CN111985524A - Improved low-voltage transformer area line loss calculation method - Google Patents
Improved low-voltage transformer area line loss calculation method Download PDFInfo
- Publication number
- CN111985524A CN111985524A CN202010621395.2A CN202010621395A CN111985524A CN 111985524 A CN111985524 A CN 111985524A CN 202010621395 A CN202010621395 A CN 202010621395A CN 111985524 A CN111985524 A CN 111985524A
- Authority
- CN
- China
- Prior art keywords
- line loss
- particle
- value
- abnormal
- loss
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
Abstract
The invention discloses an improved low-voltage transformer area line loss calculation method, which is characterized in that an improved random forest algorithm is adopted to construct a prediction model of line loss, if the line loss is abnormal, an improved particle swarm optimization algorithm is adopted to position an abnormal line loss point, and an optimal abnormal point load current is searched to calculate the line loss. The method can calculate more accurate line loss.
Description
Technical Field
The invention belongs to the technical field of intelligent distribution transformer terminals, and particularly relates to an improved low-voltage transformer area line loss calculation method.
Background
The statistical line loss electricity quantity is the difference between the power supply quantity and the power sale quantity measured and counted by the electricity consumption meter. The statistical line loss rate is the percentage of the statistical line loss power and the statistical supply power. Part of the line loss electricity quantity statistics is the line loss electricity quantity statistics caused by inaccurate statistics data, such as watt-hour meter errors, meter missing, meter error and the like, no meter electricity consumption, electricity stealing, charged equipment aging and the like. This large class is called "manage statistics power". The other type is called technical line loss capacity, which is caused by the loss of each power grid element in the transmission, transformation and distribution processes.
In the prior art, when a line loss calculation model is established by using a least square support vector, parameters to be determined first include a penalty factor C and an RBF (radial basis function) kernel function parameter σ. The calculation error magnitude has a direct relation with the two parameters, but the current research cannot optimize the selection of the two parameters. Therefore, the selection of model parameters can only be done empirically to select the optimum.
Disclosure of Invention
The invention aims to overcome the defects in the prior art, provides an improved low-voltage transformer area line loss calculation method, and solves the technical problem that the penalty factor and the radial basis function kernel function parameter in the line loss calculation model established by the least square support vector in the prior art cannot be optimal.
In order to solve the technical problem, the invention provides an improved low-voltage transformer area line loss calculation method, which is characterized by comprising the following steps:
predicting real-time line loss by adopting a random forest algorithm;
calculating to obtain a real-time line loss rate according to the real-time line loss amount;
judging whether the line loss is abnormal or not according to the comparison between the real-time line loss rate and the maximum reasonable line loss rate;
if the line loss is judged to be abnormal, calculating the load current of the abnormal point;
and calculating the bus loss caused by the abnormal point according to the load current of the abnormal point.
Further, the predicting the real-time line loss amount by using the random forest algorithm comprises the following steps:
1) acquiring historical characteristic values of voltage and current, and constructing a historical sample characteristic set;
2) obtaining an association judgment matrix Z by calculating the correlation between the historical sample characteristic set and the sample characteristic set to be predicted;
3) assigning the weight of each feature to construct a feature weight matrix W;
4) performing dot product calculation on the characteristic weight matrix W and the association judgment matrix Z to obtain an association decision matrix U;
5) randomly selecting d features from the feature set to obtain a weighted voting value a of the decision tree;
6) selecting a similar historical sample feature set from the historical sample feature vector set by utilizing a U;
7) training a decision tree in the random forest by using the selected similar historical sample feature set and the corresponding fault category to obtain a trained random forest;
8) weighting the fault prediction result of each decision tree in the random forest by using the weighted voting value a of the decision tree and the associated decision matrix U, and adjusting a to obtain a final prediction model with the prediction accuracy as a target of 100%;
9) inputting the characteristic vector of the sample to be predicted into a final prediction model to obtain a final line loss prediction result;
10) updating the value of W in a self-adaptive manner by using the prediction accuracy rate of 100%;
11) and calculating a U value by using the updated W value, and adaptively updating the value of a.
Further, the maximum reasonable line loss rate is:
taking the calculation of the a-phase reasonable line loss rate of the low-voltage branch line as an example, the maximum reasonable line loss of the a-phase N users is calculated as follows:
wherein R iskRepresenting the line resistance value between the kth user and the (k-1) th user in the delta t time; pkRepresenting the active power of the kth user per minute in the delta t time period; qkRepresenting the reactive power per minute of the kth user during the at time period;
calculating the total active power consumption of the user as follows:
maximum reasonable line loss rate eta of the A phase0The method comprises the following steps:
when the low-voltage transformer area power grid normally works, the current real-time line loss rate eta is less than or equal to the maximum reasonable line loss rate.
Further, the calculating the load current at the abnormal point includes:
the total loss and the measured abnormal loss of a certain user node are minimized as an optimization target,
and calculating to obtain the optimal solution of the optimization target, namely the load current of the abnormal point.
Further, the minimizing the total loss and the measured abnormal loss of a certain user node as an optimization target includes:
taking phase a of the low-voltage branch line as an example, the measured abnormal loss when the "abnormal" power consumption exists is as follows:
wherein S isSPower at the head end of the single-phase line for abnormal use, SkThe active power of the k-th user in abnormal power utilization is obtained. The node refers to a section of line of each user, and each section of line is a node;
assuming that the abnormal point is near the ith node, the total extra loss and the abnormal electricity consumption of all lines caused by the abnormal point are as follows:
wherein
Representing the measured current I according to the impedance of each section of line experienced by k userskAnd load current I at abnormal point#Calculating the loss power of each section of line; i is#Is the load current of the abnormal point, (U)i)*Is the voltage of the anomaly;
line loss outlier localization can be described as an optimization problem: in finding an i#So that
The above is the optimization goal.
Further, the calculating to obtain the optimal solution of the optimization objective includes:
and calculating by adopting a particle swarm algorithm to obtain the optimal solution of the optimization target.
Further, the calculating by using the particle swarm algorithm to obtain the optimal solution of the optimization target includes:
1) initializing a particle swarm and randomly initializing each particle; inputting parameters of a particle swarm optimization algorithm and difference vectors of total loss and actually measured abnormal loss of nodes as parameters; taking a difference vector of the total loss of the nodes and the actually measured abnormal loss as an optimization objective function;
2) calculating the fitness value of each particle based on the fitness function; inputting the particle individuals as system variables, and generating feasible solutions based on constraint conditions;
3) comparing the current fitness value with the historical optimal fitness value aiming at the particles, and simultaneously replacing the historical optimal value;
4) comparing the current fitness with the historical optimal fitness value of the population aiming at the particles, and replacing the historical optimal value;
5) calculating the difference between the maximum fitness value and the average fitness value of each particle to obtain an individual fitness difference value, taking the individual fitness difference value as input, adaptively adjusting the speed and the position of each particle, and updating the speed and the position of the particles to obtain a progeny population;
6) dynamically adjusting individuals of the filial generation population to generate a new generation of particle swarm;
dynamically adjusting each particle of the particle swarm to generate a new generation of particle swarm; the method comprises the following steps:
introducing immigration operators, eliminating the worst individuals with a certain elimination rate in the evolution process of each generation, and then replacing the worst individuals with the generated new individuals;
filtering similar individuals, sorting the offspring individuals according to the fitness values, and sequentially calculating the difference value between the maximum fitness value and the average fitness value of each particle and the generalized Hamming distance between the similar individuals of which the difference value is smaller than a threshold delta; if the difference value is smaller than the threshold delta and the generalized Hamming distance is smaller than the threshold delta, filtering the particle;
dynamically supplementing new filial generation individuals, randomly carrying out multiple variations on a plurality of individuals with higher fitness values in the parent generation to generate new individuals, and adding filial generations to generate a new generation of particle swarm;
7) updating the speed and position of each particle in the new generation of particle swarm through the individual extreme value and the swarm extreme value; comparing the fitness value of each particle with the optimal value, if the fitness value is better, taking the fitness value as the current best position, comparing all current individual extremum values with the group extremum values, and updating the group extremum value;
updating the speed of each particle in the new generation of particle swarm by adopting a formula (14), and updating the position of each particle in the new generation of particle swarm by adopting a formula (15):
wherein the content of the first and second substances,
represents the velocity of the particle after the (k + 1) th iteration, w represents the inertia weight,
representing the velocity of the particle after the k-th iteration, c1Denotes a first learning factor constant, c2A second learning factor constant is represented as a second learning factor constant,
representing the first random number after the kth iteration,
representing the individual extremum of the particle after the k-th iteration,
indicating the position of the particle after the k-th iteration,
representing the second random number after the kth iteration,
representing the extreme value of the particle group after the kth iteration, wherein k represents the iteration times;
wherein, the position of the particle after the (k + 1) th iteration is represented;
8) and if the minimum difference value between the total extra loss of the node and the actually measured abnormal loss is obtained, ending, and recording the user node as an abnormal line loss point. Otherwise, jumping to step 5).
Compared with the prior art, the invention has the following beneficial effects: the method adopts a self-adaptive precision weighted random forest algorithm to predict the line loss, if the line loss is abnormal, a self-adaptive particle swarm optimization algorithm is used for positioning a line loss abnormal point, and the line loss is calculated according to the load current of the optimal abnormal point. The method can improve the accuracy of the line loss.
Drawings
FIG. 1 is a schematic flow chart of line loss prediction using a modified random forest algorithm;
FIG. 2 is a schematic flow chart of an automatic abnormal point positioning process using an improved particle swarm optimization algorithm;
FIG. 3 is a schematic flow chart of the method of the present invention.
Detailed Description
The invention is further described below with reference to the accompanying drawings. The following examples are only for illustrating the technical solutions of the present invention more clearly, and the protection scope of the present invention is not limited thereby.
Based on the above technical solution, an implementation flow of the improved low-voltage transformer area line loss calculation method is shown in fig. 3. The method comprises the following steps:
step 1, calculating the maximum reasonable line loss rate of a low-voltage transformer area;
take the calculation of the a-phase reasonable line loss rate of the low-voltage branch line as an example. The loss rate calculation process of the phase B and the phase C is similar. Assuming that all loads are at the tail end of a line of a certain low-voltage transformer area, calculating the maximum reasonable line loss of N A-phase users by using active power and reactive power as follows:
wherein R iskRepresenting the line resistance value between the kth user and the (k-1) th user in the delta t time; pkRepresenting the active power of the kth user per minute in the delta t time period; qkIndicating the reactive power per minute for the kth user during the at period.
Calculating the total active power consumption of the user as follows:
maximum reasonable line loss rate eta of A phase0The method comprises the following steps:
when the low-voltage transformer area power grid normally works, the current real-time line loss rate eta is less than or equal to the maximum reasonable line loss rate. That is, the reasonable line loss rate of the line is within the range of eta ≦ eta0(ii) a If η > η0The line loss is judged to be abnormal.
And 2, predicting the real-time line loss amount by adopting an improved random forest algorithm, calculating to obtain a real-time line loss rate according to the real-time line loss amount, and comparing the real-time line loss rate with the maximum reasonable line loss rate to judge whether the line loss is abnormal or not.
In the prior art, the formula is directly adopted to calculate the real-time line loss of the line
Calculating, and obtaining the current line loss rate by using the real-time line loss ratio to the head end power of the upper line, namely directly calculating to obtain the current line loss rate by the following formula:
wherein P isSRepresenting single phase line head end power.
If the current line loss rate satisfies eta > eta0If so, the line loss is determined to be abnormal.
The method directly adopts the formula to calculate larger deviation of the result, and adopts an improved random forest algorithm to predict the real-time line loss amount in order to improve the accuracy of line loss calculation.
The process of predicting the real-time line loss amount by the improved random forest algorithm is shown in fig. 1 and comprises the following steps:
1) obtaining A, B, C historical characteristic values of three-phase voltage and current, and constructing a historical sample characteristic set;
the feature vector of the historical sample at the ith moment is SiIt can be expressed as:
Si＝[si1,si2,...,sim],i＝1,2,...,n (5)
where n is the number of historical samples, m represents the number of eigenvalues contained in each sample, simRepresenting the mth eigenvalue of the historical sample at the ith time.
S for feature vector of sample to be measuredoRepresents:
So＝[so1,so2,...,som]，o＝1,2,...g (6)
wherein g is the number of samples to be measured, somAnd the mth characteristic value represents the line loss amount at the ith time of the day to be measured.
2) And calculating a correlation judgment matrix by calculating the correlation between the historical sample characteristic vector and the sample characteristic vector to be predicted.
Historical sample feature vector S at ith momentiAnd the characteristic vector S of the sample to be measuredoThe relationship between the two sequences is called the correlation coefficient, and the correlation judgment matrix is formed:
wherein Z isijJ characteristic value representing ith historical sample and sample set S to be predictedoThe correlation between the j-th row feature values, i 1, …, n, j 1, …, m; the calculation formula is as follows:
x represents the historical sample feature set in this embodiment, y represents the sample feature set under test,
the average feature vectors of the history sample x and the sample y to be measured are respectively. n represents the historical sample size and l represents the sample size to be measured. x is the number ofejIn this embodiment, the jth characteristic value, y, of the e-th history sample is shownkjRepresenting the jth characteristic value of the kth sample to be tested, wherein i is 2, n; j is 1, 2.
3) Assigning the weight of each feature, and constructing a feature weight matrix W:
W＝[W1,W2,...Wi,...,Wn]T (8)
wherein Wi＝[w1,w2,...wj...,wm]，wjThe weight of the jth characteristic value is a random value as an initial value, and the weight is more than or equal to 0 and less than or equal to 1; the weight of each sample is the same.
4) And performing dot product calculation on the characteristic weight matrix W and the association judgment matrix Z to obtain an association decision matrix U.
5) Randomly selecting d features from the feature set to obtain a weighted voting value a of the decision tree,
wherein, λ is a parameter adjusting factor, the initial value is a random value, and the random value is more than or equal to 0 and less than or equal to 1; r isjRepresenting the correlation between the j-th column of the feature vector in the training sample set and the j-th column of the feature vector in the sample set to be detected;
wherein x represents the training sample feature set in this embodiment, y represents the sample feature set to be measured,
the average feature vectors of the training sample X and the sample Y to be detected are respectively. x is the number offjIn this embodiment, the jth characteristic value, y, of the f training sample is shownkjRepresenting the jth characteristic value of the kth sample to be tested, wherein f is 1, 2. k is 1,2,. l; j is 1, 2. t represents the scale of the training sample and l represents the scale of the sample to be measured.
rjIs a known value, i.e. during subsequent adjustment of a, no new r is selected againj；
6) Selecting a similar historical sample feature set from the historical sample feature vector set by utilizing U, namely adaptively adjusting a threshold value according to the set number of similar samples, and setting a threshold value phi for each column of feature valuesq(q 1, 2.. times, m), if the element in the U { Si } matrix is larger than or equal to the set threshold value phi #qI.e. zijwjsij≥φq(i 1, 2.. times.n, j 1, 2.. times.m), taken per row in a U { Si } matrixOnly t rows of m elements meeting the condition are taken, and a similar historical sample feature set S is selectedu。
The number of the similar samples is equal to t × m matrix element number, t is the scale of the training samples, and m is the number of the eigenvalues. Threshold value phiqAdaptive adjustment according to the training sample size t, i.e. the threshold value phiqThe criterion for adjustment is to keep m elements per row in the matrix U x { Si } for only t rows greater than a threshold value phiqLet the number of similar samples equal to the number of t m matrix elements.
Si＝[si1,si2,...,sim](i＝1,2,...,n)SiForming an n x m matrix;
u is a matrix of n m, and U { Si } is also a matrix of n m.
7) Training a decision tree in the random forest by using the selected similar historical sample feature set and the corresponding fault category to obtain a trained random forest;
in the training process, the number of historical sample features is used as the feature dimension of each decision tree, the value of the number of the decision trees refers to the number of times that the voltage and current feature values need to be judged, the depth refers to the level of a node when each decision tree is generated, namely the set number of decision time intervals, and the minimum number of samples on the node is set as the number of sampling times in one day. The actual number of samples is the number of eigenvalues multiplied by the number of samples. The minimum information gain on the nodes is 1, and the root node of each decision tree corresponds to the line loss of the similar historical sample feature set;
the fault categories include types of voltage loss, undervoltage, overvoltage, overcurrent, undercurrent, overload, reversal, phase loss, residual current fault, normal power failure and the like.
The input of the random forest is a similar historical sample characteristic set, and the output is a fault discrimination type (fault category).
8) Weighting the fault prediction result of each decision tree in the random forest by using the weighted voting value a of the decision tree and the associated decision matrix U, and adjusting a to obtain a final prediction model (namely a random forest model) with the prediction accuracy as a 100% target;
and (5) directly solving a according to the prediction accuracy rate equal to 100%, and then carrying out reverse calculation by using the formula for solving a in the step 5 to obtain the parameter adjustment factor lambda.
9) And inputting the characteristic vector of the sample to be predicted into the final prediction model to obtain the predicted fault category.
Wherein f isRF(so) Representing random forest algorithm to sample S to be testedoThe classification result of (2), i.e. the final prediction model, is used to decide the classification result. I (-) represents the number satisfying the expression in parentheses, fl tree(so) I represents that the fault prediction result of the first decision tree in the trained random forest is i, c represents the number of the fault prediction result categories of the whole random forest, and i is one of the categories c.
The final decision and classification result of the prediction model is a certain fault category, and the certain fault category can correspond to a certain historical line loss amount, so that a final line loss prediction result is obtained.
Will f isl tree(so) And i, predicting the correct times to obtain the final correct line loss prediction sample number. And the number of the feature vectors of the prediction samples is used as the number of the prediction samples.
10) According to the two formulas of step 9, by using the initial values a and I (eta) and the prediction accuracy of 100%, the correct number of samples is predicted to be equal to the predicted number of samples, U can be calculated, and the value of Z is known in step 2, so that the value of W can be updated adaptively.
11) And calculating a U value by using the updated W value, and combining the I (.) value and the prediction accuracy rate of 100% to apply the formula in the step 9 to adaptively update the value of a.
After the random forest algorithm predicts the fault type, the fault type obtains a historical line loss amount corresponding to a certain fault type, and the historical line loss amount is used as a final line loss amount prediction result.
And after the real-time line loss is predicted by the random forest algorithm, calculating to obtain the current real-time line loss rate by utilizing the real-time line loss to be compared with the total power. And if the current real-time line loss rate is less than or equal to the maximum reasonable line loss rate, judging that the line loss is not abnormal. And if the current real-time line loss rate is greater than the maximum reasonable line loss rate, judging that the line loss is abnormal. Then step 3 is performed.
And 3, if the line loss is abnormal, calculating the abnormal loss.
Take phase a of the low voltage branch as an example. At head end voltage U1And each node Pk+jQkUnder the conditions of (1), the measured abnormal loss when the "abnormal" power consumption exists is as follows:
wherein S isSPower at the head end of the single-phase line for abnormal use, SkThe active power of the k-th user in abnormal power utilization is obtained. The node refers to a section of line of each user, and each section of line is a node.
And 4, calculating the load current of the line loss abnormal point so as to position the abnormal point by using the load current of the abnormal point.
If there is "abnormal" power usage, the load current at the abnormal point should be:
wherein I1Head end current of single-phase line for abnormal power consumption, IkThe current when the kth user abnormally uses electricity is obtained.
In the invention, the load current of an optimal solution, namely an abnormal point, is obtained by minimizing the total loss and the actually measured abnormal loss of a certain user node i. The specific process is as follows:
A1) assuming that the abnormal point is near the ith node, the total extra loss and the abnormal electricity consumption of all lines caused by the abnormal point are as follows:
wherein
Representing the measured current I according to the impedance of each section of line experienced by k userskAnd load current I at abnormal point#And calculating the loss power of each section of line. U shapeiIs the voltage of the ith node, and the addition of an asterisk means that the voltage is the voltage of an abnormal node.
Line loss outlier localization can be described as an optimization problem: in finding an i#So that
To a minimum. When the total extra loss of a certain user node i is closest to the measured abnormal loss, the abnormal line loss caused by the leakage and the electricity stealing at the point can be considered. The actual line loss can be calculated after the abnormal line loss point is located.
A2) And solving the optimization problem by using an improved particle swarm optimization algorithm, and calculating to obtain the optimal load current of the abnormal line loss point.
An improved implementation process for automatically positioning line loss anomaly points by using a particle swarm optimization algorithm is shown in fig. 2, and includes:
1) initializing a particle swarm and randomly initializing each particle; inputting parameters of a particle swarm optimization algorithm and difference vectors of total loss and actually measured abnormal loss of nodes as parameters; taking a difference vector of the total loss of the nodes and the actually measured abnormal loss as an optimization objective function;
2) calculating the fitness value of each particle based on the fitness function; inputting the particle individuals as system variables, and generating feasible solutions based on constraint conditions;
3) comparing the current fitness value with the historical optimal fitness value aiming at the particles, and simultaneously replacing the historical optimal value;
4) comparing the current fitness with the historical optimal fitness value of the population aiming at the particles, and replacing the historical optimal value;
5) calculating the difference between the maximum fitness value and the average fitness value of each particle to obtain an individual fitness difference value, taking the individual fitness difference value as input, adaptively adjusting the speed and the position of each particle, and updating the speed and the position of the particles to obtain a progeny population;
6) dynamically adjusting individuals of the filial generation population to generate a new generation of particle swarm;
dynamically adjusting each particle of the particle swarm to generate a new generation of particle swarm; the method comprises the following steps:
introducing immigration operators, eliminating the worst individuals with a certain elimination rate in the evolution process of each generation, and then replacing the worst individuals with the generated new individuals;
filtering similar individuals, sorting the offspring individuals according to the fitness values, and sequentially calculating the difference value between the maximum fitness value and the average fitness value of each particle and the generalized Hamming distance between the similar individuals of which the difference value is smaller than a threshold delta; if the difference value is smaller than the threshold delta and the generalized Hamming distance is smaller than the threshold delta, filtering the particle;
dynamically supplementing new filial generation individuals, randomly carrying out multiple variations on a plurality of individuals with higher fitness values in the parent generation to generate new individuals, and adding filial generation to generate a new generation of particle swarm.
7) Updating the speed and position of each particle in the new generation of particle swarm through the individual extreme value and the swarm extreme value; comparing the fitness value of each particle with the optimal value, if the fitness value is better, taking the fitness value as the current best position, comparing all current individual extremum values with the group extremum values, and updating the group extremum value;
in the step 7, the speed of each particle in the new-generation particle swarm is updated by formula (14), and the position of each particle in the new-generation particle swarm is updated by formula (15):
wherein the content of the first and second substances,
represents the velocity of the particle after the (k + 1) th iteration, w represents the inertia weight,
representing the velocity of the particle after the k-th iteration, c1Denotes a first learning factor constant, c2A second learning factor constant is represented as a second learning factor constant,
representing the first random number after the kth iteration,
representing the individual extremum of the particle after the k-th iteration,
indicating the position of the particle after the k-th iteration,
representing the second random number after the kth iteration,
representing the extreme value of the particle group after the kth iteration, wherein k represents the iteration times;
where the position of the particle after the (k + 1) th iteration is indicated.
8) And if the minimum difference value between the total extra loss of the node and the actually measured abnormal loss is obtained, ending, and recording the user node as an abnormal line loss point. Otherwise jump to step 5) above.
And 5, calculating the bus loss caused by the abnormal point by using the optimal load current of the abnormal line loss point.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above description is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, several modifications and variations can be made without departing from the technical principle of the present invention, and these modifications and variations should also be regarded as the protection scope of the present invention.
Claims (7)
1. An improved low-voltage transformer area line loss calculation method is characterized by comprising the following steps:
predicting real-time line loss by adopting a random forest algorithm;
calculating to obtain a real-time line loss rate according to the real-time line loss amount;
judging whether the line loss is abnormal or not according to the comparison between the real-time line loss rate and the maximum reasonable line loss rate;
if the line loss is judged to be abnormal, calculating the load current of the abnormal point;
and calculating the bus loss caused by the abnormal point according to the load current of the abnormal point.
2. The improved low-voltage transformer area line loss calculation method as claimed in claim 1, wherein the predicting the real-time line loss amount by using the random forest algorithm comprises:
1) acquiring historical characteristic values of voltage and current, and constructing a historical sample characteristic set;
2) obtaining an association judgment matrix Z by calculating the correlation between the historical sample characteristic set and the sample characteristic set to be predicted;
3) assigning the weight of each feature to construct a feature weight matrix W;
4) performing dot product calculation on the characteristic weight matrix W and the association judgment matrix Z to obtain an association decision matrix U;
5) randomly selecting d features from the feature set to obtain a weighted voting value a of the decision tree;
6) selecting a similar historical sample feature set from the historical sample feature vector set by utilizing a U;
7) training a decision tree in the random forest by using the selected similar historical sample feature set and the corresponding fault category to obtain a trained random forest;
8) weighting the fault prediction result of each decision tree in the random forest by using the weighted voting value a of the decision tree and the associated decision matrix U, and adjusting a to obtain a final prediction model with the prediction accuracy as a target of 100%;
9) inputting the characteristic vector of the sample to be predicted into a final prediction model to obtain a final line loss prediction result;
10) updating the value of W in a self-adaptive manner by using the prediction accuracy rate of 100%;
11) and calculating a U value by using the updated W value, and adaptively updating the value of a.
3. The improved low-voltage transformer area line loss calculation method as claimed in claim 1, wherein the maximum reasonable line loss rate is:
taking the calculation of the a-phase reasonable line loss rate of the low-voltage branch line as an example, the maximum reasonable line loss of the a-phase N users is calculated as follows:
wherein R iskRepresenting the line resistance value between the kth user and the (k-1) th user in the delta t time; pkRepresenting the active power of the kth user per minute in the delta t time period; qkRepresenting the reactive power per minute of the kth user during the at time period;
calculating the total active power consumption of the user as follows:
maximum reasonable line loss rate of the a phaseη0The method comprises the following steps:
when the low-voltage transformer area power grid normally works, the current real-time line loss rate eta is less than or equal to the maximum reasonable line loss rate.
4. The improved low-voltage transformer area line loss calculation method as claimed in claim 1, wherein the step of calculating the load current of the abnormal point comprises the following steps:
the total loss and the measured abnormal loss of a certain user node are minimized as an optimization target,
and calculating to obtain the optimal solution of the optimization target, namely the load current of the abnormal point.
5. The improved low-voltage transformer area line loss calculation method as claimed in claim 4, wherein the optimization objective of minimizing the total loss and the measured abnormal loss of a certain user node comprises:
taking phase a of the low-voltage branch line as an example, the measured abnormal loss when the "abnormal" power consumption exists is as follows:
wherein S isSPower at the head end of the single-phase line for abnormal use, SkThe active power of the k-th user in abnormal power utilization is obtained. The node refers to a section of line of each user, and each section of line is a node;
assuming that the abnormal point is near the ith node, the total extra loss and the abnormal electricity consumption of all lines caused by the abnormal point are as follows:
wherein
Representing the measured current I according to the impedance of each section of line experienced by k userskAnd load current I at abnormal point#Calculating the loss power of each section of line; i is#Is the load current of the abnormal point, (U)i)*Is the voltage of the anomaly;
line loss outlier localization can be described as an optimization problem: in finding an i#So that
The above is the optimization goal.
6. The improved low-voltage transformer area line loss calculation method as claimed in claim 4, wherein the calculating to obtain the optimal solution of the optimization objective comprises:
and calculating by adopting a particle swarm algorithm to obtain the optimal solution of the optimization target.
7. The improved low-voltage transformer area line loss calculation method as claimed in claim 6, wherein the obtaining of the optimal solution of the optimization target by using particle swarm optimization comprises:
1) initializing a particle swarm and randomly initializing each particle; inputting parameters of a particle swarm optimization algorithm and difference vectors of total loss and actually measured abnormal loss of nodes as parameters; taking a difference vector of the total loss of the nodes and the actually measured abnormal loss as an optimization objective function;
2) calculating the fitness value of each particle based on the fitness function; inputting the particle individuals as system variables, and generating feasible solutions based on constraint conditions;
3) comparing the current fitness value with the historical optimal fitness value aiming at the particles, and simultaneously replacing the historical optimal value;
4) comparing the current fitness with the historical optimal fitness value of the population aiming at the particles, and replacing the historical optimal value;
5) calculating the difference between the maximum fitness value and the average fitness value of each particle to obtain an individual fitness difference value, taking the individual fitness difference value as input, adaptively adjusting the speed and the position of each particle, and updating the speed and the position of the particles to obtain a progeny population;
6) dynamically adjusting individuals of the filial generation population to generate a new generation of particle swarm;
dynamically adjusting each particle of the particle swarm to generate a new generation of particle swarm; the method comprises the following steps:
introducing immigration operators, eliminating the worst individuals with a certain elimination rate in the evolution process of each generation, and then replacing the worst individuals with the generated new individuals;
filtering similar individuals, sorting the offspring individuals according to the fitness values, and sequentially calculating the difference value between the maximum fitness value and the average fitness value of each particle and the generalized Hamming distance between the similar individuals of which the difference value is smaller than a threshold delta; if the difference value is smaller than the threshold delta and the generalized Hamming distance is smaller than the threshold delta, filtering the particle;
dynamically supplementing new filial generation individuals, randomly carrying out multiple variations on a plurality of individuals with higher fitness values in the parent generation to generate new individuals, and adding filial generations to generate a new generation of particle swarm;
7) updating the speed and position of each particle in the new generation of particle swarm through the individual extreme value and the swarm extreme value; comparing the fitness value of each particle with the optimal value, if the fitness value is better, taking the fitness value as the current best position, comparing all current individual extremum values with the group extremum values, and updating the group extremum value;
updating the speed of each particle in the new generation of particle swarm by adopting a formula (14), and updating the position of each particle in the new generation of particle swarm by adopting a formula (15):
wherein the content of the first and second substances,
represents the velocity of the particle after the (k + 1) th iteration, w represents the inertia weight,
representing the velocity of the particle after the k-th iteration, c1Denotes a first learning factor constant, c2A second learning factor constant is represented as a second learning factor constant,
representing the first random number after the kth iteration,
representing the individual extremum of the particle after the k-th iteration,
indicating the position of the particle after the k-th iteration,
representing the second random number after the kth iteration,
representing the extreme value of the particle group after the kth iteration, wherein k represents the iteration times;
wherein, the position of the particle after the (k + 1) th iteration is represented;
8) and if the minimum difference value between the total extra loss of the node and the actually measured abnormal loss is obtained, ending, and recording the user node as an abnormal line loss point. Otherwise, jumping to step 5).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010621395.2A CN111985524A (en) | 2020-07-01 | 2020-07-01 | Improved low-voltage transformer area line loss calculation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010621395.2A CN111985524A (en) | 2020-07-01 | 2020-07-01 | Improved low-voltage transformer area line loss calculation method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111985524A true CN111985524A (en) | 2020-11-24 |
Family
ID=73437680
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010621395.2A Pending CN111985524A (en) | 2020-07-01 | 2020-07-01 | Improved low-voltage transformer area line loss calculation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111985524A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114113885A (en) * | 2021-11-19 | 2022-03-01 | 国网甘肃省电力公司电力科学研究院 | Redundancy check-based accurate positioning method for abnormal low-voltage phase-splitting line loss |
-
2020
- 2020-07-01 CN CN202010621395.2A patent/CN111985524A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114113885A (en) * | 2021-11-19 | 2022-03-01 | 国网甘肃省电力公司电力科学研究院 | Redundancy check-based accurate positioning method for abnormal low-voltage phase-splitting line loss |
| CN114113885B (en) * | 2021-11-19 | 2023-09-22 | 国网甘肃省电力公司电力科学研究院 | Redundancy check-based abnormal low-voltage split-phase line loss accurate positioning method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107516170B (en) | Difference self-healing control method based on equipment failure probability and power grid operation risk | |
| CN111985571B (en) | Low-voltage intelligent monitoring terminal fault prediction method, device, medium and equipment | |
| CN109298330B (en) | High-voltage circuit breaker fault diagnosis method based on GHPSO-BP | |
| CN112734128A (en) | 7-day power load peak value prediction method based on optimized RBF | |
| CN110929835B (en) | Novel silicon carbide-based aviation power converter fault diagnosis method and system | |
| CN111460001A (en) | Theoretical line loss rate evaluation method and system for power distribution network | |
| Gensler et al. | An analog ensemble-based similarity search technique for solar power forecasting | |
| CN111985524A (en) | Improved low-voltage transformer area line loss calculation method | |
| CN114971090A (en) | Electric heating load prediction method, system, equipment and medium | |
| CN117170980B (en) | Early warning method, device, equipment and storage medium for server hardware abnormality | |
| CN113328467B (en) | Probability voltage stability evaluation method, system, terminal device and medium | |
| Yasin et al. | Quantum-inspired evolutionary programming-artificial neural network for prediction of undervoltage load shedding | |
| CN117330963A (en) | Energy storage power station fault detection method, system and equipment | |
| CN112818525A (en) | Hydroelectric generating set state evaluation method and system | |
| CN114611799B (en) | Time sequence neural network new energy output multi-step prediction method based on supervised learning | |
| CN113420912B (en) | Method for identifying users with low-voltage abnormality of power distribution network | |
| CN115907228A (en) | Short-term power load prediction analysis method based on PSO-LSSVM | |
| CN113496255B (en) | Power distribution network mixed observation point distribution method based on deep learning and decision tree driving | |
| CN115133556A (en) | Power system transient stability prevention control method based on deep cross network | |
| CN113901622A (en) | Genetic algorithm optimization-based SVM power distribution network topology identification method | |
| CN110571791B (en) | Optimal configuration method for power transmission network planning under new energy access | |
| CN114429172A (en) | Load clustering method, device, equipment and medium based on transformer substation user constitution | |
| CN110942195A (en) | Power load prediction method and device | |
| CN113762591A (en) | Short-term electric quantity prediction method and system based on GRU and multi-core SVM counterstudy | |
| CN108009686B (en) | Photovoltaic power generation power prediction method, device and system |
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 |