CN116992311B - Energy storage power supply state analysis method based on machine learning - Google Patents
Energy storage power supply state analysis method based on machine learning Download PDFInfo
- Publication number
- CN116992311B CN116992311B CN202311244772.5A CN202311244772A CN116992311B CN 116992311 B CN116992311 B CN 116992311B CN 202311244772 A CN202311244772 A CN 202311244772A CN 116992311 B CN116992311 B CN 116992311B
- Authority
- CN
- China
- Prior art keywords
- capacity
- sequence
- time point
- time
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000004146 energy storage Methods 0.000 title claims abstract description 75
- 238000004458 analytical method Methods 0.000 title claims abstract description 26
- 238000010801 machine learning Methods 0.000 title claims abstract description 19
- 230000008929 regeneration Effects 0.000 claims abstract description 122
- 238000011069 regeneration method Methods 0.000 claims abstract description 122
- 230000036541 health Effects 0.000 claims abstract description 16
- 238000001514 detection method Methods 0.000 claims abstract description 15
- 238000005457 optimization Methods 0.000 claims abstract description 10
- 238000012545 processing Methods 0.000 claims abstract description 6
- 238000000034 method Methods 0.000 claims description 58
- 230000006870 function Effects 0.000 claims description 20
- 238000010606 normalization Methods 0.000 claims description 15
- 230000001174 ascending effect Effects 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000010586 diagram Methods 0.000 claims description 4
- 230000010354 integration Effects 0.000 claims description 2
- 230000001172 regenerating effect Effects 0.000 claims description 2
- 230000008569 process Effects 0.000 description 15
- 230000008859 change Effects 0.000 description 11
- 238000006731 degradation reaction Methods 0.000 description 7
- 238000011156 evaluation Methods 0.000 description 5
- 230000015556 catabolic process Effects 0.000 description 4
- YHXISWVBGDMDLQ-UHFFFAOYSA-N moclobemide Chemical compound C1=CC(Cl)=CC=C1C(=O)NCCN1CCOCC1 YHXISWVBGDMDLQ-UHFFFAOYSA-N 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 2
- 239000002245 particle Substances 0.000 description 2
- WHXSMMKQMYFTQS-UHFFFAOYSA-N Lithium Chemical compound [Li] WHXSMMKQMYFTQS-UHFFFAOYSA-N 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000006227 byproduct Substances 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000007599 discharging Methods 0.000 description 1
- 150000002500 ions Chemical class 0.000 description 1
- 229910052744 lithium Inorganic materials 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 239000000047 product Substances 0.000 description 1
- 238000002759 z-score normalization Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/40—Testing power supplies
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/27—Regression, e.g. linear or logistic regression
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention relates to the technical field of digital data processing, and provides an energy storage power supply state analysis method based on machine learning, which comprises the following steps: acquiring power supply detection data; obtaining a fitting equation of the power supply capacity according to an optimization algorithm; acquiring a capacity residual sequence and a capacity first-order difference sequence according to the real capacity sequence and the capacity fitting sequence; acquiring local capacity fluctuation between two time points according to the window capacity sequence; obtaining a symbol value according to the difference of each element value in the capacity first-order differential sequence; acquiring capacity regeneration matching degree of each time point according to the capacity local fluctuation and the symbol value; acquiring the capacity regeneration quantity of the capacity regeneration point according to the capacity regeneration matching degree; and acquiring a capacity predicted value of each time point according to the capacity regeneration quantity, and acquiring an analysis result of the health state of the energy storage power supply. According to the invention, the predicted value of the time point when the capacity regeneration phenomenon occurs is corrected by utilizing the capacity regeneration matching degree, so that the prediction accuracy of the health state at different time points is improved.
Description
Technical Field
The invention relates to the technical field of digital data processing, in particular to an energy storage power supply state analysis method based on machine learning.
Background
The energy storage power supply is a device for storing electric energy, generally refers to a power supply with high battery capacity and higher charge and discharge current, has a wide application range, and occupies low positions in the fields of power systems, communication base stations, data centers, rail transit and the like. Because of the important low level of the energy storage power supply in various fields, in order to avoid the direct or indirect loss caused by the failure of the energy storage power supply, the state analysis of the energy storage power supply is very necessary, wherein the state analysis of the energy storage power supply generally refers to the process that in the working process of the energy storage power supply, data such as voltage, current and the like are collected through a sensor, the data are processed, the obtained health characteristics describe the current health state of the energy storage power supply, and the current available capacity data are usually used for describing the current health state of a lithium battery.
The change condition of the available capacity of the energy storage power supply is complex, time-varying, dynamic, nonlinear and other time sequence characteristics exist, and capacity regeneration phenomenon is caused, so that the accuracy rate of the traditional machine learning prediction model for predicting the health state of the energy storage power supply is low, and the accurate analysis result of the health state of the energy storage power supply is difficult to obtain.
Disclosure of Invention
The invention provides an energy storage power supply state analysis method based on machine learning, which aims to solve the problem that the capacity regeneration phenomenon in the existing energy storage power supply capacity changing process causes errors to the power supply capacity predicted value, and the adopted technical scheme is as follows:
one embodiment of the invention provides a machine learning-based energy storage power supply state analysis method, which comprises the following steps:
acquiring energy storage power supply detection data, wherein the energy storage power supply detection data comprises voltage data and current data; acquiring the real power capacity of each time point according to the stored energy power detection data;
obtaining an initial fitting value of each time point by using an optimization algorithm; taking a sequence formed by the real power capacities of all time points in a time window according to the ascending order of time as a window capacity sequence; acquiring local capacity fluctuation between any two elements in a time window according to the window capacity sequence;
acquiring a capacity fitting value of each time point in a time window according to a fitting equation, and acquiring a symbol value according to a feature sequence constructed by the capacity fitting values of all the time points in the time window;
acquiring capacity regeneration matching degree of each time point according to the capacity local fluctuation and the symbol value; acquiring the capacity regeneration quantity of the capacity regeneration point according to the capacity regeneration matching degree;
and acquiring a capacity predicted value of each time point according to the capacity regeneration quantity, and acquiring an analysis result of the health state of the energy storage power supply according to the capacity predicted value.
Preferably, the method for obtaining the real power capacity of each time point according to the stored energy power detection data includes:
respectively acquiring voltage normalization data and current normalization data of each time point, and taking a three-dimensional curve acquired by taking the voltage normalization data as an x-axis, the current normalization data as a y-axis and the time point as a z-axis as a discharge curve;
the integral value of each time point on the discharge curve is obtained by using a curve integration method, and the integral value of each time point is used as the real capacity of the power supply of each time point.
Preferably, the method for obtaining the initial fitting value of each time point by using the optimization algorithm comprises the following steps:
taking the real capacity of the power supply as an ordinate and taking a two-dimensional broken line obtained by taking a time point as an abscissa as a capacity broken line diagram;
obtaining a fitting function expression between the real capacity of the power supply and a time point in the capacity line diagram by using a function fitting method, and determining all constant parameters in the fitting function expression by using an optimization algorithm;
and obtaining an initial fitting value of each time point according to the fitting function expression after the constant parameters are determined.
Preferably, the method for obtaining the local fluctuation of the capacity between any two elements in the time window according to the window capacity sequence comprises the following steps:
respectively acquiring time points corresponding to two elements in a window capacity sequence, and taking the difference value between the time points as the time interval between the two elements;
taking the inverse number of the ratio of the real capacity of the power supply at each time point and the real capacity of the power supply at the previous time point in the time interval as an index, taking the calculation result taking the natural constant as the base number as the adjacent deviation amount, and taking the accumulated sum of the inverse number of the sum of the adjacent deviation amount and the preset parameter between two elements as a first accumulated value;
the local capacity fluctuation between the two elements consists of a time interval and a first accumulated value, wherein the local capacity fluctuation is in inverse proportion to the time interval, and the local capacity fluctuation is in direct proportion to the first accumulated value.
Preferably, the method for obtaining the symbol value according to the sequence constructed by the capacity fitting values of all the time points in the time window comprises the following steps:
acquiring a characteristic sequence corresponding to the time window according to the capacity fitting values of all time points in the time window;
obtaining the maximum value of elements in a capacity first-order differential sequence in a time window corresponding feature sequence taken by each time point, and the element value corresponding to each time point in the capacity first-order differential sequence, and taking the difference value between the element value and the maximum value of the elements as a first difference value of each time point;
and taking the first difference value of each time point as the input of a sign function, and acquiring the sign value of each time point according to the comparison result of the first difference value and the preset parameter.
Preferably, the method for obtaining the feature sequence corresponding to the time window according to the capacity fitting values of all the time points in the time window comprises the following steps:
taking a sequence formed by capacity fitting values of all time points in a time window according to the ascending order of time as a capacity fitting sequence;
taking a sequence formed by difference values of the same position elements in the window capacity sequence and the capacity fitting sequence according to time ascending order as a capacity residual sequence, and taking a sequence obtained by performing first-order difference processing on the capacity residual sequence as a capacity first-order difference sequence;
and taking the capacity residual error sequence and the capacity first-order difference sequence as the characteristic sequences.
Preferably, the method for obtaining the capacity regeneration matching degree of each time point according to the capacity local fluctuation and the symbol value comprises the following steps:
respectively taking a sequence formed by the real power capacities of a left time point and a right time point of a central point in a time window as a left adjacent sequence and a right adjacent sequence;
taking the product of the ratio of the element value of the central point in the capacity residual error sequence corresponding to the time window taken by each time point to the maximum value of the element in the capacity residual error sequence and the symbol value of each time point as a molecule;
and respectively acquiring the local fluctuation of the capacity between the first element and the last element in the left adjacent sequence and the right adjacent sequence, taking the sum of the local fluctuation of the capacity as a denominator, and taking the ratio of the numerator and the denominator as the capacity regeneration matching degree of each time point.
Preferably, the method for obtaining the capacity regeneration amount of the capacity regeneration point according to the capacity regeneration matching degree comprises the following steps:
acquiring a capacity regeneration point according to the capacity regeneration matching degree, and acquiring the ratio of the capacity regeneration matching degree of the capacity regeneration point to the corresponding time of the capacity regeneration point;
the capacity regeneration quantity consists of the ratio and the parameter adjusting factor, wherein the capacity regeneration quantity is in direct proportion relation with the ratio and the parameter adjusting factor.
Preferably, the method for obtaining the capacity regeneration point according to the capacity regeneration matching degree comprises the following steps:
and acquiring the capacity regeneration matching degree of each time point, and taking the time point with the capacity regeneration matching degree larger than a preset threshold value as a capacity regeneration point.
Preferably, the method for obtaining the capacity prediction value of each time point according to the capacity regeneration amount comprises the following steps:
taking a sequence consisting of the difference values of the regenerative fitting values and the power supply true values of all time points in the time window taken by each time point as the input of a prediction model, and acquiring local random fluctuation of each time point by using the prediction model;
the sum of the initial fitting value, the capacity regeneration amount, the local random fluctuation at each time point is taken as the capacity prediction value at each time point.
The beneficial effects of the invention are as follows: according to the invention, the change characteristics of the capacity of the energy storage power supply along with time are analyzed, so that the capacity change of the energy storage power supply is decomposed into a normal degradation trend part, a capacity regeneration part and a local random fluctuation part. Constructing a capacity regeneration matching degree according to the characteristics when the capacity regeneration phenomenon occurs, wherein the capacity regeneration matching degree reflects the possibility of the capacity regeneration phenomenon at each time point, and constructing a capacity regeneration amount according to the capacity regeneration matching degree, wherein the capacity regeneration amount reflects the characteristics of capacity change of the energy storage power supply caused by the capacity regeneration phenomenon; secondly, obtaining local random fluctuation by using an autoregressive moving average model ARIMA, wherein the local random fluctuation reflects the influence degree of each time point on the power supply health state of the adjacent time point; the method utilizes the capacity regeneration quantity and the local random fluctuation to comprehensively reflect the characteristic of the change of the available capacity of the energy storage power supply along with time, corrects the predicted value of the time point when the capacity regeneration phenomenon occurs, and improves the prediction accuracy of the health state at different time points.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions of the prior art, the drawings which are used in the description of the embodiments or the prior art will be briefly described, it being obvious that the drawings in the description below are only some embodiments of the invention, and that other drawings can be obtained according to these drawings without inventive faculty for a person skilled in the art.
Fig. 1 is a flowchart of a method for analyzing a state of an energy storage power supply based on machine learning according to an embodiment of the invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1, a flowchart of a method for analyzing a state of an energy storage power supply based on machine learning according to an embodiment of the invention is shown, and the method includes the following steps:
and S001, acquiring detection data of the energy storage power supply, and preprocessing the detection data.
The voltage data in the energy storage power supply is collected through a voltage sensor in the energy storage power supply, the current data in the energy storage power supply is collected through a current sensor, the collection interval is recorded as T, the experience value is usually taken as 30s, the collection times are N, and the experience value is usually taken as 900.
Because the collected data may have abnormal conditions such as missing values, the missing values need to be supplemented, common missing value filling methods include mean value filling, mode filling, nearest neighbor interpolation method, regression interpolation method and the like, in order to accurately reflect the change relation among the data, the missing values are filled by using the regression interpolation method, and meanwhile, in order to avoid influence on subsequent calculation results caused by different dimensions, the obtained data are normalized by using the Z-score method. The regression interpolation method and the Z-score normalization method are known techniques, and the specific process is not repeated.
Therefore, the normalization result of the detection data of each time point is obtained, and the analysis of the time-varying characteristics of the follow-up energy storage power supply is facilitated.
Step S002, an initial fitting value of each time point is obtained by using an optimization algorithm, and the local fluctuation of the capacity is obtained based on the capacity change of different time points in the window capacity sequence.
When the energy storage power supply is used, the available capacity of the energy storage power supply is continuously reduced along with the increase of the charge and discharge times, namely, the capacity degradation process, and in the capacity degradation process, a capacity regeneration phenomenon exists, namely, the capacity is slightly increased after a plurality of charge and discharge cycles in the capacity degradation process, namely, the capacity regeneration phenomenon, and the capacity regeneration phenomenon is the capacity regeneration phenomenon.
And fitting the normal degradation trend part of the capacity according to the available capacity data of the energy storage power supply. According to the normalization result of the current data and the normalization result of the voltage data at each time point, the three-dimensional distribution curve of the detection data of the energy storage power supply is obtained according to the normalization result of the current data and the normalization result of the voltage data at all time points, the voltage normalization data are respectively taken as the x axis of the three-dimensional curve, the current normalization data are taken as the y axis of the three-dimensional curve, the obtained three-dimensional curve is taken as the z axis of the three-dimensional curve, the discharge curve is integrated, the real power supply capacity corresponding to each time point can be obtained, namely, the data of the available capacity of the energy storage power supply changing along with time can be obtained, a corresponding capacity line graph can be made based on the data, the abscissa is taken as the time point, the real power supply capacity of the energy storage power supply can be seen, the normal degradation trend of the real capacity of the energy storage power supply presents a nonlinear trend, and is relatively close to the power function change trend, accordingly, the power supply capacity is fitted by utilizing an exponential function, and the obtained equation is as follows:
；
wherein the method comprises the steps ofInitial fitting value of the energy storage power supply capacity representing the time point t,/->、/>、/>、/>Respectively represent constant parameters corresponding to different exponentiations.
Because the fitting equation contains more parameters, the calculation efficiency is lower when each parameter in the fitting equation is calculated by using the traditional least square method, and therefore, the invention uses the least mean square error as an objective function and uses the particle swarm optimization algorithm to calculate the fitting equationThe constant parameter in (1), wherein the particle swarm optimization algorithm is a well-known technique, and the specific process is not repeated.
When the capacity regeneration phenomenon occurs, the available capacity of the energy storage power supply can be increased to a certain extent in a subsequent period of time, and after the available capacity is increased, the normal degradation trend is continuously displayed on the basis of the added value, the capacity regeneration phenomenon cannot continuously occur, and the capacity regeneration phenomenon usually occurs once and then occurs again after a plurality of periods. In the invention, a time window is formed by taking adjacent n time points around one time point as the center, the length of the time window is 2n+1, and the size of n is taken as the checked value 10. The reason for this is that in a time window of a short duration, if a capacity regeneration phenomenon occurs in the time window, there is and only one capacity regeneration phenomenon in the time window.
For a time point t, a time window is acquired by taking the time point t as a center pointRespectively obtain->The real power capacities of all time points in the range are arranged according to the ascending order of time, and the sequence obtained by the arrangement is recorded as window capacity sequence +.>Wherein->Is the real capacity of the power supply at the first point in time within the time window. Acquiring initial fitting value data of the capacity of each time point of a time window by utilizing the fitting equation, and taking a sequence formed by 2n+1 initial fitting values according to the sequence of time ascending as a capacity fitting sequence +.>Recording the sequence of window capacity->Fitting sequence to Capacity->The sequence formed by the difference values of the same position elements in the sequence is a capacity residual sequence. If the capacity regeneration phenomenon occurs at the time point t, the real capacity of the power supply at the time t should be larger than the fitting value of the fitting equation at the time t; time window->The time point adjacent to the time point t is influenced by the time point t to generate fluctuation with a certain amplitude, namely, the time window +.>There is an intra-amplitude difference between the real power supply capacity and the fitting value at a time point adjacent to the time point t.
Based on the above analysis, local capacity fluctuations are constructed here for characterizing local fluctuations between different elements within a time window, calculating the time windowLocal capacity fluctuation between inner elements s, j +.>：
；
In the method, in the process of the invention,、/>the time points of the element j and the element s in the time window are respectively, i is the ith element in the partial sequence from the element s to the element j, and the value range of i is [ s, j ]]，/>、/>Window capacity sequences +.>Parameter values corresponding to the elements i and i+1.
The larger the power supply real capacity difference between the elements i and i +1 in the window capacity sequence,the larger the value of (2), the adjacent deviation amount +.>The smaller the value of (2); first accumulated value +.>The greater the value of +.>The greater the value of (2); time interval between elements s, j +.>The smaller the value of +.>The greater the value of (2); that is, the larger the difference in the true capacity of the power supply at adjacent time points in the partial sequence between the elements s, j, the partial capacity fluctuation +.>The larger.
So far, the capacity local fluctuation between any two elements in the time window is obtained, and the assessment index of the possibility of capacity regeneration phenomenon at each time point is conveniently obtained later.
Step S003, a capacity regeneration matching degree for each time point is acquired based on the capacity local fluctuation and the symbol value, and a capacity regeneration amount is acquired based on the capacity regeneration matching degree.
For a time point t, acquiring a corresponding capacity residual sequence, performing first-order differential processing, and taking the sequence formed by the first-order differential processing as a capacity first-order differential sequence。
Further, a time window is obtained at time point tPerforming sequence division to respectively divide the left time point of the central point in the time window,The sequence consisting of the true capacity of the power supply at the right time point is taken as the left adjacent sequence +.>Right adjacent sequence->. If the time point t is the time point at which the capacity regeneration sequence takes place, the left adjacent sequence +.>Right adjacent sequence->The time points in (a) are affected by the capacity regeneration phenomenon to a similar extent, i.e. left adjacent sequence +.>Right adjacent sequence->The difference between the elements is small; besides, the time window->The difference between the power supply real capacity value of the internal capacity regeneration point and the power supply real capacity value at the previous moment is positive, and the difference is large, namely the position of the capacity regeneration point is the maximum value in the capacity residual error sequence.
Based on the analysis, a capacity regeneration matching degree M is constructed, which is used for representing the possibility of capacity regeneration phenomenon in a time window corresponding to each time point, and the capacity regeneration matching degree at the time point t is calculated：
；
；
In the method, in the process of the invention,is the sign value of time t, +.>、/>Capacity first order differential sequences>Element value, element maximum value, < > -corresponding to time point t in>Is a sign function->The function of the function input is that a symbol value corresponding to the function input is obtained according to the magnitude relation between the function input and 0, when the function input in the bracket is more than or equal to 0, the symbol value is 1, and when the function input in the bracket is less than 0, the symbol value is-1.
Is the capacity regeneration matching degree at time point t, +.>、/>Respectively, the element value corresponding to the time point t in the capacity residual sequence, the element maximum value, and the +.>Is the left adjacent sequence->Local fluctuation of capacity between the 1 st time point and the last time point,/and the like>Is the right adjacent sequence->The capacity between the 1 st time point and the last time point locally fluctuates.
The capacity regeneration matching degree reflects the probability of capacity regeneration phenomenon in the time window taken by each time point. The greater the probability of capacity regeneration at time t, the capacity first order differential sequenceThe more likely the element value at time t is the element maximum, the first difference +.>The closer the value of (2) is to 0, the sign value +.>The more likely it is 1; the larger the element value corresponding to the time point t in the capacity residual sequence, i.e. +.>Maximum value of element->The closer the (the)>The closer the value of (2) is to 1; left adjacent sequence->Right adjacent sequence->The more similar the time points of the same time interval are influenced by the time point t, +.>、/>The smaller the value of (2); i.e. < ->The larger the value of (c), the more likely the capacity regeneration phenomenon occurs at the time point t.
Further, the capacity regeneration matching degree of each time point obtained by the steps is larger than the capacity regeneration matching threshold value when the capacity regeneration matching degree of a certain time point is larger than the capacity regeneration matching threshold valueWhen the size is 0.55, the time point is defined as the capacity regeneration point, and the set of the capacity regeneration points is defined as S, the capacity regeneration amount can be constructed for the capacity regeneration point, and the constructed capacity regeneration amount is ∈>：
；
In the method, in the process of the invention,indicating the capacity regeneration of the energy storage power supply at time t, < >>The capacity regeneration matching degree at time t is shown.
In the working process of the energy storage power supply, the charge and discharge process is completed by continuously carrying out circulated physicochemical reflection of various ions, various byproducts can be generated in the process to reduce the available capacity of the energy storage power supply, when the capacity regeneration phenomenon occurs, the capacity regeneration amount gradually decreases, so that the time is longer, namely t is longer, the capacity regeneration amount is reduced along with t, the higher the capacity regeneration matching degree is, the more likely the fluctuation of the point is caused by the capacity regeneration phenomenon, and the higher the corresponding capacity regeneration amount is; and vice versa.
Thus, the capacity regeneration quantity of each capacity regeneration point is obtained according to the capacity regeneration matching degree, so that the subsequent correction of the initial fitting value is facilitated.
And S004, acquiring a health state evaluation value according to the capacity prediction value, and completing the state analysis of the energy storage power supply.
Further, a local random wave equation is constructed based on the fitting equation and the capacity regeneration amount. And taking the result of adding the fitting equation and the capacity regeneration amount as the fitting regeneration equation, wherein the function operation is a known technology, and the specific process is not repeated. According to the method, the energy storage power supply capacity regeneration fitting value of each time point is obtained according to a fitting regeneration equation, and the power supply real capacity of the energy storage power supply of all time points in a window taken by each time point is differenced from the regeneration fitting value to obtain a local random fluctuation sequence corresponding to each time point.
Taking a differential autoregressive moving average model ARIMA as a prediction model, taking a local random fluctuation sequence of all time points in a time window taken by each time point as a model input, and acquiring the local random fluctuation of each time point by using the prediction model ARIMA, wherein the ARIMA time sequence model is a known technology, and the detailed process is not repeated.
Based on the initial fitting value, the capacity regeneration amount and the local random fluctuation obtained in the steps, an available capacity change equation of the available capacity of the energy storage power supply, which changes with time, can be constructed, and the available capacity change equation can be specifically expressed as the following form:
；
in the method, in the process of the invention,predictive value representing the capacity of the energy storage power supply at point in time t, < >>Preliminary fitting value of the available capacity of the energy storage power supply representing the time point t, < >>Indicates the capacity regeneration amount at time t, +.>Representing local random fluctuations of the time point t.
The method comprises the steps of obtaining a predicted value of the energy storage power supply capacity at each time point through an available capacity change equation of the energy storage power supply, taking the ratio of the predicted value of the energy storage power supply capacity at each time point to the initial capacity of the energy storage power supply as a state evaluation index, considering the time point of which the state evaluation index is smaller than or equal to a detection standard threshold value as a time point of which the power supply health state does not reach the standard, and considering the time point of which the state evaluation index is larger than the detection standard threshold value as a time point of which the power supply health state reaches the standard.
The power health state of each time point is obtained through the state evaluation index, the stability of the energy storage power supply is analyzed through the stability of the power health state, and then stable voltage, current and the like are provided for various electric equipment, so that the various electric equipment can work stably, and meanwhile, the stable operation of the energy storage belt energy source plays an important role in improving the scheduling plan tracking capacity of new energy power generation and improving the new energy utilization rate. For example, in a wind energy storage system, if the collected wind energy overflows, the energy storage power supply is powered, namely, an energy storage process is performed; if the collected wind energy cannot meet the energy used by the plan and the energy storage power supply is in an energy sufficient state, starting the discharging operation of the energy storage power supply.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, alternatives, and improvements that fall within the spirit and scope of the invention.
Claims (10)
1. The energy storage power supply state analysis method based on machine learning is characterized by comprising the following steps of:
acquiring energy storage power supply detection data, wherein the energy storage power supply detection data comprises voltage data and current data; acquiring the real power capacity of each time point according to the stored energy power detection data;
obtaining an initial fitting value of each time point by using an optimization algorithm; taking a sequence formed by the real power capacities of all time points in a time window according to the ascending order of time as a window capacity sequence; acquiring local capacity fluctuation between any two elements in a time window according to the window capacity sequence;
acquiring a capacity fitting value of each time point in a time window according to a fitting equation, and acquiring a symbol value according to a feature sequence constructed by the capacity fitting values of all the time points in the time window;
acquiring capacity regeneration matching degree of each time point according to the capacity local fluctuation and the symbol value; acquiring the capacity regeneration quantity of the capacity regeneration point according to the capacity regeneration matching degree;
and acquiring a capacity predicted value of each time point according to the capacity regeneration quantity, and acquiring an analysis result of the health state of the energy storage power supply according to the capacity predicted value.
2. The machine learning-based energy storage power state analysis method according to claim 1, wherein the method for obtaining the real power capacity of each time point according to the energy storage power detection data comprises the following steps:
respectively acquiring voltage normalization data and current normalization data of each time point, and taking a three-dimensional curve acquired by taking the voltage normalization data as an x-axis, the current normalization data as a y-axis and the time point as a z-axis as a discharge curve;
the integral value of each time point on the discharge curve is obtained by using a curve integration method, and the integral value of each time point is used as the real capacity of the power supply of each time point.
3. The machine learning based energy storage power state analysis method according to claim 1, wherein the method for obtaining the initial fitting value of each time point by using the optimization algorithm is as follows:
taking the real capacity of the power supply as an ordinate and taking a two-dimensional broken line obtained by taking a time point as an abscissa as a capacity broken line diagram;
obtaining a fitting function expression between the real capacity of the power supply and a time point in the capacity line diagram by using a function fitting method, and determining all constant parameters in the fitting function expression by using an optimization algorithm;
and obtaining an initial fitting value of each time point according to the fitting function expression after the constant parameters are determined.
4. The machine learning-based energy storage power state analysis method according to claim 1, wherein the method for acquiring the local fluctuation of the capacity between any two elements in the time window according to the window capacity sequence is as follows:
respectively acquiring time points corresponding to two elements in a window capacity sequence, and taking the difference value between the time points as the time interval between the two elements;
taking the inverse number of the ratio of the real capacity of the power supply at each time point and the real capacity of the power supply at the previous time point in the time interval as an index, taking the calculation result taking the natural constant as the base number as the adjacent deviation amount, and taking the accumulated sum of the inverse number of the sum of the adjacent deviation amount and the preset parameter between two elements as a first accumulated value;
the local capacity fluctuation between the two elements consists of a time interval and a first accumulated value, wherein the local capacity fluctuation is in inverse proportion to the time interval, and the local capacity fluctuation is in direct proportion to the first accumulated value.
5. The machine learning-based energy storage power state analysis method according to claim 1, wherein the method for obtaining the symbol value according to the sequence constructed by the capacity fitting values of all time points in the time window is as follows:
acquiring a characteristic sequence corresponding to the time window according to the capacity fitting values of all time points in the time window;
obtaining the maximum value of elements in a capacity first-order differential sequence in a time window corresponding feature sequence taken by each time point, and the element value corresponding to each time point in the capacity first-order differential sequence, and taking the difference value between the element value and the maximum value of the elements as a first difference value of each time point;
and taking the first difference value of each time point as the input of a sign function, and acquiring the sign value of each time point according to the comparison result of the first difference value and the preset parameter.
6. The machine learning-based energy storage power state analysis method according to claim 5, wherein the method for obtaining the feature sequence corresponding to the time window according to the capacity fitting values of all time points in the time window is as follows:
taking a sequence formed by capacity fitting values of all time points in a time window according to the ascending order of time as a capacity fitting sequence;
taking a sequence formed by difference values of the same position elements in the window capacity sequence and the capacity fitting sequence according to time ascending order as a capacity residual sequence, and taking a sequence obtained by performing first-order difference processing on the capacity residual sequence as a capacity first-order difference sequence;
and taking the capacity residual error sequence and the capacity first-order difference sequence as the characteristic sequences.
7. The machine learning-based energy storage power state analysis method according to claim 1, wherein the method for obtaining the capacity regeneration matching degree of each time point according to the capacity local fluctuation and the sign value is as follows:
respectively taking a sequence formed by the real power capacities of a left time point and a right time point of a central point in a time window as a left adjacent sequence and a right adjacent sequence;
taking the product of the ratio of the element value of the central point in the capacity residual error sequence corresponding to the time window taken by each time point to the maximum value of the element in the capacity residual error sequence and the symbol value of each time point as a molecule;
and respectively acquiring the local fluctuation of the capacity between the first element and the last element in the left adjacent sequence and the right adjacent sequence, taking the sum of the local fluctuation of the capacity as a denominator, and taking the ratio of the numerator and the denominator as the capacity regeneration matching degree of each time point.
8. The machine learning-based energy storage power state analysis method according to claim 1, wherein the method for acquiring the capacity regeneration amount of the capacity regeneration point according to the capacity regeneration matching degree is as follows:
acquiring a capacity regeneration point according to the capacity regeneration matching degree, and acquiring the ratio of the capacity regeneration matching degree of the capacity regeneration point to the corresponding time of the capacity regeneration point;
the capacity regeneration quantity consists of the ratio and the parameter adjusting factor, wherein the capacity regeneration quantity is in direct proportion relation with the ratio and the parameter adjusting factor.
9. The machine learning-based energy storage power state analysis method according to claim 8, wherein the method for acquiring the capacity regeneration point according to the capacity regeneration matching degree is as follows:
and acquiring the capacity regeneration matching degree of each time point, and taking the time point with the capacity regeneration matching degree larger than a preset threshold value as a capacity regeneration point.
10. The machine learning-based energy storage power state analysis method according to claim 1, wherein the method for obtaining the capacity prediction value of each time point according to the capacity regeneration amount is as follows:
taking a sequence consisting of the difference values of the regenerative fitting values and the power supply true values of all time points in the time window taken by each time point as the input of a prediction model, and acquiring local random fluctuation of each time point by using the prediction model;
the sum of the initial fitting value, the capacity regeneration amount, the local random fluctuation at each time point is taken as the capacity prediction value at each time point.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311244772.5A CN116992311B (en) | 2023-09-26 | 2023-09-26 | Energy storage power supply state analysis method based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311244772.5A CN116992311B (en) | 2023-09-26 | 2023-09-26 | Energy storage power supply state analysis method based on machine learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116992311A CN116992311A (en) | 2023-11-03 |
| CN116992311B true CN116992311B (en) | 2023-12-19 |
Family
ID=88523435
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311244772.5A Active CN116992311B (en) | 2023-09-26 | 2023-09-26 | Energy storage power supply state analysis method based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116992311B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117791626B (en) * | 2024-02-23 | 2024-04-30 | 广东佰林电气设备厂有限公司 | Intelligent comprehensive power box power supply optimization method |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110221225A (en) * | 2019-07-08 | 2019-09-10 | 中国人民解放军国防科技大学 | Spacecraft lithium ion battery cycle life prediction method |
| JP2020060404A (en) * | 2018-10-09 | 2020-04-16 | 株式会社豊田中央研究所 | Full charge capacity estimation device of power supply |
| DE102019007510A1 (en) * | 2019-10-28 | 2020-07-16 | Daimler Ag | Method for determining a state of health of an electrical energy store by means of a stored capacity model depending on an anode overhang effect, and electronic computing device |
| CN113075574A (en) * | 2021-03-30 | 2021-07-06 | 上海交通大学 | Battery health state prediction method and equipment based on self-adaptive information fusion |
| CN114545270A (en) * | 2022-01-13 | 2022-05-27 | 上海电力大学 | Lithium battery health state estimation method considering health factor delay |
| CN114660497A (en) * | 2022-03-17 | 2022-06-24 | 哈尔滨理工大学 | Lithium ion battery service life prediction method aiming at capacity regeneration phenomenon |
| CN115508706A (en) * | 2022-09-22 | 2022-12-23 | 电子科技大学 | Lithium battery residual life prediction method considering capacity regeneration phenomenon |
-
2023
- 2023-09-26 CN CN202311244772.5A patent/CN116992311B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2020060404A (en) * | 2018-10-09 | 2020-04-16 | 株式会社豊田中央研究所 | Full charge capacity estimation device of power supply |
| CN110221225A (en) * | 2019-07-08 | 2019-09-10 | 中国人民解放军国防科技大学 | Spacecraft lithium ion battery cycle life prediction method |
| DE102019007510A1 (en) * | 2019-10-28 | 2020-07-16 | Daimler Ag | Method for determining a state of health of an electrical energy store by means of a stored capacity model depending on an anode overhang effect, and electronic computing device |
| CN113075574A (en) * | 2021-03-30 | 2021-07-06 | 上海交通大学 | Battery health state prediction method and equipment based on self-adaptive information fusion |
| CN114545270A (en) * | 2022-01-13 | 2022-05-27 | 上海电力大学 | Lithium battery health state estimation method considering health factor delay |
| CN114660497A (en) * | 2022-03-17 | 2022-06-24 | 哈尔滨理工大学 | Lithium ion battery service life prediction method aiming at capacity regeneration phenomenon |
| CN115508706A (en) * | 2022-09-22 | 2022-12-23 | 电子科技大学 | Lithium battery residual life prediction method considering capacity regeneration phenomenon |
Non-Patent Citations (1)
| Title |
|---|
| 基于变分模态分解和高斯过程回归的锂离子电池剩余寿命预测方法;吴祎 等;计算机与现代化(第02期);第83-88页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116992311A (en) | 2023-11-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111398833B (en) | Battery health state assessment method | |
| CN112540317B (en) | Battery health state estimation and residual life prediction method based on real vehicle data | |
| CN116992311B (en) | Energy storage power supply state analysis method based on machine learning | |
| CN111695301A (en) | Method and device for predicting battery charge state, storage medium and equipment | |
| CN114372417A (en) | Electric vehicle battery health state and remaining life evaluation method based on charging network | |
| CN113219343A (en) | Lithium battery health state prediction method, system, equipment and medium based on elastic network | |
| CN110941929A (en) | Battery health state assessment method based on ARMA and Elman neural network combined modeling | |
| CN115902647B (en) | Intelligent battery state monitoring method | |
| CN112068003B (en) | Method and device for predicting service life of cadmium-nickel storage battery based on linear wiener process | |
| JP6305988B2 (en) | Device and method for determining energy status based on data derived from processing methods | |
| CN110443377B (en) | Support vector machine storage battery life prediction method based on immune algorithm optimization | |
| CN113836756B (en) | Intelligent monitoring method and system for annealing process of three-dimensional wound core transformer | |
| CN114740388A (en) | Lithium battery residual life state evaluation method based on improved TCN | |
| CN116430817B (en) | Data acquisition processing method and system applied to photovoltaic module production system | |
| CN116031888A (en) | Dynamic load prediction-based power flow optimization method, system and storage medium | |
| CN117554846B (en) | Lithium battery life prediction method and system considering constraint conditions | |
| CN117783745A (en) | Data online monitoring method and system for battery replacement cabinet | |
| CN114594398A (en) | Energy storage lithium ion battery data preprocessing method | |
| CN112287597B (en) | Lead-acid storage battery SOH estimation method based on VPGA-GPR algorithm | |
| CN116743180A (en) | Intelligent storage method for energy storage power supply data | |
| CN116953547A (en) | Energy storage battery health evaluation method, system, equipment and storage medium | |
| CN112630658A (en) | Method and system for evaluating service life loss of storage battery | |
| CN117110884A (en) | Lithium battery remaining service life prediction method based on multi-core correlation vector machine | |
| CN115219907A (en) | Lithium battery SOC estimation method, system, medium, equipment and terminal | |
| CN116754960A (en) | Cloud SOH estimation and residual life prediction method, device and program for sodium ion battery |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |