METHOD AND DEVICE FOR FORMULATING COORDINATED ACTION STRATEGY OF SSTS AND DVR FOR VOLTAGE SAG MITIGATION

Abstract

The present invention discloses a method and device for formulating a coordinated action strategy of SSTS and DVR for voltage sag mitigation. The influence of voltage sag on a whole industrial process of a sensitive user is analyzed, and the sensitive loads of SSTS and DVR which satisfy a governance need are grouped; a practical governance scenario of installing a plurality of DVR is considered to install a minimum-capacity DVR to realize a target of a minimum interruption probability of the whole industrial process of the user; an optimal coordinated governance solution of SSTS and DVR based on sensitive load grouping is proposed; a classification result is obtained for duration time at a time when a voltage sag event occurs through a decision tree constructed based on historical data.

Claims

1. A method for formulating a coordinated action strategy of SSTS and DVR for voltage sag mitigation, comprising the following steps: step 1: grouping mitigation of sensitive loads: considering the whole industrial process of a sensitive user, and realizing the grouping of two groups of sensitive loads of SSTS mitigation and DVR mitigation; conducting grouping optimization again for the sensitive loads with the installation of DVR for compensation and mitigation; and finally outputting a grouping matrix and grouping compensation voltage; step 2: classification of voltage sag duration time: based on the characteristics of historical voltage sag monitoring data, constructing a decision tree to predict whether the duration time of a newly generated voltage sag event is less than SSTS switching time to conduct classification; step 3: according to a grouping mitigation solution of the sensitive loads obtained in step 1 and the classification of the voltage sag duration time obtained in step 2, when voltage sag is monitored, formulating an action strategy of SSTS and DVR as follows: if the voltage sag duration time is less than the SSTS switching time, SSTS acts; otherwise, not acts; if the amplitude of voltage sag is less than or equal to a minimum value of a voltage sag tolerance amplitude in a group of sensitive loads, the group of DVR acts; the grouping mitigation of the sensitive loads in step 1 specifically comprises: step 1.1: grouping the sensitive loads according to connection modes and function types of the sensitive loads in the industrial process, comprising: S.sub.1-type sensitive loads: electrical series sensitive loads which cause a sub-process to be interrupted when tripping; S.sub.2-type sensitive loads: electrical parallel sensitive loads which may not cause the sub-process to be interrupted when tripping and may cause the sub-process to be interrupted when all the S.sub.2-type sensitive loads trip; S.sub.3-type sensitive loads: control parallel sensitive loads which act on the industrial process through signal control, which are not directly connected to the industrial process, but may cause a control process to be interrupted when tripping; step 1.2: based on the above grouping, further conducting secondary grouping for each sensitive load: 1. dividing the S.sub.2-type sensitive loads into two categories according to whether the loads can recover automatically after suffering from voltage sag: sensitive loads capable of automatic recovery and sensitive loads incapable of automatic recovery; 2. dividing the S.sub.1-type and S.sub.3-type sensitive loads and the sensitive loads incapable of automatic recovery in the S.sub.2-type sensitive loads into two categories T.sub.tolerance≤T.sub.1 and T.sub.tolerance>T.sub.1 according to tolerant voltage sag duration time T.sub.tolerance, wherein T.sub.1 is the SSTS switching time; the former is compensated and governed by installing DVR, and the latter is governed by SSTS switching lines; step 1.3: conducting grouping optimization of DVR mitigation on the sensitive loads with the tolerant voltage sag duration time T.sub.tolerance≤T.sub.1: a grouping goal is to install minimum-capacity DVR, and a governance goal is to achieve a minimum interruption probability of a sensitive industrial process of the user; therefore, a grouping optimization model of the sensitive loads is constructed by taking the minimization of the capacity of the installed DVR and the minimization of the interruption probability of the sensitive industrial process as an objective function, and decision variables are the grouping matrix and the grouping compensation voltage.

2. The method for formulating the coordinated action strategy of SSTS and DVR for voltage sag mitigation according to claim 1, wherein according to the grouping in the step 1.1, a calculation method of the probability that the industrial process of the sensitive loads is interrupted comprises: assuming that an industrial user has n sensitive loads and M sensitive industrial processes; j is made to represent a sensitive load, and m represents a sensitive industrial process, i.e., j=1,2, . . . ,n, m=1,2, . . . , M; then the tripping probabilities P.sub.process-m.sub.1, P.sub.process-m.sub.m and P.sub.process-m.sub.3, of the sub-processes of S.sub.1-type, S.sub.2-type and S.sub.3-type sensitive loads are respectively: $\begin{array}{cc}{P}_{\mathrm{process}-m_1}=1-\underset{j=1}{\overset{A}{.Math.}}\left(1-P_{S_1-j}\right)& \left(1\right)\end{array}$ $\begin{array}{cc}{P}_{\mathrm{process}-m_2}=\underset{j=1}{\overset{B}{.Math.}}{P}_{S_2-j}& \left(2\right)\end{array}$ $\begin{array}{cc}{P}_{\mathrm{process}-m_3}=1-\underset{j=1}{\overset{C}{.Math.}}\left(P_{S_3-j}{P}_{S_3-\mathrm{control}-j}\right)& \left(3\right)\end{array}$ wherein A, B and C are the number of three types of sensitive loads respectively; P.sub.S.sub.1.sub.-j, P.sub.S.sub.2.sub.-j and P.sub.S.sub.3.sub.-j are the tripping probabilities of the three types of sensitive loads respectively; and P.sub.S.sub.3.sub.-control-j is the tripping probability of a device controlled by the S.sub.3-type sensitive loads; assuming that an m-th industrial process contains K, L and Q of the above three sub-processes respectively, the interruption probability of the industrial process is: $\begin{array}{cc}{P}_{\mathrm{process}-m}=\underset{m_1=1}{\overset{K}{.Math.}}{P}_{\mathrm{process}-m_1}\underset{m_2=1}{\overset{L}{.Math.}}{P}_{\mathrm{process}-m_2}\underset{m_3=1}{\overset{Q}{.Math.}}{P}_{\mathrm{process}-m_3}& \left(4\right)\end{array}$

3. The method for formulating the coordinated action strategy of SSTS and DVR for voltage sag mitigation according to claim 1, wherein the grouping mitigation optimization of the sensitive loads in step 1.3 specifically comprises: 1) setting an objective function: setting the grouping matrix [α.sub.0,α.sub.1, . . . ,α.sub.n].sub.i=[α.sub.j], wherein i represents an i-th group, α.sub.j=0 or 1, α.sub.j=0 indicates that the sensitive load is not in the i-th group, and α.sub.j=1 indicates that the sensitive load is in the i-th group; a first optimization objective is to minimize the sum of the capacities of installed DVR: $\begin{array}{cc}\min S_{\mathrm{DVR}}=\underset{i=1}{\overset{N}{.Math.}}\left(\frac{U_i}{U_n}.Math.S_{\mathrm{load}-i}\right)& \left(5\right)\end{array}$ wherein SDVR is the sum of the capacities of i DVRs, N is the number of groups, U.sub.i is the compensation voltage of the i-th group, U.sub.n is the rated voltage of the sensitive user, and S.sub.load-i is the sum of the capacities of the i-th group of sensitive loads to be governed; a second optimization objective is to minimize the interruption probability of the sensitive industrial process: $\begin{array}{cc}\min P_{\mathrm{process}}=\underset{m=1}{\overset{M}{.Math.}}{P}_{\mathrm{process}-m}& \left(6\right)\end{array}$ wherein P.sub.process-m is the interruption probability of an m-th sensitive industrial process; 2) setting constraints {circle around (1)} capacity constraints of the sensitive loads $\begin{array}{cc}{S}_{\mathrm{load}-i}=\underset{j=1}{\overset{n}{.Math.}}\left({\alpha }_j{S}_j\right)& \left(7\right)\end{array}$ wherein S.sub.j is the rated capacity of a jth sensitive load; {circle around (2)} tripping probability constraints of the sensitive loads the tripping probability P.sub.j of a single sensitive load is:
P.sub.j=∫∫.sub.Ωp(T.sub.sag)p(U.sub.sag)dU.sub.sagdT.sub.sag  (8) wherein U.sub.sag and T.sub.sag are amplitude and duration time of voltage sag respectively; p(U.sub.sag) and p(T.sub.sag) are probability density functions of the amplitude and the duration time of voltage sag respectivelyΩ is a fault region determined by a load VTC; the specific data of each sensitive load is substituted into the above formula to obtain P.sub.S.sub.1.sub.-j, and P.sub.S.sub.2.sub.-j and P.sub.S.sub.3.sub.-j; {circle around (3)} DVR compensation voltage constraints U.sub.i is a compensation voltage amplitude that the DVR installed in the i-th group should output, i.e., a maximum value of compensation voltage required by the sensitive load with α.sub.j=1 in the grouping matrix of the i-th group, and an expression is:
U.sub.i=max{U.sub.α.sub.0,I.sub.α.sub.1, . . . ,U.sub.α.sub.n|α.sub.j}  (9)
U.sub.α.sub.j≤U.sub.demand-α.sub.j   (10) wherein U.sub.α.sub.j is the compensation voltage of the jth sensitive load, and U.sub.demand-α.sub.j is the highest compensation voltage of the jth sensitive load to satisfy the requirement for voltage sag mitigation; {circle around (4)} grouping constraints of the sensitive loads there are only two cases for the grouping of any sensitive load: a. the sensitive load does not belong to any group, i.e.: α.sub.j=0∈[α.sub.0,α.sub.1, . . . ,α.sub.n].sub.i, and α.sub.j=0∈[α.sub.0, α.sub.1, . . . ,α.sub.n].sub.else-i; b. if the sensitive load is divided into a certain group, the sensitive load is and can only be in the group; i.e.: when α.sub.j=0∈[α.sub.0,α.sub.1, . . . ,α.sub.n].sub.i, and α.sub.j=0∈[α.sub.0, α.sub.1, . . . ,α.sub.n].sub.else-i; wherein [α.sub.0,α.sub.1, . . . ,α.sub.n].sub.else-i is a grouping matrix of other groups except the i-th group; 3) model solving solving a Pareto optimal solution set of the model by an NSGA-II algorithm, and giving satisfaction to each objective function corresponding to each group of solutions in the Pareto optimal solution set by a slightly small fuzzy satisfaction function, as shown in formula (11): $\begin{array}{cc}{\mu }_{\mathrm{vo}}=\{\begin{array}{cc}1,& f_{\mathrm{vo}}\le f_{o\min }\\ \frac{f_{o\max }-f_{\mathrm{vo}}}{f_{o\max }-f_{o\min }},& f_{o\min }\le f_{\mathrm{vo}}\le f_{o\max }\\ 0,& f_{\mathrm{vo}}\ge f_{o\min }\end{array}& \left(11\right)\end{array}$ in the formula, o∈{1,2, . . . , O}; O is the number of objective functions; μ.sub.vo is the satisfaction of an oth objective function corresponding to a vth group of Pareto solutions; f.sub.vo is a function value of the oth objective function corresponding to the vth group of solutions in the Pareto solution set; f.sub.omin is a minimum value of the function values of the oth objective function corresponding to all the solutions in the Pareto solution set; and fomax is a maximum value of the function values of the oth objective function corresponding to all the solutions in the Pareto solution set; solving the satisfaction μ.sub.v of each Pareto solution based on the satisfaction of each objective function corresponding to each Pareto solution; $\begin{array}{cc}{\mu }_v=\frac{1}{O}\underset{v=1}{\overset{O}{.Math.}}{\mu }_{\mathrm{vo}}& \left(12\right)\end{array}$ using a Pareto solution with largest satisfaction μ.sub.v as a final solution of a decision variable.

4. The method for formulating the coordinated action strategy of SSTS and DVR for voltage sag mitigation according to claim 1, wherein the step 2 specifically comprises: step 2.1: discretizing conditional attribute data for the classification of the voltage sag duration time, selecting four characteristics of amplitude, phase jump, date and weather from multi-dimensional attributes as conditional attributes; for two types of continuous attribute data of the amplitude and the phase jump, merging adjacent sections according to chi-square test by a ChiMerge discrete method until criterion conditions are met; when discretizing date data, converting the date data into a digital quantity having a value changed continuously between 1 and 365 by taking days as a unit and years as a cycle; for the language description data of weather, dividing the weather according to weather categories; step 2.2: building a decision tree root node 1) calculating an information entropy E(T) of the voltage sag duration time T.sub.sag<T.sub.1:
E(T)=−(p.sub.1log.sub.2p.sub.1+p.sub.2log.sub.2p.sub.2)   (13) wherein T.sub.1 is the SSTS switching time; p.sub.1 and p.sub.2 are probabilities that a datum which is greater than or equal to T.sub.1 and less than T.sub.1 is randomly selected from a historical data set T of the voltage sag duration time; 2) calculating a conditional entropy E(T,X) between the above four conditional attributes and T.sub.sag<T.sub.1: $\begin{array}{cc}E\left(T,X\right)=\underset{c\in X}{.Math.}P\left(c\right)E\left(c\right)& \left(14\right)\end{array}$ wherein X represents four conditional attributes; c represents a conditional attribute; P(c) is a joint probability that a conditional attribute and T.sub.sag<T.sub.1 appear at the same time; and E(c) is a conditional probability of T.sub.sag<T.sub.1 under a conditional attribute and with different values; 3) calculating mutual information Gain(T,X) between the above four conditional attributes and T.sub.sag<T.sub.1:
Gain(T,X)=E(T)=E(T,X)   (15) the larger the value of mutual information is, the higher the correlation with T.sub.sag<T.sub.1 is; a conditional attribute with largest mutual information is selected as a decision tree root node; step 2.3: building decision tree branch nodes and leaf nodes a specific operation process is the same as three points in step 2.2: calculating the information entropy, the conditional entropy and the mutual information; gradually discretizing results with the largest mutual information with T.sub.sag<T.sub.1 from results by using the conditional attributes, and attributes with largest mutual information with T.sub.sag<T.sub.1 from other conditional attributes as branch nodes; and iterating the process repeatedly until a complete decision tree with “Yes, No” of T.sub.sag<T.sub.1 as a leaf node is finally constructed based on historical data; step 2.4: conducting classification by the decision tree when a monitoring device monitors a voltage sag event, inputting four conditional attribute data, and classifying the duration time of the new voltage sag event by a generated decision tree logic; and when the input data is consistent with the decision tree, entering a next branch for judgment, until a classification result of yes or no is obtained finally through the leaf node.

5. A device for formulating a coordinated action strategy of SSTS and DVR for voltage sag mitigation, comprising a sensitive load grouping mitigation module, a voltage sag duration time classification module and an SSTS and DVR coordinated action strategy module; the sensitive load grouping mitigation module comprises a first grouping module, a second grouping module and a sensitive load grouping mitigation optimization module; the first grouping module divides the sensitive loads into electrical series sensitive loads, electrical parallel sensitive loads and control parallel sensitive loads according to the connection modes and function types of the sensitive loads in the industrial process; the second grouping module divides the electrical parallel sensitive loads into the sensitive loads which can automatically recover and cannot recover automatically, and further classifies the sensitive loads which cannot recover automatically in the electrical series sensitive loads, the control parallel sensitive loads and the electrical parallel sensitive loads according to size relationships between the tolerant voltage sag duration time and the SSTS switching time; the sensitive load grouping mitigation optimization module further groups and optimizes the sensitive loads having tolerant voltage sag duration time less than or equal to the SSTS switching time; the voltage sag duration time classification module constructs a decision tree based on the characteristics of historical voltage sag monitoring data to predict whether the duration time of the newly generated voltage sag event is less than the SSTS switching time, to conduct classification; the SSTS and DVR coordinated action strategy module formulates the action strategy of SSTS and DVR according to classification structures of the sensitive load grouping mitigation module and the voltage sag duration time classification module: if the voltage sag duration time is less than the SSTS switching time, SSTS acts; otherwise, not acts; if the amplitude of voltage sag is less than or equal to a minimum value of a voltage sag tolerance amplitude in a group of sensitive loads, the group of DVR acts; the sensitive load grouping mitigation module groups the sensitive loads according to connection modes and function types of the sensitive loads in the industrial process, comprising: S.sub.1-type sensitive loads: electrical series sensitive loads which cause a sub-process to be interrupted when tripping; S.sub.2-type sensitive loads: electrical parallel sensitive loads which may not cause the sub-process to be interrupted when tripping and may cause the sub-process to be interrupted when all the S.sub.2-type sensitive loads trip; S.sub.3-type sensitive loads: control parallel sensitive loads which act on the industrial process through signal control, which are not directly connected to the industrial process, but may cause a control process to be interrupted when tripping; based on the above grouping, further conducting secondary grouping for each sensitive load: 1) dividing the S.sub.2-type sensitive loads into two categories according to whether the loads can recover automatically after suffering from voltage sag: sensitive loads capable of automatic recovery and sensitive loads incapable of automatic recovery; 2) dividing the S.sub.1-type and S.sub.3-type sensitive loads and the sensitive loads incapable of automatic recovery in the S.sub.2-type sensitive loads into two categories T.sub.tolerance≤T.sub.1 and T.sub.tolerance>T.sub.1 according to tolerant voltage sag duration time T.sub.tolerance, wherein T.sub.1 is the SSTS switching time; the former is compensated and governed by installing DVR, and the latter is governed by SSTS switching lines; step 1.3: conducting grouping optimization of DVR mitigation on the sensitive loads with the tolerant voltage sag duration time T.sub.tolerance≤T.sub.1: a grouping goal is to install minimum-capacity DVR, and a governance goal is to achieve a minimum interruption probability of a sensitive industrial process of the user; therefore, a grouping optimization model of the sensitive loads is constructed by taking the minimization of the capacity of the installed DVR and the minimization of the interruption probability of the sensitive industrial process as an objective function, and decision variables are the grouping matrix and the grouping compensation voltage.

Description

DESCRIPTION OF DRAWINGS

[0089] FIG. 1 is a sensitive load VTC.

[0090] FIG. 2 is a structural diagram of a device for formulating a coordinated action strategy of SSTS and DVR for voltage sag mitigation in the present invention.

[0091] FIG. 3 is a flow chart of solving by NSGA-algorithm.

DETAILED DESCRIPTION

[0092] The present invention will be further described in detail below in combination with the drawings and the embodiments.

[0093] A device for formulating a coordinated action strategy of SSTS and DVR in the present invention comprises three modules: “module I: sensitive load grouping mitigation module”, “module II: voltage sag duration time T.sub.sag classification module” and “module III: SSTS and DVR coordinated action strategy module” to realize a coordinated action strategy of SSTS and DVR for voltage sag mitigation. An overall flow chart is shown in FIG. 2. Module I: sensitive load grouping mitigation module

[0094] The module is a sensitive load grouping mitigation module, which considers the whole industrial process of a sensitive user, and realizes the grouping of two groups of sensitive loads of SSTS mitigation and DVR mitigation; grouping optimization is conducted again for the sensitive loads with the installation of DVR for compensation and mitigation; and a grouping matrix and grouping compensation voltage are finally outputted. The module comprises the following steps:

[0095] Step 1: sensitive load grouping I

[0096] grouping the sensitive loads according to connection modes and function types of the sensitive loads in the industrial process.

[0097] (1) electrical series (S.sub.1): sensitive loads which cause a sub-process to be interrupted when tripping;

[0098] (2) electrical parallel (S.sub.2): sensitive loads which may not cause the sub-process to be interrupted when tripping and assuming that the sub-process is interrupted when all the S.sub.2-type sensitive loads trip;

[0099] (3) control parallel (S.sub.3): sensitive loads which act on the industrial process through signal control, which are not directly connected to the industrial process, but may cause a control process to be interrupted when tripping.

[0100] It is assumed that an industrial user has n sensitive loads and M sensitive industrial processes; j is made to represent a sensitive load, and m represents a sensitive industrial process, i.e., j=1,2, . . . ,n, m=1,2, . . . , M. It is assumed that the SSTS switching time is T.sub.1 and DVR switching time is T.sub.2. In practice, T.sub.1>T.sub.2.

[0101] The tripping probabilities P.sub.process-m.sub.1, P.sub.process-m.sub.2 and P.sub.process-m.sub.3 of the sub-processes of the three types of sensitive loads are respectively:

[00009]$\begin{array}{cc}{P}_{\mathrm{process}-m_1}=1-\underset{j=1}{\overset{A}{.Math.}}\left(1-P_{S_1-j}\right)& \left(1\right)\end{array}$$\begin{array}{cc}{P}_{\mathrm{process}-m_2}=\underset{j=1}{\overset{B}{.Math.}}{P}_{S_2-j}& \left(2\right)\end{array}$$\begin{array}{cc}{P}_{\mathrm{process}-m_3}=1-\underset{j=1}{\overset{C}{.Math.}}\left(P_{S_3-j}{P}_{S_3-\mathrm{control}-j}\right)& \left(3\right)\end{array}$

[0102] wherein A, B and C are the number of three types of sensitive loads respectively; P.sub.S.sub.1.sub.-j, P.sub.S.sub.2.sub.-j and P.sub.S.sub.3.sub.-j are the tripping probabilities of the three types of sensitive loads respectively; and P.sub.S.sub.51.sub.-control-j is the tripping probability of a device controlled by the S.sub.3-type sensitive loads.

[0103] It is assumed that an m-th industrial process contains K, L and Q of the above three sub-processes respectively, the interruption probability of the industrial process is:

[00010]$\begin{array}{cc}{P}_{\mathrm{process}-m}=\underset{m_1=1}{\overset{K}{.Math.}}{P}_{\mathrm{process}-m_1}\underset{m_2=1}{\overset{L}{.Math.}}{P}_{\mathrm{process}-m_2}\underset{m_3=1}{\overset{Q}{.Math.}}{P}_{\mathrm{process}-m_3}& \left(4\right)\end{array}$

[0104] Step 2: sensitive load grouping II

[0105] (1) dividing the S.sub.2-type sensitive loads into two categories according to whether the loads can recover automatically after suffering from voltage sag: sensitive loads capable of automatic recovery and sensitive loads incapable of automatic recovery;

[0106] (2) dividing the S.sub.1-type and S3-type sensitive loads and the sensitive loads incapable of automatic recovery in the S.sub.2-type sensitive loads into two categories T.sub.tolerance≤T.sub.1 and T.sub.tolerance>T.sub.1 according to tolerant voltage sag duration time T.sub.tolerance, wherein the former is compensated and governed by installing DVR, and the latter is governed by SSTS switching lines;

[0107] (3) conducting grouping optimization of DVR mitigation on the sensitive loads with the tolerant voltage sag duration time ≤T.sub.1.

[0108] Step 3: sensitive load grouping mitigation optimization

[0109] conducting grouping optimization on the sensitive loads with installed DVR for compensation and mitigation, with a grouping goal to install minimum-capacity DVR, and a governance goal to achieve a minimum interruption probability of a sensitive industrial process of the user; and therefore, constructing a grouping optimization model of the sensitive loads by taking the minimization of the capacity of the installed DVR and the minimization of the interruption probability of the sensitive industrial process as an objective function, wherein decision variables are the grouping matrix and the grouping compensation voltage.

[0110] (1) Objective function

[0111] setting the grouping matrix [α.sub.0,α.sub.1, . . . ,α.sub.n].sub.i=1, wherein i represents an i-th group, α.sub.j=0 or 1, α.sub.j=0 indicates that the sensitive load is not in the i-th group, and α.sub.j=1 indicates that the sensitive load is in the i-th group; a first optimization objective is to minimize the sum of the capacities of installed DVR:

[00011]$\begin{array}{cc}\min S_{\mathrm{DVR}}=\underset{i=1}{\overset{N}{.Math.}}\left(\frac{U_i}{U_n}.Math.S_{\mathrm{load}-i}\right)& \left(5\right)\end{array}$

[0112] wherein SDVR is the sum of the capacities of i DVRs, N is the number of groups, U.sub.1 is the grouping compensation voltage, U.sub.n is the rated voltage of the sensitive user, and S.sub.load-i is the sum of the capacities of the i-th group of sensitive loads to be governed.

[0113] In addition, because each group has one DVR, i groups have i DVRs.

[0114] A second optimization objective is to minimize the interruption probability of the sensitive industrial process:

[00012]$\begin{array}{cc}\min P_{\mathrm{process}}=\underset{m=1}{\overset{M}{.Math.}}{P}_{\mathrm{process}-m}& \left(6\right)\end{array}$

[0115] wherein P.sub.process-m is the interruption probability of an m-th sensitive industrial process.

[0116] (2) Constraints

[0117] {circle around (1)} Capacity constraints of the sensitive loads

[00013]$\begin{array}{cc}{S}_{\mathrm{load}-i}=\underset{j=1}{\overset{n}{.Math.}}\left({\alpha }_j{S}_j\right)& \left(7\right)\end{array}$

[0118] wherein S.sub.j is the rated capacity of a jth sensitive load.

[0119] {circle around (2)} Tripping probability constraints of the sensitive loads

[0120] the tripping probability P.sub.j of a single sensitive load is

P.sub.j=∫∫.sub.Ωp(T.sub.sag)p(U.sub.sag)dU.sub.sagdT.sub.sag   (8)

[0121] wherein U.sub.sag and T.sub.sag are amplitude and duration time of voltage sag respectively; p(U.sub.sag) and p(T.sub.sag) are probability density functions of the amplitude and the duration time of voltage sag respectively, which are obtained by fitting according to the historical monitoring data; Ω is a fault region determined by a load VTC. with the change of U.sub.i, a knee point of the VTC changes, and Ω changes accordingly. The specific data of each sensitive load is substituted into the above formula to obtain P.sub.S.sub.1.sub.-j, P.sub.S.sub.2.sub.-j and P.sub.S.sub.3.sub.-j.

[0122] {circle around (3)} DVR compensation voltage constraints

[0123] U.sub.i is a compensation voltage amplitude that the DVR installed in the i-th group should output, i.e., a maximum value of compensation voltage required by the sensitive load with α.sub.j=1 in the grouping matrix of the i-th group, and an expression is:

U.sub.i=max{U.sub.α.sub.0, U.sub.α.sub.1, . . . , U.sub.α.sub.n|α.sub.j−1}  (9)

U.sub.α.sub.j≤U.sub.demand-α.sub.j   (10)

[0124] wherein U.sub.α.sub.1 is the compensation voltage of the jth sensitive load, and U.sub.demand-α.sub.j is the highest compensation voltage of the jth sensitive load to satisfy the requirement for voltage sag mitigation.

[0125] {circle around (4)} Grouping constraints of the sensitive loads

[0126] There are only two cases for the grouping of any sensitive load:

[0127] a. the sensitive load does not belong to any group, i.e.: α.sub.j=0∈[α.sub.0,α.sub.1, . . . ,α.sub.n].sub.i, and α.sub.j=0∈[α.sub.0,α.sub.1, . . . ,α.sub.n].sub.else-i;

[0128] b. if the sensitive load is divided into a certain group, the sensitive load is and can only be in the group. i.e.: when α.sub.j=1∈[α,α.sub.1, . . . ,α.sub.n].sub.i, α.sub.j=0∈[α.sub.0,α.sub.1, . . . ,α.sub.n].sub.else-i.

[0129] Wherein [α.sub.0,α.sub.1, . . . , α.sub.n].sub.else-i is a grouping matrix of other groups except the i-th group.

[0130] (3) Model solving

[0131] The minimum DVR capacity in the optimization model and the minimum interruption probability in the industrial process are two contradictory goals. When the decision variable is changed in a given feasible region, the optimization of the DVR capacity will cause the degradation of the interruption probability of the industrial process, so that a set of solutions which make the objective functions reach the minimum values at the same time does not exist, and the Pareto solution set can only be solved. NSGA-II algorithm is an effective method for searching Pareto frontier based on a genetic algorithm, and is suitable for solving the multi-objective optimization model here. The specific solving process is shown in FIG. 3.

[0132] After solving the Pareto optimal solution set by the NSGA-II algorithm, a set of optimal compromise solutions needs to be selected as a final solution for sensitive load grouping and compensation voltage for each group.

[0133] Two objective functions of the optimization model here pursue the minimum values. Satisfaction is given to each objective function corresponding to each group of solutions in the Pareto optimal solution set by a slightly small fuzzy satisfaction function, as shown in formula (11):

[00014]$\begin{array}{cc}{\mu }_{\mathrm{vo}}=\{\begin{array}{cc}1,& f_{\mathrm{vo}}\le f_{o\min }\\ \frac{f_{o\max }-f_{\mathrm{vo}}}{f_{o\max }-f_{o\min }},& f_{o\min }\le f_{\mathrm{vo}}\le f_{o\max }\\ 0,& f_{\mathrm{vo}}\ge f_{o\min }\end{array}& \left(11\right)\end{array}$

[0134] in the formula, o∈{1,2, . . . ,O}; O is the number of objective functions; μ.sub.vo is the satisfaction of an oth objective function corresponding to a vth group of Pareto solutions; f.sub.vo is a function value of the oth objective function corresponding to the vth group of solutions in the

[0135] Pareto solution set; f.sub.omin is a minimum value of the function values of the oth objective function corresponding to all the solutions in the Pareto solution set; and Lax is a maximum value of the function values of the oth objective function corresponding to all the solutions in the Pareto solution set;

[0136] The satisfaction μ.sub.v of each Pareto solution is solved based on the satisfaction of each objective function corresponding to each Pareto solution;

[00015]$\begin{array}{cc}{\mu }_v=\frac{1}{O}\underset{v=1}{\overset{O}{.Math.}}{\mu }_{\mathrm{vo}}& \left(12\right)\end{array}$

[0137] A Pareto solution with largest satisfaction μ.sub.v is used as a final solution of a decision variable.

[0138] Module II: voltage sag duration time T.sub.sag classification module

[0139] The module is a voltage sag duration time T.sub.sag classification module which constructs a decision tree based on the characteristics of historical voltage sag monitoring data to predict the classification of the duration time T.sub.sag<T.sub.1 of the newly generated voltage sag event: yes or no. “Yes” indicates T.sub.sag<T.sub.1 and “No” indicates T.sub.sag>T.sub.1. The module comprises the following steps:

[0140] Step 1: discretizing conditional attribute data

[0141] For the classification of the voltage sag duration time, selecting four characteristics of amplitude, phase jump, date and weather from multi-dimensional attributes as conditional attributes. For two types of continuous attribute data of the “amplitude” and the “phase jump”, merging adjacent sections according to chi-square test by a ChiMerge discrete method until criterion conditions are met; when discretizing “date” data, converting the date data into a digital quantity having a value changed continuously between “1 and 365” by taking days as a unit and years as a cycle; for the language description data of “weather”, dividing the weather into four categories: “sunny, snowy, thunderstorm and cloudy”.

[0142] Step 2: building a decision tree root node

[0143] (1) Calculating an information entropy E(T) of the voltage sag duration time T.sub.sag<T.sub.1:

E(T)=−(p.sub.1log.sub.2p.sub.1+p.sub.2log2p.sub.2)   (13)

[0144] wherein p.sub.1 and p.sub.2 are probabilities that a datum≥T.sub.1 and <T.sub.1 is randomly selected from a historical data set T of the voltage sag duration time.

[0145] (2) Calculating a conditional entropy E(T,X) between the four conditional attributes and T.sub.sag<T.sub.1:

[00016]$\begin{array}{cc}E\left(T,X\right)=\underset{c\in X}{.Math.}P\left(c\right)E\left(c\right)& \left(14\right)\end{array}$

[0146] wherein X represents four conditional attributes; c represents a conditional attribute; P(c) is a joint probability that a conditional attribute and T.sub.sag <T.sub.1 appear at the same time; and E(c) is a conditional probability of T.sub.sag<T.sub.1 under a conditional attribute and with different values.

[0147] (3) Calculating mutual information Gain(T,X) between the four conditional attributes and T.sub.sag<T.sub.1:

Gain(T,X)=E(T)−E(T,X)   (15)

[0148] The larger the value of mutual information is, the higher the correlation with T.sub.sag<T.sub.1 is. A conditional attribute with largest mutual information is selected as a decision tree root node.

[0149] Step 3: building decision tree branch nodes and leaf nodes

[0150] A specific operation process is the same as three points in step 2: calculating the information entropy, the conditional entropy and the mutual information; gradually discretizing results with the largest mutual information with T.sub.sag<T.sub.1 from results by using the conditional attributes, and attributes with largest mutual information with T.sub.sag<T.sub.1 from other conditional attributes as branch nodes; and iterating the process repeatedly until a complete decision tree with “Yes, No” of T.sub.sag<T.sub.1 as a leaf node is finally constructed based on historical data.

[0151] Step 4: conducting classification by the decision tree

[0152] When a monitoring device monitors a voltage sag event, inputting four conditional attribute data, and classifying the duration time of the new voltage sag event by a generated decision tree logic; and when the input data is consistent with the decision tree, entering a next branch for judgment, until a classification result of yes or no is obtained finally through the leaf node.

[0153] Module III: SSTS and DVR coordinated action strategy module

[0154] The module is an SSTS and DVR coordinated action strategy module. The output result of module I is used to determine the grouping mitigation solution for the sensitive loads. When the voltage sag event is monitored, the module II is used to output T.sub.sag classification. Based on the output results of module I and module II, when voltage sag is monitored, the action strategy of SSTS and DVR is formulated as follows:

[0155] (1) If T.sub.sag<T.sub.1, SSTS acts; and if T.sub.sag≥T.sub.1, SSTS does not act.

[0156] (2)U.sub.sag≤U.sub.tolercance-i, and the i-th group of DVR acts. U.sub.tolercance-i is the minimum value of the voltage sag tolerance amplitude in the i-th group of sensitive loads.

[0157] To sum up, the solution of the present invention is summarized as follows:

[0158] 1) For the problems that whether the sensitive load should be governed and mitigation is conducted by SSTS or DVR, the present invention proposes a grouping method for sensitive loads with consideration of the whole industrial process of the user. From the perspective of the probability that a single sensitive load trips and causes interruption of the whole industrial process of the user, the method divides the loads into two categories based on the operating characteristics of SSTS and DVR;

[0159] 2) For the sensitive loads governed by DVR, the present invention proposes a grouping mitigation optimization model for the sensitive loads. The model takes the minimum sum of capacities of installed DVR and the minimum interruption probability of the industrial process of the user as the goals, and considers four constraints. The NSGA-II algorithm and the slightly small fuzzy satisfaction function are used to finally determine the grouping solution and the compensation voltage of each group;

[0160] 3) For the problem that whether the SSTS acts depends on the key factor of the voltage sag duration time, the present invention proposes a method for classifying T.sub.sag through the decision tree, and finally outputs the classification of yes or no for T.sub.sag<T.sub.1;

[0161] 4) Based on the above three points, the present invention finally proposes a coordinated action strategy of SSTS and DVR for voltage sag mitigation. The action basis of SSTS is determined through points 1) and 3), and the action basis of each group of DVRs is determined through points 1) and 2).