The invention content is as follows:
in order to simplify the energy storage battery state evaluation process, reduce the influence of human factors, strengthen the complete description of the possible running state of the energy storage battery to be detected and improve the robustness of the identification technology of the possible running state of the energy storage battery to be detected, the invention provides an energy storage battery state judgment method based on mutual information. The technical scheme adopted by the invention is as follows:
a mutual information-based energy storage battery state judgment method comprises the following steps:
step 1: preprocessing the terminal voltage of the energy storage battery in different historical operating states; the method comprises the following specific steps:
step 1.1: the method comprises the steps that terminal voltage signals of m energy storage batteries in a one-time charging and discharging test process are collected by using measuring equipment, the m energy storage batteries are divided into a normal type and a fault type, and the fault type is divided into a specific type in S-1, so that the m energy storage batteries comprise S types, and the type L of each signal is markeds,s=1,2,…,S;
Step 1.2: performing first-order difference preprocessing on the m historical terminal voltage data, and combining the operation state class L of the energy storage batterysForming a data sample set A, setting the number R of extracted sample subsets, the capacity Q of the subsets and a comparison threshold value delta, and enabling R to be 1;
step 2: generating an energy storage battery state judgment evidence; randomly sampling the data sample set A to form a subset B with the capacity of QrR is 1,2, …, R, and the sampling process guarantees the subset BrThe middle samples are mutually exclusive;
and step 3: acquiring a terminal voltage signal of an energy storage battery to be detected in a primary charging and discharging test process by using measuring equipment, calculating a first-order difference of the terminal voltage, defining the first-order difference to be represented by a random variable Y, and setting j to be 1;
and 4, step 4: calculating to-be-detected energy storage battery and subset B
rThe mutual information between each sample in the group; assuming that random variable Y and subset B represented by first-order differential data of terminal voltage of energy storage battery to be detected
rThe sample of the jth in the sequence represents a random variable of
And p (y) represents a random variable
And the edge probability distribution function of Y,
representing random variables
And Y, the mutual information calculation formula is as follows:
and 5: judging whether j is greater than or equal to Q, if so, entering a step 6, otherwise, j is j +1, and returning to the step 4;
step 6: sum of calculated mutual information values
Accumulating the mutual information values from large to small in sequence until the ratio of the accumulated result to the sum of the mutual information values is greater than or equal to a set threshold value delta;
and 7: statistical subset B
rThe number of different types corresponding to the medium history energy storage battery samples
And the frequency of the different types of samples participating in the accumulation process
The predicate evidence vector for this subset is computed as:
and 8: forming a fused final energy storage battery state judgment result; judging subset BrAnd if the number of the energy storage batteries reaches a set value, namely whether R is smaller than R, making R be R +1 and returning to the step 2 to continue generating judgment evidence vectors, if not, calculating the average value of the judgment evidence vectors generated by the R difference subsets to form a final judgment evidence vector, judging the running state type of the energy storage battery to be detected according to the numerical condition in the final judgment evidence vector, returning to the final judgment evidence vector, and finishing the judgment process of the running state of the energy storage battery to be detected, wherein the energy storage battery represented by the maximum value in the final judgment evidence vector is normal or the fault type is a final judgment result.
The first preferred scheme is as follows: the step 4 represents a random variable
And the marginal probability distribution function of Y
The distribution calculation process for p (y) is as follows:
step 4.1: taking the maximum value y of the first-order differential signal of the terminal voltage of the detection batterymaxAnd the minimum value yminAnd is in [ ymin,ymax]N parts of inner equal intervals;
step 4.2: and counting the frequency of each point of the first-order difference signal of the end voltage of the energy storage battery to be detected in different intervals, and dividing the frequency by the total data of the first-order difference signal of the end voltage of the energy storage battery to be detected to obtain the probability of occurrence of each interval, and recording the probability as the edge probability distribution p (Y) of the random variable Y.
Further, the random variable in step 4
And joint probability score of YCloth function
The calculation method of (2) is as follows: acquiring a first order differential signal of battery terminal voltage and subset B
rThe frequency of the jth sample occurring in the interval divided by the marginal probability distribution function at the same time is divided by the total data of the first-order differential signals of the terminal voltage of the energy storage battery to be detected to obtain the probability of each interval, and the probability is recorded as the random variable Y and the sum
Joint probability distribution of
The preferred scheme II is as follows: the specific steps of the step 6 comprise:
step 6.1: sum of calculated mutual information values
Let j equal 1 and MidInfo equal 0;
step 6.2: sorting mutual information values from big to small
And accumulating the mutual information values in turn, i.e.
Until the ratio MidInfo/SI of the sum of the accumulated result and the mutual information value
(r)Greater than or equal to delta, record subset B
rThe type of historical energy storage battery participating in the accumulation process.
The preferable scheme is three: the specific steps of step 8 include:
step 8.1: judging subset BrIf the number reaches the set value, namely, if R is smaller than R, if R is equal to R +1, returning to the step 2 to continue generating the judgment evidence vector Evi(r)If not, go to step 8.2;
step 8.2: computing the average of the decision evidence vectors generated by the R difference subsets
Form the final decision evidence vector Evi ═ η
1,η
2,...,η
S]Wherein
Step 8.3: according to the eta in the final judgment evidence vector1,η2,...,ηSThe running state type of the energy storage battery to be detected is judged according to the numerical condition, and the normal or fault type of the energy storage battery represented by the maximum value in the final judgment evidence vector is returned as a final judgment result LsAnd arg (·) represents a function of the type of the energy storage battery represented by the returned maximum value, and the judgment process of the operation state of the energy storage battery to be detected is completed.
Compared with the closest prior art, the excellent effects of the invention are as follows:
in the technical scheme of the invention, mutual information of the first-order difference of the voltage of the energy storage battery to be detected and the first-order difference data of the voltage of the energy storage batteries in the historical records is directly calculated, and the similarity between the energy storage battery to be detected and the energy storage batteries in different states in the historical records is sequentially analyzed in an accumulated manner according to the sequence from large to small so as to form the state judgment of the energy storage battery to be detected. Compared with the traditional machine learning method which requires an artificial characteristic extraction process, the method disclosed by the invention does not need the characteristic extraction process based on mutual information calculation of different energy storage battery voltages, is beneficial to influence of artificial subjective factors, and improves the applicability of the method.
In the technical scheme of the invention, on the basis of calculating and sorting the energy storage batteries to be detected and recording the energy storage mutual information of different running states in history, the proportion of normal and different faults of the history recording energy storage batteries in the subset sample and the screened result is counted to form a diagnosis evidence. Compared with a simple method only outputting the most sample types, the method fully considers the composition conditions of all the subset samples and the historical recording energy storage batteries in the screened result, completely and accurately reflects the description of the possible running state of the batteries to be detected, and is beneficial to more comprehensively and accurately judging the running condition of the energy storage batteries to be detected.
In the technical scheme of the invention, a plurality of differentiated diagnosis evidence modes are generated by randomly generating a plurality of data subsets, so that the variety of diagnosis evidences of the energy storage battery to be detected is enriched. Compared with a single diagnosis evidence, the method disclosed by the invention has the advantages that the influence of singular samples and noise is favorably relieved by utilizing an integration process and an average value calculation method, and the robustness of a diagnosis result is improved.
The specific implementation mode is as follows:
example (b):
the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application.
A mutual information-based energy storage battery state judgment method comprises the following steps:
step 1: preprocessing the terminal voltage of the energy storage battery in different historical operating states;
step 1.1: the method comprises the steps that terminal voltage signals of m energy storage batteries in a one-time charging and discharging test process are collected by using measuring equipment, the m energy storage batteries are divided into a normal type and a fault type, and the fault type is divided into a specific type in S-1, so that the m energy storage batteries comprise S types, and the type L of each signal is markeds,s=1,2,…,S;
Step 1.2: performing first-order difference preprocessing on the m historical terminal voltage data, and combining the operation state class L of the energy storage batterysForming a data sample set A, setting the number R of extracted sample subsets, the capacity Q of the subsets and a comparison threshold value delta, and enabling R to be 1;
step 2: generating an energy storage battery state judgment evidence; randomly sampling the data sample set A to form a subset B with the capacity of Qr(R1, 2, …, R) and the sampling process guarantees the subset BrThe middle samples are mutually exclusive;
and step 3: acquiring a terminal voltage signal of an energy storage battery to be detected in a primary charging and discharging test process by using measuring equipment, calculating a first-order difference of the terminal voltage, defining the first-order difference to be represented by a random variable Y, and setting j to be 1;
and 4, step 4: define subset B
rThe jth sample in (1) is composed of random variables
Representing, calculating the random variables Y and
edge probability distribution of
And p (y); the first-order differential signal of the terminal voltage of the energy storage battery to be detected represents a random variable Y and a random variable
And the marginal probability distribution function of Y
The distribution calculation process for p (y) is as follows:
step 4.1: taking the maximum value y of the first-order differential signal of the terminal voltage of the detection batterymaxAnd the minimum value yminAnd is in [ ymin,ymax]N parts of inner equal intervals;
step 4.2: counting the frequency of each point of the first-order differential signal at the end voltage of the energy storage battery to be detected in different intervals, and dividing the frequency by the total data of the first-order differential signal at the end voltage of the energy storage battery to be detected to obtain the probability of occurrence of each interval, and recording the probability as the edge probability distribution p (Y) of the random variable Y;
step 4.3: random variable
And the joint probability distribution function of Y
The calculation method of (2) is as follows: battery terminal acquisitionVoltage first order difference signal sum subset B
rThe frequency of the jth sample occurring in the interval divided by the marginal probability distribution function at the same time is divided by the total data of the first-order differential signals of the terminal voltage of the energy storage battery to be detected to obtain the probability of each interval, and the probability is recorded as the random variable Y and the sum
Joint probability distribution of
First-order differential signal of terminal voltage of energy storage battery to be detected and subset BrThe mutual information of the jth sample in (1) is:
and 5: judging whether j is greater than or equal to Q, if so, entering a step 6, otherwise, j is j +1, and returning to the step 4;
step 6: screening out subset B based on mutual informationrThe energy storage battery sample with similar running characteristics to the energy storage battery to be detected is obtained; the method specifically comprises the following steps:
step 6.1: sum of calculated mutual information values
Let j equal 1 and MidInfo equal 0;
step 6.2: sorting mutual information values from big to small
And accumulating the mutual information values in turn, i.e.
Until the ratio MidInfo/SI of the sum of the accumulated result and the mutual information value
(r)Greater than or equal to delta, record subset B
rThe type of the historical energy storage battery participating in the accumulation process;
and 7: statistical subset B
rHistory of China storeNumber of different types of corresponding battery samples
And the frequency of the different types of samples participating in the accumulation process
Computing a predicate evidence vector for the subset
Comprises the following steps:
and 8: forming a fused final energy storage battery state judgment result; the method comprises the following specific steps:
step 8.1 judge subset BrThe number reaches a set value (namely whether R is smaller than R), if so, the R is made to be R +1, the step 2 is returned, and the judgment evidence vector Evi is continuously generated(r)If not, entering step 3.2;
step 8.2 calculate the average of the decision evidence vectors generated by the R difference subsets
Form the final decision evidence vector Evi ═ η
1,η
2,...,η
S]Wherein
According to the eta in the final judgment evidence vector
1,η
2,...,η
SThe running state type of the energy storage battery to be detected is judged according to the numerical condition, and the normal or fault type of the energy storage battery represented by the maximum value in the final judgment evidence vector is returned as a final judgment result L
sAnd arg (·) represents a function of the type of the energy storage battery represented by the returned maximum value, and the judgment process of the operation state of the energy storage battery to be detected is completed.
Finally, it should be noted that: the described embodiments are only some embodiments of the present application and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.