Active power distribution network operation state information mining and prediction system and method

Abstract

The application provides an active power distribution network operation state information mining and prediction system and method, at least comprising: a collection module: based on a distributed measurement principle, collecting power distribution network different node electronic equipment operation state information and collecting static information including power grid topology and line impedance; a processing module: for processing the operation state information, including effective data extraction and bad data identification and cleaning; finding key features of static information and updating the state; a prediction module: for state prediction through the operation state information obtained by the processing module, the state prediction at least including one of load prediction, new energy output prediction and power distribution network operation state prediction. The optimal state of the distributed power supply access in the power distribution network can be predicted according to the analysis result.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of electrical automation, and in particular to a system and method for mining and predicting operation state information of an active power distribution network. BACKGROUND

[0002] The power distribution network is an intermediate bridge connecting the large power grid and the power users, and has the most direct influence on the reliability of user power supply. Most of the power failure of users is caused by the failure of the power distribution network, and thus the reliable and economic operation of the power distribution network has an important influence on the stability of the entire power system. The active power distribution network refers to a power distribution network with a large number of distributed power sources and bidirectional power flow, and is also called a proactive power distribution network. At present, with the access of new energy equipment and power electronic equipment, the network structure of the power distribution network is constantly changing. Meanwhile, with the development of sensing technology and communication technology, a large number of sensing, monitoring and communication equipment are connected to the active power distribution network, which can efficiently obtain the operation state information of the power distribution network, but also easily suffers from network attacks from the outside world, increasing the difficulty of the safety and stability analysis of the power distribution network.

[0003] The data directly collected by the measurement system has problems such as redundancy, noise interference and poor quality, and the access of different distributed complex node new energy equipment increases the difficulty of mining and analyzing the operation state of the power distribution network, and at the same time, there is a lack of a systematic and practical operation state evaluation and prediction system that can be directly applied to power grid analysis. Therefore, a new scheme suitable for mining and analyzing the operation state information of the active power distribution network is urgently needed. SUMMARY

[0004] In order to solve the problems existing in the prior art and meet the actual needs of the operation of the power distribution network, the present application provides a system and method for mining and predicting operation state information of an active power distribution network. Based on the current situation of the active power distribution network, a method for mining and analyzing the operation state information of the power distribution network is proposed, and the optimal state of the distributed power source access in the power distribution network can be predicted according to the analysis result. The power grid dispatching and operation personnel can determine the current operation state of the power grid according to the method of the present application, and prepare relevant power grid operation and engineering plans, so as to optimize the power grid structure. The present application is suitable for various types of active power distribution networks, including medium-voltage and low-voltage active power distribution networks.

[0005] The present application specifically adopts the following technical solutions:

[0006] A system for mining and predicting operation state information of an active power distribution network comprises:

[0007] The acquisition module is used to acquire the operation state information of the electronic equipment of different nodes of the power distribution network and collect static information including the power grid topology and line impedance based on the distributed measurement principle;

[0008] Processing module: used for processing the operation state information, including extraction of valid data and identification and cleaning of bad data; finding key features of static information and updating state;

[0009] Prediction module: used for state prediction based on the operation state information obtained by the processing module, the state prediction at least including one of load prediction, new energy output prediction and distribution network operation state prediction.

[0010] As a more optimal selection, generally further comprising: an evaluation module; used for classifying the operation state information and static information processed by the processing module to form an index type system, on the basis of rationality test, establishing the weight of each index based on the combination method of TFAHP-anti-entropy weight method, and then optimizing the weight by using the variable weight theory method to evaluate the operation of the distribution network.

[0011] Further, in the collection module, if there is a problem of insufficient measurement configuration of the distribution network, the measurement redundancy is improved by using pseudo measurement or virtual collection method:

[0012] For the active distribution network, the specific method of the virtual collection is:

[0013] If the branch current and the voltage at one end of the branch are known, the voltage at the other end of the branch is virtually measured;

[0014] If the branch current and the voltage at one end of the branch are known, the voltage at the other end of the branch is virtually measured;

[0015] If all the adjacent node voltage vectors of the unknown injection node are known, the voltage of the node can be virtually measured by the node equation;

[0016] If the currents of all the branches except one branch of the node are known, the current of the unknown branch is virtually measured;

[0017] The pseudo measurement includes load prediction pseudo measurement and distributed power output prediction pseudo measurement.

[0018] Further, the method of identification and cleaning of bad data in the processing module is specifically:

[0019] The backtracking method is used to identify bad data, including empty data, abnormal data and repeated data;

[0020] For empty data, one of the following two methods is used:

[0021] Empty: directly deleting from the data set;

[0022] Completion: using Lagrange interpolation method for interpolation processing: according to the empty data x iFor the first two records and the last record, establish a second-order equidistant Lagrange interpolation equation to perform interpolation on the empty data;

[0023] For outlier data, the Wright test is used to process the data values ​​and then evaluate them. If |V i If |>3σ, then the error is a gross error, and the data is considered abnormal and should be deleted.

[0024] For duplicate data, merge them directly;

[0025] After cleaning the raw data, the entire dataset is subjected to data quality testing using the Wright test. The processing method for abnormal data is repeated until the data quality Q meets the preset accuracy.

[0026] Furthermore, the specific steps of the evaluation module based on the TFAHP-anti-entropy weight method combination method are as follows:

[0027] Construct a fuzzy judgment matrix for the obtained indicator system;

[0028] For a factor at level k-1, its subordinate indicators at level k are compared pairwise, and the fuzzy judgment matrix is ​​obtained by quantitatively representing it using triangular fuzzy numbers. The element a of the fuzzy judgment matrix ij =(l ij ,m ij ,u ij ) is a word with m ij Let n be a closed interval containing the median. k Let be the size of the matrix, i and j be the coordinates of the elements, l ≤ m ≤ u, and l and u be the lower and upper bounds supporting a, respectively.

[0029] Calculate the overall score:

[0030] The comprehensive triangular fuzzy number of the k-th layer is:

[0031]

[0032] In the formula, X represents the number of experts, thus obtaining the comprehensive judgment matrix of all factors in the k-th layer on each factor in the (k-1)-th layer;

[0033]

[0034] Hierarchical single sorting:

[0035] Based on the fundamental properties of triangular fuzzy numbers, we obtain Possibility:

[0036]

[0037] but

[0038]

[0039] denotes the single ordering of the kth layer on the (k-1)th layer of the hth factor, denotes the i th factor on the kth layer, obtained from the above After normalization, we get:

[0040]

[0041] That is, the relative weight of the kth layer of each factor on the (k-1)th layer of the hth factor, that is, the subjective weight;

[0042] The objective weight is determined by the anti-entropy weight method:

[0043] Suppose there are m evaluation objects and n evaluation indexes, and the index value is x ij (i = 1, 2,..., n; j = 1, 2,..., m), and the evaluation matrix is X = (x ij )n×m; According to the evaluation matrix X, the anti-entropy of each index is determined;

[0044]

[0045] In the formula, According to the anti-entropy value, the objective weight of each index is further determined:

[0046]

[0047] Then the combined weight

[0048] ω i =sP i +(1-s)Q i

[0049] In the formula, s is the subjective coefficient, which is taken in 0-1.

[0050] Further, in the process of weight optimization, a rationality check is adopted, and the standard is that the remaining indexes screened can reflect more than 90% of the initial index information, and the information contribution rate is used to measure, and the calculation formula is:

[0051]

[0052] In the formula: σ j is the variance of the index; m is the total number of indexes in the screened index system; h represents the total number of initial indexes, and the threshold value of IN is 90%;

[0053] If the power distribution network is transformed or abnormal, the variable weight theory method is used for further optimization, and the calculation formula is as follows:

[0054]

[0055] In the formula, ω i is the variable weight value of the i-th index, is the original weight value of the i-th index, x i is the normalized value of the i-th index, n is the number of indexes; μ is the equilibrium factor.

[0056] Further, the prediction module performs state prediction for the source, load and network respectively.

[0057] The load prediction adopts LSSVM-RBF for ultra-short-term prediction. First, the data is trained by the least square support vector machine LSSVM; then the prediction error of the LSSVM predicted value is predicted by using the RBF neural network.

[0058] The prediction error of the same time of the previous three days is taken as the input to predict the error of the same time of the day; the prediction error of the LSSVM is collected as the sample of the RBF neural network for further training, and on this basis, the load at time t is predicted.

[0059] For the prediction of new energy output and distribution network operation state, the historical data of new energy output and distribution network operation state are combined to use the same LSSVM-RBF method for short-term new energy output and distribution network operation state prediction.

[0060] And a method for mining and analyzing the operation state information of an active distribution network, comprising the following steps:

[0061] Based on the principle of distributed measurement, the operation state information of the electronic equipment of different nodes of the distribution network is collected, and the static information including the power grid topology and line impedance is collected;

[0062] The collected state information is processed, including extraction of valid data and identification and cleaning of bad data; the key features of the static information are found and the state is updated;

[0063] The operation state information obtained by processing is used for state prediction, which at least includes one of load prediction, new energy output prediction and distribution network operation state prediction.

[0064] Among them, generally, the operation state information at least includes: the voltage of each node, the branch current, the active power and the reactive power of the branch, the voltage, the current, the active power and the reactive power of the distributed power grid connection point, the line, the load rate of the distribution transformer;

[0065] The static information at least includes: power distribution network architecture, line parameters, line switch protection action time, average power supply radius, operation life, insulation rate, automation coverage, tie-in rate, inter-station tie-in rate, equipment operation life, type, power outage information, distributed power supply parameters, type, installed capacity, grid connection rate, heavy and light load distribution transformers, line proportion, distribution transformer, line maximum load rate, FA correct action rate, FA operation rate, automation terminal coverage rate, automation terminal online rate, system outage time, CO2 and SO2 gas emissions;

[0066] According to the collected operation state information and the static information, the power distribution network operation state information is obtained, at least including: line loss rate, low voltage user proportion, average three-phase imbalance rate, voltage qualification rate, low voltage total harmonic distortion rate, voltage drop percentage, load transferable percentage, system average outage frequency, equipment line fault rate, fault average outage proportion, outage operation on-time power restoration rate, power supply fault healing rate;

[0067] After processing, the power distribution network operation state information is obtained by combining the collected operation state information and the static information, at least including: line loss rate, low voltage user proportion, average three-phase imbalance rate, voltage qualification rate, low voltage total harmonic distortion rate, voltage drop percentage, load transferable percentage, system average outage frequency, equipment line fault rate, fault average outage proportion, outage operation on-time power restoration rate, power supply fault healing rate.

[0068] Corresponding to the system, as preferred, the method also includes the following steps: classifying the processed operation state information and the static information to form an index type system, on the basis of rationality inspection, establishing the weight of each index based on the TFAHP-anti-entropy weight combination method, and then optimizing the weight by using the variable weight theory method to evaluate the operation of the power distribution network.

[0069] Further, the load prediction uses LSSVM-RBF for ultra-short-term prediction. First, the data is trained by the least squares support vector machine LSSVM; then the prediction error of the LSSVM predicted value is predicted by using the RBF neural network;

[0070] The prediction error of the same time of the previous three days is used as the input to predict the error of the same time of the day; the prediction error of the LSSVM is collected as the sample of the RBF neural network for further training, and on this basis, the load at time t is predicted;

[0071] For the prediction of new energy output and the operation state of the power distribution network, the same LSSVM-RBF method is used for short-term prediction of new energy output and the operation state of the power distribution network by using the historical data of new energy output and the operation state of the power distribution network.

[0072] Compared with the prior art, the present application and its preferred schemes have at least the following beneficial effects and advantages:

[0073] 1. Based on the current situation of the active power distribution network, an operation state information mining and analysis system architecture and scheme of the active power distribution network are provided, covering all architecture types of the existing power distribution network. Dispatching and operation personnel can mine the operation state information of the power distribution network according to the method of the present application, and flexibly find the optimal solution of the optimal, most reliable power distribution network architecture, load, distributed energy penetration rate and installation location.

[0074] 2. It is proposed that, while collecting the operation state of the power distribution network through distributed measurement, in combination with line impedance, power grid architecture and other static information, in view of the problem of insufficient measurement configuration, the method of adopting pseudo-measurement and virtual collection is used to improve the measurement redundancy, which can improve the accuracy of the method of analyzing the state of the power distribution network and the observability of the power distribution network. In the process of processing information, the method of adopting the Bezier formula and the Lett test is proposed for different types of bad data, which can solve the problems of data redundancy, noise interference and poor quality of the power distribution network, and improve the reliability of the data.

[0075] 3. In the evaluation of the state of the power distribution network, the TFAHP-anti-entropy weight method combination method is adopted, which is improved on the basis of the traditional AHP and entropy weight method. In combination with subjective weight and objective weight, the subjectivity caused by human evaluation is avoided, and the independence and integrity between the indexes are optimized, realizing the effective combination of expert experience and objective theory. On the basis of obtaining the preliminary weight, the variable weight theory is used for optimization, which can make up for the shortcoming that the original weight cannot change with the operation state, and improve the effectiveness and rationality of the output results of the evaluation module.

[0076] 4. In the prediction module, improved prediction methods are provided for the source, load and network, respectively, which improve the universality and accuracy of the prediction of the state information of the power distribution network. The contents of the state information collection module, the processing module, the evaluation module and the prediction module are innovatively combined, and the operation state information mining and analysis method is further proposed based on the active power distribution network with a large number of distributed power sources, which has strong practicality and can be directly applied to the actual power distribution network construction. BRIEF DESCRIPTION OF DRAWINGS

[0077] The present application will be further described in detail below in combination with the drawings and specific embodiments:

[0078] Figure 1 Fig. 1 is a schematic diagram of the topology structure of the active power distribution network corresponding to the virtual collection in the first embodiment of the present application;

[0079] Figure 2 Fig. 2 is a schematic diagram of the simple power distribution network topology structure corresponding to the pseudo-measurement in the first embodiment of the present application;

[0080] Figure 3 The active power distribution network operation state information evaluation system schematic diagram provided in the second embodiment of the present application;

[0081] Figure 4 The method flow chart in the second embodiment of the present application;

[0082] Figure 5 The working flow chart of the prediction module in the third embodiment of the present application. DETAILED DESCRIPTION

[0083] In the following, the specific embodiments of the present application will be described in detail with reference to the accompanying drawings, and the skilled in the art can clearly understand the present application and implement the present application according to these detailed descriptions. The features in each different embodiment can be combined to obtain new implementation modes, or replace some features in some embodiments to obtain other preferred implementation modes, without departing from the principles of the present application.

[0084] It should be noted that the terms used herein are only for describing specific embodiments, and are not intended to limit the exemplary embodiments according to the present application. As used herein, the singular form is intended to include the plural form, unless the context clearly indicates otherwise, and it should also be understood that when the terms "comprise" and / or "include" are used in the specification, there is a feature, step, operation, device, component and / or combination thereof.

[0085] In order to make the features and advantages of the present patent more obvious and easy to understand, the following specific embodiments are described in detail as follows:

[0086] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present application should be understood as the general meaning understood by those skilled in the art to which the present application belongs. The terms "first", "second" and similar words used in the present application do not represent any order, number or importance, but are only used to distinguish different components. The terms "include" or "contain" and the like mean that the elements or objects before the word cover the elements or objects listed after the word and their equivalents, and do not exclude other elements or objects. The terms "connect" or "connect" and the like are not limited to physical or mechanical connection, but can include electrical connection, whether direct or indirect. The terms "up", "down", "left", "right" and the like are only used to represent relative positional relationship, and when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0087] Firstly, in the system scheme provided in the embodiments of the present application, four modules are included: a state information acquisition module, a processing module, an evaluation module, and a prediction module, wherein:

[0088] The collection module: based on the distributed measurement principle, collects the relevant operation state information of the intelligent electronic device of different nodes of the power distribution network, and collects the static information such as power grid topology and line impedance.

[0089] The processing module: processes the state information collected by the collection module, including extraction of valid data, identification and cleaning of bad data; and finds key features for static information (power grid topology, line impedance, etc.) and updates the status in time.

[0090] The evaluation module: comprehensively classifies the state information and static information processed by the processing module, forms an index system covering different aspects of the power distribution network, performs rationality test, establishes the weight of each index based on the TFAHP-anti-entropy weight method combination method, and finally optimizes the weight by using the variable weight theory method, so as to measure the operation of the power distribution network. The running result of the power distribution network is used to measure the operation of the power distribution network.

[0091] The prediction module: based on the operation state information obtained by the evaluation module, the state of the source, load and network is predicted, including load prediction, new energy output prediction and power distribution network operation state prediction, which helps the dispatching and operation personnel to control and schedule in advance, and improves the economy and reliability of the active power distribution network.

[0092] The evaluation module is a preferred module of the present application, and the present application can still achieve the basic purpose of the application without the module. The function of the module makes the function of the whole system more complete, and the output result has stronger reference value for the operation personnel.

[0093] As a preferred scheme of the present embodiment, in the collection module, if there is a problem of insufficient measurement configuration of the power distribution network, pseudo measurement and virtual collection can be used to improve the measurement redundancy and enhance the observability of the power distribution network.

[0094] There are four types of virtual collection for the active power distribution network.

[0095] 1. If the voltage of one end node of a branch and the current of the branch are known, the voltage of the other end node of the branch is virtually measured.

[0096] 2. If the voltages of both end nodes of a branch are known, the current of the branch is virtually measured.

[0097] 3. If all adjacent node voltage vectors of an unknown injection node are known, the voltage of the node can be virtually measured by the node equation.

[0098] 4. If the currents of all branches except one branch of a node are known, the current of the unknown current branch is virtually measured.

[0099] Pseudo-measurements are mainly divided into two categories: load forecast pseudo-measurements and distributed generation output forecast pseudo-measurements. Load forecast pseudo-measurements are prone to error due to the large and complex nature of the distribution network load. Distributed generation output forecast pseudo-measurements can be obtained by combining weather forecasts with historical data.

[0100] As a preferred embodiment, the method for identifying and cleaning bad data in the processing module is as follows:

[0101] The backtracking method is used to identify bad data, which includes empty data (data that was not successfully recorded), abnormal data (data whose recorded value is outside the normal range), and duplicate data (repeated data).

[0102] Two methods are used for empty data.

[0103] 1. Remove empty strings. Delete them directly from the dataset.

[0104] 2. Completion. If the corresponding data is important, use Lagrange interpolation for interpolation: based on the empty data x... i Using the first two records and the last record, establish a second-order isometric Lagrange interpolation equation P2(x) i ), to perform interpolation on empty data.

[0105]

[0106] For outlier data, the Wright test is used as follows:

[0107] Export the data values ​​and calculate the data values ​​(X1, X2, ..., X). n mean With residual V i And calculate the standard deviation σ according to Bessel's formula.

[0108]

[0109] Then, the data values ​​are evaluated; if |V i If |>3σ, then the error is a gross error, and the data is considered abnormal and should be deleted.

[0110] For duplicate data, perform a direct check and merge.

[0111] After cleaning the original data, the entire dataset needs to be tested for data quality. The Wright test is used to repeat the process for handling outlier data until the data quality Q meets the required precision. Here, the precision value is set to 95%.

[0112]

[0113] In the formula, N represents the total number of data that satisfy the Wright criterion, and M represents the total number of data.

[0114] The evaluation module can be based on various indicators, including economic efficiency, power supply reliability, power quality, network resilience, sustainable development, and grid security.

[0115] As a preferred embodiment, the evaluation module is based on the TFAHP-anti-entropy weight method combination method, and the specific steps are as follows:

[0116] After establishing the indicator system, construct the fuzzy judgment matrix.

[0117] For a factor in layer k-1, its subordinate indicators in layer k are compared pairwise, and the fuzzy judgment matrix is ​​obtained by quantitative representation using triangular fuzzy numbers. The element a of the fuzzy judgment matrix ij =(l ij ,m ij ,u ij ) is a word with m ij Let m be a closed interval containing the median. ij It is usually taken as an integer between 1 and 9.

[0118] Calculate the overall score.

[0119] The comprehensive triangular fuzzy number of the k-th layer:

[0120]

[0121] In the formula, X represents the number of experts, from which the comprehensive judgment matrix of all factors in the k-th layer on each factor in the (k-1)-th layer can be obtained.

[0122]

[0123] Hierarchical single sorting.

[0124] Based on the fundamental properties of triangular fuzzy numbers, we can obtain... Possibility

[0125]

[0126] but

[0127]

[0128] This indicates a single ranking of the h-th factor in the (k-1)-th layer from the factor in the k-th layer. This represents the i-th factor at the k-th level, obtained from the above. After normalization, we can obtain:

[0129]

[0130] That is, the relative weight of the kth layer to the (k-1)th layer hth factor, namely the subjective weight.

[0131] The objective weight is determined by the anti-entropy weight method.

[0132] Suppose that there are m evaluation objects and n evaluation indexes in the evaluation problem, and the index value is x ij (i = 1, 2,..., n; j = 1, 2,..., m), and the evaluation matrix is X = (x ij )n×m. According to the evaluation matrix X, the anti-entropy of each index is determined.

[0133]

[0134] In the formula, The objective weight of each index is further determined according to the anti-entropy value

[0135]

[0136] The combined weight is

[0137] ω i =sP i +(1-s)Q i

[0138] In the formula, s is a subjective coefficient, which can be 0-1.

[0139] As a preferred scheme of the embodiment, the prediction module performs state prediction for three types of sources, loads and networks, and different methods are used for different aspects. The active power distribution network includes a medium and low voltage distribution network with different types, different penetration rates and different capacities of distributed power sources such as small hydropower, distributed photovoltaic power, wind power and energy storage power.

[0140] The specific process is specifically introduced in the following embodiment three.

[0141] The key technical details of the scheme of the application are further introduced through the following three specific embodiments:

[0142] Embodiment one

[0143] In the acquisition module, if there is a problem of insufficient measurement configuration of the power distribution network, a virtual acquisition method can be used to obtain Figure 1 As shown in the figure, the architecture diagram of the power distribution network of this example contains multiple different types of distributed power sources, and then the following can be obtained:

[0144] 1) The known node voltage and the branch current The node voltage is a virtual measurement.

[0145]

[0146] 2) Known node voltage and then branch current is a virtual measurement.

[0147]

[0148] 3) Known node voltage and load injection current and photovoltaic source current then node voltage is a virtual measurement.

[0149]

[0150] 4) Known branch current load injection current and then branch current is a virtual measurement.

[0151]

[0152] The virtual measurement accuracy is high, and it can be treated as real-time measurement.

[0153] Similarly, the pseudo measurement method can be combined, and if there is no historical data, the following method can be used. As shown in Figure 2 , the distribution network of this example is a simple distribution network topology diagram, P G is the root node injection measurement; P T1 , P T2 , P T3 , P T4 , P T5 , P T6 are load measurements:

[0154] PT1=KT1(t)PG(t)

[0155] Where K T1 (t) is a load distribution coefficient that changes over time, which can be processed as follows.

[0156]

[0157] P T1 (t) can be determined in combination with the capacity of different distribution transformers and the load type. Since the time and seasonal changes will cause the load type to change, the K T1 (t) corresponding to each load can be obtained through the load curve corresponding to different typical weather conditions.

[0158] The beneficial effect of the embodiment scheme is that, considering that due to the quality of the acquisition equipment, the data has problems such as redundancy, noise interference, and poor quality, the rationality and applicability of the data can be improved through virtual measurement and pseudo measurement.

[0159] Embodiment two

[0160] The present application is applicable to various types of power distribution networks, especially active power distribution networks with a large number of distributed power sources. Figure 3 and Figure 4 As shown in the system design based on the overall scheme, a method for mining and analyzing the operating state information of an active power distribution network is further provided, including the following steps:

[0161] A1, based on the principle of distributed measurement, collecting the relevant operating state information of the intelligent electronic device of different nodes of the power distribution network, collecting the static information such as power grid topology and line impedance based on the device and record.

[0162] Among them, the operating state information includes: the voltage of each node, the branch current, the active power and the reactive power of the branch, the voltage, the current, the active power and the reactive power of the distributed power grid connection point, the load rate of the line and the distribution transformer, etc.

[0163] The static information includes: the power distribution network architecture, the line parameters (the line length, the type (the impedance and the capacitive reactance of each type of line) in the active power distribution network), the line (the switch protection action time, the average power supply radius, the operation time, the insulation rate, the automation coverage rate, the connection rate, the inter-station connection rate), the equipment (the operation time, the type, the power outage information), the distributed power source (the parameter, the type, the installed capacity, the grid connection rate), the heavy and light load distribution transformer, the proportion of the line, the highest load rate of the distribution transformer and the line, the FA correct action rate, the FA operation rate, the automation terminal coverage rate, the automation terminal online rate, the system power outage time, the CO2 and SO2 gas emission, etc.

[0164] The above-mentioned operating state information and static information can obtain other required power distribution network operating state information, including the line loss rate (theoretical line loss rate, fixed line loss rate), the proportion of low-voltage users, the average three-phase imbalance rate, the voltage qualification rate (bus voltage qualification rate, substation voltage qualification rate, user voltage qualification rate), low-voltage total harmonic distortion rate, voltage drop percentage, load transfer percentage, system average power outage frequency, equipment line fault rate, fault average power outage proportion, power outage operation on-time power restoration rate, power supply fault healing rate, etc.

[0165] A2, processing the collected state information, including extracting valid data, identifying and cleaning bad data; and finding key features for static information (power grid topology, line impedance, etc.) and updating the current situation in time.

[0166] The identification and cleaning of bad data adopts Lagrange interpolation method and Wright test method respectively for different types of bad data, and after cleaning by the above methods, the obtained data is subjected to Wright test method again, and then the data precision value is set to 95%, so that the obtained data meets the relevant requirements.

[0167] A3, the processed state information and static information are comprehensively classified, an index type system covering different aspects of the distribution network is formed, the weight of each index is established based on the TFAHP-anti-entropy weight combination method, and finally the weight is optimized by using the variable weight theory, so that the operation result of the distribution network is comprehensively evaluated to measure the operation of the distribution network.

[0168] The index type system given by the preferred embodiment of the present application is as shown in the figure Figure 3 The index type system given by the preferred embodiment of the present application is as shown in the figure

[0169] At this time, the rationality of the index system can be checked, the standard is that the retained index after screening can reflect more than 90% of the initial index information, that is, it is considered effective and reasonable, and the information contribution rate can be used to measure, and the calculation formula is

[0170]

[0171] In the formula, σ j is the variance of the index, m is the total number of indexes in the index system after screening, and h represents the total number of initial indexes, and the threshold value of IN can be set to 90%.

[0172] If the above index weight cannot truly reflect the operation state of the distribution network due to the reconstruction or abnormality of the distribution network, the variable weight theory method is used for the last step of optimization on this basis, and the accuracy of the result is improved, and the calculation formula is as follows:

[0173]

[0174] In the formula, ω i is the variable weight value of the i-th index, is the original weight value of the i-th index, x i is the normalized value of the i-th index, n is the number of indexes, and μ is the balance factor, which can be taken as 0.5 in general cases.

[0175] A4, through the operation state information obtained by the above excavation, the state prediction is carried out for the source, load and network, including load prediction, new energy output prediction and distribution network operation state prediction, the specific method can be seen in embodiment 3, through the state prediction, the related adjustment of the distribution network can be helped to be carried out in advance by the dispatching and operation personnel, and the economy and reliability of the active distribution network operation can be improved.

[0176] Embodiment three

[0177] Further, the load prediction in the prediction module of the scheme can adopt LSSVM-RBF for ultra-short-term prediction, and the least square support vector machine (LSSVM) is excellent in training speed, and the main process steps are as follows:

[0178] Given the training set S={(x i , y i ), i=1, 2…, l}, wherein x i is the i-th input vector, y i is the target value corresponding to x i , and l is the number of samples. The regression problem is to determine the optimal regression function f(x), wherein ω is the weighting vector, b is the constant bias, and x is the input variable matrix.

[0179] f(x) = ω T x + b

[0180] In the least square support vector machine, the optimization problem corresponding to the regression problem is shown in the following formula, e is the slack variable, and γ is the penalty coefficient.

[0181]

[0182]

[0183] The corresponding Lagrange function of the least square support vector machine and the optimal value condition thereof are as follows, wherein α is the Lagrange coefficient:

[0184]

[0185] The Gaussian radial basis RBF kernel is selected as the kernel function, as shown below, and σ is the kernel function parameter.

[0186]

[0187] The above formula is written in matrix form and ω and e are eliminated to obtain the model of the least square support vector machine regression

[0188]

[0189] Then, the prediction error of the LSSVM prediction value is predicted by using the RBF neural network.

[0190] The RBF neural network is generally divided into three layers of input layer, hidden layer and output layer. The input layer is composed of input nodes of samples, only plays a role of sample data transmission, the hidden layer is generally composed of neurons with Gaussian radial basis function G(x) as kernel function, the kernel function can realize mapping of samples from low-dimensional linear non-separable space to high-dimensional linear separable space, the number of neurons is determined by specific problems and network types, is generally set to (N+2) times of input layer data type, and the output layer is generally described by a simple linear function, and the final prediction result is obtained by linear weighting calculation of outputs of the hidden layer neurons.

[0191] The core of the RBF neural network is the hidden layer part, and the kernel function generally used in the hidden layer is a Gaussian function, and the formula is as follows:

[0192]

[0193] In the formula, R i is an output value of the i-th unit of the hidden layer neuron, x is an input sample, c i is a center vector of the i-th unit of the basis function with the same dimension as the input sample, and sigma is the width of the basis function. i

[0194] The prediction error of the same time point of the previous three days is taken as the input to predict the error of the same time point of the day. The prediction error of the LSSVM is collected as the sample of the RBF neural network for a large amount of training, and on this basis, the load at time t is predicted, and the basic process is as shown in Figure 5 .

[0195] The prediction method of the new energy output and the operation state of the power distribution network is similar to the load prediction, and the historical data of the new energy output and the operation state of the power distribution network can be combined to use the LSSVM-RBF method to predict the short-term new energy output and the operation state of the power distribution network.

[0196] The embodiments of the application are described in detail in combination with the drawings, but the application is not limited to the described embodiments. For those skilled in the art, various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirits of the application, and still fall within the protection scope of the application.

[0197] The above system and method provided by the embodiment can be stored in a computer readable storage medium in a coded form, and be realized in a computer program manner, and the basic parameter information required for calculation is input by computer hardware, and the calculation result is output.

[0198] ​Those skilled in the art will appreciate that embodiments of the application can be readily used as a method, a system or a computer program product. Accordingly, the application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the application can take the form of a computer program product on one or more computer readable storage media (including, but not limited to, disk memory, CD-ROMs, optical storage devices, etc.) embodying computer readable program code.

[0199] The present application is described in reference to the flowchart and / or block diagrams of the method, apparatus (system) and computer program product according to embodiments of the application. It will be understood that each block of the flowchart and / or block diagrams, and combinations of blocks in the flowchart and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can 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 and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0200] These computer program instructions can 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 instructions which implement the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0201] These computer program instructions can 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 flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0202] Finally, it should be noted that the above-mentioned embodiments are merely intended for describing and illustrating, not limiting, the technical solutions of the present application. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that the specific embodiments of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, and any modification or equivalent replacement without departing from the spirit and scope of the present application should be covered in the protection scope of the claims of the present application.

[0203] The patent is not limited to the above best mode, anyone can draw other various forms of active power distribution network operation state information mining and prediction system and method under the inspiration of the patent, any equivalent changes and modifications made in the patent application scope shall belong to the scope of the patent.

Claims

1. A system for mining and predicting the operating status information of an active power distribution network, characterized in that, include: Acquisition module: Used to collect operating status information of electronic equipment at different nodes of the distribution network based on the principle of distributed measurement, as well as collect static information including power grid topology and line impedance; Processing module: Used to process the running status information, including the extraction of valid data and the identification and cleaning of bad data; Find key features in static information and update its status; Prediction module: used to perform status prediction based on the operating status information obtained by the processing module. The status prediction includes at least one of load prediction, new energy output prediction, and distribution network operation status prediction. Evaluation module; This is used to classify the operational status information and static information processed by the processing module to form an index-based system. Based on the rationality test, the weights of each index are established based on the TFAHP-anti-entropy weight method combination method, and then the weights are optimized using the variable weight theory method to evaluate the operation of the distribution network. If there is insufficient distribution network measurement configuration in the acquisition module, then pseudo-measurement or virtual acquisition methods are used to improve measurement redundancy. For active power distribution networks, the specific method for virtual data acquisition is as follows: Given the node voltage and current at one end of a branch, the node voltage at the other end of the branch is virtually measured. Given the voltages at both ends of a branch, the branch current is virtually measured. When the voltage vectors of all adjacent nodes of an unknown injection node are known, the voltage of this node can be virtually measured by the node equations. If the current of all branches except one of the nodes is known, then the current of the unknown current branch is virtually measured. The pseudo-measures include: load forecast pseudo-measures and distributed generation output forecast pseudo-measures. The specific steps of the TFAHP-anti-entropy weight method combination method in the evaluation module are as follows: Construct a fuzzy judgment matrix for the obtained indicator system; For a factor at level k-1, its subordinate indicators at level k are compared pairwise, and the fuzzy judgment matrix is ​​obtained by quantitatively representing it using triangular fuzzy numbers. The element a of the fuzzy judgment matrix ij =(l ij ,m ij ,u ij ) is a word with m ij Let n be a closed interval containing the median. k Let be the size of the matrix, i and j be the coordinates of the elements, l ≤ m ≤ u, and l and u be the lower and upper bounds supporting a, respectively. Calculate the overall score: The comprehensive triangular fuzzy number of the k-th layer is: In the formula, E represents the number of experts, thus obtaining the comprehensive judgment matrix of all factors in the k-th layer on each factor in the (k-1)-th layer; Hierarchical single sorting: Based on the fundamental properties of triangular fuzzy numbers, we obtain Possibility: but This indicates a single ranking of the h-th factor in the (k-1)-th layer from the factor in the k-th layer. This represents the i-th factor at the k-th level, obtained from the above. After normalization, we get: That is, the relative weight of each factor in the k-th layer to the h-th factor in the (k-1)-th layer, which is the subjective weight; Determining objective weights using the anti-entropy weight method: Suppose there are m evaluation objects and n evaluation indicators, with indicator values ​​x. ij Let i = 1, 2, ..., n; j = 1, 2, ..., m, and the evaluation matrix be X = (x ij n×m; Based on the evaluation matrix X, determine the inverse entropy of each index; In the formula, The objective weights of each indicator are further determined based on the inverse entropy value: Then the combined weight ω i =sP i +(1-s)Q i In the formula, s is a subjective coefficient, which takes values ​​between 0 and 1.

2. The active power distribution network operation status information mining and prediction system according to claim 1, characterized in that: The specific method for identifying and cleaning bad data in the processing module is as follows: A backtracking method is used to identify bad data, which includes empty data, abnormal data, and duplicate data. For empty data, use one of the following two methods; Remove empty: Delete directly from the dataset; Completion: Lagrange interpolation is used for interpolation processing: based on the empty data x i For the first two records and the last record, establish a second-order equidistant Lagrange interpolation equation to interpolate the empty data; For outlier data, the Wright test is used to process the data values ​​and then evaluate them. If |V i If |>3σ, then the error is a gross error, and the data is considered abnormal and should be deleted. For duplicate data, merge them directly; After cleaning the raw data, the entire dataset is subjected to data quality testing using the Wright test. The processing method for abnormal data is repeated until the data quality Q meets the preset accuracy.

3. The active power distribution network operation status information mining and prediction system according to claim 1, characterized in that: The weight optimization process employs a rationality check, with the standard being that the selected and retained indicators can reflect more than 90% of the initial indicator information. This is measured using the information contribution rate, calculated using the following formula: Where: σ j denoted as σ0, z is the total number of indicators in the screened indicator system; H represents the initial total number of indicators, and the threshold for IN is 90%. If the problem is due to power distribution network upgrades or anomalies, the variable weight theory method is used for further optimization. The calculation formula is as follows: In the formula, ω i It is the variable weight value of the i-th type of indicator. It is the original weight value of the i-th type of indicator, x i is the normalized value of the i-th type of indicator, N is the number of indicators, and μ is the equilibrium factor.

4. The active power distribution network operation status information mining and prediction system according to claim 1, characterized in that: The prediction module performs state predictions for three types: source, load, and network. Load forecasting uses LSSVM-RBF for ultra-short-term forecasting. First, the data is trained using LSSVM (Least Squares Support Vector Machine); then, RBF neural network is used to predict the forecast error of the LSSVM prediction. The prediction error at the same time three days prior to the prediction date is used as input to predict the error at the same time on the current day; the prediction error of LSSVM is collected as samples for further training of RBF neural network, and load prediction is performed at time t based on this. For the prediction of renewable energy output and distribution network operation status, the same LSSVM-RBF method is used in combination with historical data of renewable energy output and distribution network operation status to make short-term predictions of renewable energy output and distribution network operation status.

5. A method for mining and analyzing the operating status information of an active distribution network, based on the active distribution network operating status information mining and prediction system described in any one of claims 1 to 4, characterized in that, Includes the following steps: Based on the principle of distributed measurement, the system collects the operating status information of electronic equipment at different nodes of the distribution network, as well as static information including the network topology and line impedance. Process the collected status information, including the extraction of valid data and the identification and cleaning of bad data; Find key features in static information and update its status; By processing the obtained operating status information, a status prediction is performed, which includes at least one of load prediction, new energy output prediction, and distribution network operating status prediction.

6. The method for mining and analyzing the operating status information of an active power distribution network according to claim 5, characterized in that: It also includes the following steps: The processed operational status information and static information are classified to form an index-based system. Based on the rationality test, the weights of each index are established using the TFAHP-anti-entropy weight method combination method. Then, the weights are optimized using the variable weight theory method to evaluate the operation of the distribution network.

7. The method for mining and analyzing the operating status information of an active power distribution network according to claim 5, characterized in that: The load forecasting uses LSSVM-RBF for ultra-short-term forecasting. First, the data is trained using LSSVM (Least Squares Support Vector Machine); then, RBF neural network is used to predict the forecasting error of the LSSVM forecast. The prediction error at the same time three days prior to the prediction date is used as input to predict the error at the same time on the current day; the prediction error of LSSVM is collected as samples for further training of RBF neural network, and load prediction is performed at time t based on this. For the prediction of renewable energy output and distribution network operation status, the same LSSVM-RBF method is used in combination with historical data of renewable energy output and distribution network operation status to make short-term predictions of renewable energy output and distribution network operation status.

Images