monitoring data segment includes: Obtain the mean value of a type of environmental data corresponding to a monitoring data segment in time series, as the environmental mean value corresponding to that monitoring data segment; calculate the absolute value of the sum of the differences between the environmental means values ​​corresponding to every two adjacent monitoring data segments, and take the mean value to obtain the average change of environmental data; calculate the reciprocal of the mean value of the standard deviation of the type of environmental data corresponding to each monitoring data segment in time series, and record it as the stationary quantity of environmental data; multiply the average change of environmental data by the stationary quantity of environmental data and normalize to obtain the degree of influence of the type of environmental data on the type to be analyzed.

5. The controller optimization processing method for an engine servo system according to claim 1, characterized in that, The method of obtaining the influence of each type of environmental data on the initial cluster center at a given moment based on the average influence of each type of environmental data on a target species, and various environmental data at a historical moment and the current moment, includes: Obtain the absolute value of the difference between environmental data at a historical moment and the environmental data at the current moment, add it to the hyperparameter, and then calculate the inverse to obtain the environmental similarity. Multiply the environmental similarity corresponding to that historical moment by the average influence of that environmental data on a category to be analyzed to obtain the first feature value corresponding to that environmental data at that moment. Sum the first feature values ​​corresponding to each type of environmental data at that moment to obtain the influence of that category to be analyzed on the initial cluster center at that moment.

6. The controller optimization processing method for an engine servo system according to claim 1, characterized in that, The process of updating the initial cluster centers based on the influence of each type to be analyzed at each time point on the initial cluster centers, to obtain the final clusters, includes: The influence weight of a species to be analyzed on the initial cluster center at a given time is obtained by comparing the influence of a species to be analyzed on the initial cluster center at a given time within an initial cluster with the sum of the influence of the species to be analyzed on the initial cluster center at all times within the initial cluster. The average monitoring data of a species to be analyzed at each time step is weighted and averaged to obtain the average monitoring data value corresponding to that species. Similarly, the average monitoring data values ​​corresponding to other species to be analyzed are obtained. The data points composed of the average monitoring data values ​​of all species to be analyzed in the initial cluster are the first-updated initial cluster centers corresponding to the initial cluster centers of the initial cluster. Clustering is performed based on the first-updated initial cluster centers corresponding to the initial cluster centers of each initial cluster to obtain the first clustering result. This process is repeated to update the first-updated initial cluster centers in the first clustering result until the data points in the clusters no longer change, at which point the update stops, and the final cluster is obtained.

7. The controller optimization processing method for an engine servo system according to claim 1, characterized in that, The process of obtaining the adjustment effect after each instruction was issued in history includes: The parameters that need to be controlled for the engine are recorded as control parameters. The difference between the stable control parameters after the engine control parameters have stabilized after each command is issued in history and the set target control parameters is obtained as the adjustment effect after each command is issued.

8. The controller optimization processing method for an engine servo system according to claim 1, characterized in that, The step of obtaining the average adjustment effect of a final cluster based on the adjustment effect after each instruction is issued in the final cluster includes: Based on the historical times when instructions were issued, the time of instruction issuance in each final cluster is searched to obtain the instruction issuance time in each final cluster; the reciprocal of the distance between the data point corresponding to the instruction issuance time in a final cluster and the cluster center of the final cluster is obtained, and it is recorded as the weight of the data point corresponding to the instruction issuance time; the adjustment effect after each instruction issuance in the final cluster is weighted and averaged using the weight of the data point corresponding to each instruction issuance time, and the weighted average value is recorded as the average adjustment effect of the final cluster.

9. The controller optimization processing method for an engine servo system according to claim 1, characterized in that, The adjustment of the integral gain parameter based on the average adjustment effect of the final cluster to which it belongs at the current time includes: The data points consisting of monitoring data of each type to be analyzed at the current time are recorded as target data points. The distance between the target data points and the cluster centers of each final cluster is calculated, and the final cluster with the smallest distance is the final cluster to which the current time belongs. The average adjustment effect of the final cluster to which the current time belongs is compared with the maximum absolute value of the average adjustment effect of each final cluster to obtain the proportional coefficient. The proportional coefficient is added to the first preset value and multiplied by the original integral gain parameter to obtain the adjusted integral gain parameter at the current time.

Images