Device and method for evaluating power supply reliability of source-containing network load storage power distribution network

Abstract

The invention discloses a power supply reliability evaluation device and method for a load storage power distribution network with a source network, belongs to the technical field of power supply reliability evaluation, and can achieve active improvement and accurate management and control of a load end during operation through multidirectional analysis of an instant operation parameter value and a previous power supply reference value. And the reliability and coordination of the power distribution network to the load end power supply performance are improved. Through automatic distinctive analytic operation, the effect of the change of a power supply scene and the reliability of the load end on the power supply process of the power distribution network on the load end is efficiently introduced, and the small supernormal value in the numerical value is checked, so that the power supply of the power distribution network on the load end can execute feedback on the small supernormal value, and a management and control mode is configured; the continuous improvement of the power supply performance is guaranteed, the power supply change can be matched and the power supply process of the power distribution network to the load end can be continuously improved according to the response learning of the previous power supply value and the improvement of the control mode, and the power supply performance of the power distribution network to the load end is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of power supply reliability assessment, and in particular relates to a power supply reliability assessment device and method for a source-grid-load-storage distribution network. Background Art

[0002] As an important public infrastructure, the distribution network plays an important role in ensuring power supply, supporting economic and social development, and serving to improve people's livelihood. With the advancement of the construction of new power systems, the distribution network is gradually transforming from a power network that simply receives and distributes electric energy to users to a power network that aggregates and interacts with sources, grids, loads, and storage, and is flexibly coupled with the upper power grid. Its functions in promoting the nearby consumption of distributed power sources and carrying new loads are becoming increasingly prominent.

[0003] For the management and control of power supply of source-grid-load-storage-distribution network, the technical solution with patent publication number "CN210011617U" is often used to achieve the management and control of power supply of source-grid-load-storage-distribution network, which includes collecting power usage data at the load end and constructing a model to perform load power prediction Pload.

[0004] However, on the other hand, the current distribution network's control methods for the load end generally rely on pre-set constant parameters, lacking flexible configuration and automatic improvement functions, resulting in the distribution network often having poor matching with the load-end power supply method during the actual power supply period. In addition, under the influence of factors such as the application scenario of the load end and the degradation or wear of the load-end components themselves, the control method based on pre-set constant parameters is not easy to meet the high-reliability power supply requirements. Summary of the invention

[0005] In order to solve the defects in the prior art, the present invention proposes a power supply reliability assessment device and method for a source-grid-load-storage distribution network, which is suitable for performing abnormal detection on the target operating parameter values ​​during the operation of the load end, and can be used to find out the hidden abnormalities in the values. The load end is affected by scene factors and its own consumption during operation. There are often slight differences between the parameters sent during operation and the parameters of the actual operation of the load end. Various factors make the value fluctuations often cover the slight abnormalities in the values. Therefore, the current collected values ​​can be combined with past power supply values ​​to perform abnormal detection. According to the target power supply value, several clusters of past power supply reference values ​​with benchmarks are selected from the information table. It is involved that the load end is affected by the load end's own factors and external scene factors during operation. When the load end is powered on according to the same target power supply value, it is necessary to calibrate the errors of the actual parameters and the sent principle parameters caused by several types of factors, and often slight changes are made to the sent principle parameters.

[0006] The present invention uses the following technical solutions.

[0007] A method for evaluating the reliability of power supply in a source-grid-load-storage distribution network, comprising:

[0008] Collect the power usage data at the load end and construct a model to perform load power prediction Pload;

[0009] The power supply reliability evaluation method of the source-grid-load-storage distribution network also includes:

[0010] Step 1: Obtain the target power supply value, target operation parameter value and power supply scenario value 1 of the load end, and construct an import parameter vector 1 and a detection parameter vector 1 according to the operation parameter value;

[0011] Step 2: extracting a number of groups of past power supply reference values ​​from the power supply values ​​used to register the power supply of the distribution network to the load end according to the target power supply value and constructing a power supply reference value group, extracting the corresponding input parameter vector 2 and detection parameter vector 2 of each group of past power supply reference values;

[0012] Step 3: Perform a discriminative analysis 1 on the target operating parameter value and a plurality of clusters of past power supply reference values, extract a parameter discriminative vector 1 of the target operating parameter value, and a parameter discriminative vector 2 of the past power supply reference values ​​of each cluster;

[0013] Step 4: Performing a second distinguishing analysis on a plurality of clusters of past power supply reference values ​​according to the target operating parameter value, and extracting detection parameter distinguishing vectors corresponding to the target operating parameter value and the plurality of clusters of past power supply reference values;

[0014] Step 5: Determine the difference reference coefficient of each cluster's past power supply reference value based on parameter difference vector 1 and several clusters' parameter difference vector 2, form an abnormal detection value of the target operating parameter value based on several clusters' detection parameter difference vectors and the difference reference coefficient of each cluster's past power supply reference value, and configure the load-side control mode based on the abnormal detection value.

[0015] Further, in step 1, the target power supply value is used to represent the power supply specification value of the distribution network currently supplying power to the load end. The target power supply value includes the type of device at the load end), the power consumption capacity of the load end, the rated parameters of the load end, and the control specification value of the distribution network for the load end. The target operation parameter value includes an operation import parameter 1 and an operation detection parameter 1, which respectively represent the parameters sent to the load end by the distribution network and the working value of the load end obtained through detection during the actual distribution network supplying power to the load end. The power supply scenario value 1 includes the scenario parameters of the load end during the operation period. The operation import parameter 1 and the operation detection parameter 1 are respectively extracted to obtain an import parameter vector 1 and a detection parameter vector 1. Each value in the parameter vector represents a parameter.

[0016] Further, in step 2, the power supply values ​​used to register the power supply of the distribution network to the load end record the past power supply values ​​of the distribution network to the load end under different power supply modes of the distribution network to the load end, and the past power supply values ​​that are the same as the target power supply values ​​are selected from the power supply values ​​used to register the power supply of the distribution network to the load end, and they are several clusters of past power supply values ​​without abnormalities during the period of power supply of the distribution network to the load end, so as to obtain several clusters of past power supply benchmark values ​​and construct a power supply benchmark value group.

[0017] Further, in step 2, a random cluster of past power supply reference values ​​includes past target power supply values, past operation parameter values ​​and power supply scenario value 2, and the past operation parameter values ​​include operation import parameter 2 and operation detection parameter 2. Value extraction is performed on the past operation parameter values ​​to obtain import parameter vector 2 corresponding to operation import parameter 2 and detection parameter vector 2 corresponding to operation detection parameter 2.

[0018] Further, in step 3, the difference analysis 1 analyzes the difference between the imported parameter vector and the detected parameter vector, just as the difference between each parameter between the imported parameter vector 1 and the detected parameter vector is calculated, and the difference is combined to construct a parameter difference vector 1 for obtaining the target operating parameter value, and the difference between each parameter between the imported parameter vector 2 and the detected parameter vector 2 of the past power supply reference value of each cluster is calculated respectively, and the difference is combined to construct a parameter difference vector 2 for obtaining the past power supply reference value of each cluster.

[0019] Further, in step 4, the second distinguishing analysis includes analyzing the difference between a detection parameter vector 1 of the target operating parameter value and a detection parameter vector 2 respectively corresponding to a number of clusters of past power supply benchmark values, and using the difference between an imported parameter vector 1 of the target operating parameter value and an imported parameter vector 2 respectively corresponding to a number of clusters of past power supply benchmark values ​​as a calibration reference to obtain detection parameter difference vectors respectively corresponding to the target operating parameter value and a number of clusters of past power supply benchmark values.

[0020] Further, in step 5, it is determined whether the target difference values ​​corresponding to the previous power supply benchmark values ​​of each cluster in the abnormal detection value are all lower than the preset difference threshold. If so, the control mode of the load end is not configured;

[0021] On the contrary, the target power supply value and the power supply scenario value of the load end are sent into the control mode formation mode to form the corresponding operating parameter benchmark value of the load end, and the control mode of the load end is configured according to the operating parameter benchmark value.

[0022] Furthermore, in step 4, a method of performing a second distinguishing analysis on a plurality of clusters of past power supply reference values ​​according to the target operating parameter value to extract detection parameter distinguishing vectors corresponding to the target operating parameter value and the plurality of clusters of past power supply reference values ​​includes:

[0023] The differences of each parameter between a detection parameter vector one and several clusters of detection parameter vector twos are calculated respectively, and the differences are combined to obtain several detection difference vectors; the differences of each parameter between an input parameter vector one and several input parameter vector twos are calculated respectively, and the differences are combined to obtain multiple input difference vectors.

[0024] Furthermore, in step 5, the method of determining the difference reference coefficient of the previous power supply reference value of each cluster according to the parameter difference vector 1 and the plurality of cluster parameter difference vectors 2 includes:

[0025] Determine the scenario effect coefficient of the previous power supply benchmark value of each cluster according to the power supply scenario value 1 of the load end and the power supply scenario value 2 corresponding to the previous power supply benchmark value of each cluster;

[0026] Performing approximation analysis on a plurality of clusters of parameter distinguishing vectors 2 according to the parameter distinguishing vector 1, and determining the distinguishing correlation coefficient of the previous power supply benchmark values ​​of each cluster according to the approximation between the parameter distinguishing vector 1 and the plurality of clusters of parameter distinguishing vectors 2;

[0027] After the scene effect coefficient and the difference correlation coefficient of the previous power supply benchmark values ​​of each cluster are obtained by calculation, the difference benchmark coefficient of the previous power supply benchmark values ​​of each cluster is determined according to the scene effect coefficient and the difference correlation coefficient, as shown in the following equation:

[0028]

[0029] In the equation, Esf j is the difference benchmark coefficient of the previous power supply benchmark value of the jth cluster, Fp j is the scenario effect coefficient of the jth cluster’s previous power supply benchmark value, St j is the difference correlation coefficient of the previous power supply benchmark values ​​of the jth cluster, θ, They are respectively the pre-set importance parameters of the scene effect coefficient and the distinction correlation coefficient.

[0030] Furthermore, in step 5, the method for calculating the scenario effect coefficient includes: respectively calculating the KL divergence between the power supply scenario value 1 at the load end and the power supply scenario value 2 corresponding to the previous power supply benchmark value of each cluster, and performing standardization processing on several KL divergences using the Z-score method.

[0031] Furthermore, in step 5, the method for calculating the difference correlation coefficient includes respectively calculating the L2 norm between the parameter difference vector 1 and several clusters of parameter difference vector 2 to evaluate the correlation between the parameter difference vectors, and performing standardization processing on several L2 norms using the Z-score method to obtain the difference correlation coefficient of the past power supply benchmark values ​​of each cluster.

[0032] Furthermore, in step 5, the method of forming an abnormal detection value of a target operation parameter value according to a plurality of cluster detection parameter difference vectors and a difference reference coefficient of a previous power supply reference value of each cluster includes:

[0033] According to the detection parameter distinction vectors respectively corresponding to the target operation parameter value and a plurality of clusters of past power supply reference values, respectively, the distinction reference values ​​of the past power supply reference values ​​of each cluster corresponding to the target operation parameter value are determined;

[0034] Calibrate the difference reference values ​​of the previous power supply reference values ​​of each cluster according to the difference reference coefficients of the previous power supply reference values ​​of each cluster to the difference reference values ​​of the target operation parameter values, and obtain the target difference values ​​of the previous power supply reference values ​​of each cluster in the power supply reference value group corresponding to the target operation parameter values. Here, the calculation equation of the target difference value is:

[0035] H k =v k W k

[0036] In the equation, H k is the target difference value of the kth cluster's previous power supply benchmark value, W k is the difference reference value of the previous power supply reference value of the kth cluster, ν k It is the distinguishing benchmark coefficient of the kth cluster of previous power supply benchmark values ​​set in advance.

[0037] Furthermore, in step 5, the variance of the detection parameter difference vector of each cluster is calculated, and the variance of the detection parameter difference vector of each cluster is used as the difference reference value of the previous power supply reference value of each cluster corresponding to the target operation parameter value.

[0038] Furthermore, in the face of the control mode, the formation of the modality also includes:

[0039] The control mode formation mode is obtained by inputting sample learning. The sample contains several clusters of operation parameter values, and the target power supply values ​​and power supply scenario values ​​corresponding to each cluster of operation parameter values ​​are as follows:

[0040] The restricted Boltzmann machine is used to construct the control mode formation model to grasp the relationship between several clusters of operating parameter values ​​in the sample, the target power supply values ​​and the power supply scenario values ​​corresponding to each cluster of operating parameter values. During the model learning, the target power supply values ​​and the power supply scenario values ​​corresponding to each cluster of operating parameter values ​​are used as the input values ​​of the control mode formation model and fed into the restricted Boltzmann machine, and several clusters of operating parameter values ​​are used as the learning objectives of the control mode formation model to learn and obtain the control mode formation model.

[0041] A power supply reliability evaluation device for a source-grid-load-storage distribution network, comprising:

[0042] A construction module, which is used to obtain a target power supply value, a target operation parameter value and a power supply scenario value 1 of the load end, and to obtain an input parameter vector 1 and a detection parameter vector 1 according to the operation parameter value construction;

[0043] An extraction module is used to extract a number of clusters of past power supply reference values ​​from the power supply values ​​used to register the power supply of the distribution network to the load end according to the target power supply value and construct a power supply reference value group, and extract the corresponding input parameter vector 2 and detection parameter vector 2 of each cluster of past power supply reference values;

[0044] The first analysis module is used to perform a first distinguishing analysis on the target operation parameter value and a plurality of clusters of past power supply reference values, and extract a first parameter distinguishing vector of the target operation parameter value and a second parameter distinguishing vector of the past power supply reference values ​​of each cluster;

[0045] The second analysis module is used to perform a second distinguishing analysis on a plurality of clusters of past power supply reference values ​​according to the target operation parameter value, and extract detection parameter distinguishing vectors corresponding to the target operation parameter value and the plurality of clusters of past power supply reference values;

[0046] The control module is used to determine the difference reference coefficient of the previous power supply reference value of each cluster based on the parameter difference vector 1 and the parameter difference vector 2 of several clusters, form the abnormal detection value of the target operation parameter value based on the detection parameter difference vectors of several clusters and the difference reference coefficient of the previous power supply reference value of each cluster, and configure the control mode of the load end according to the abnormal detection value.

[0047] The beneficial effects of the present invention are that, compared with the prior art, the technical effects of the present invention include:

[0048] Through comprehensive analysis of real-time operating parameter values ​​and past power supply benchmark values, active improvement and precise control during load operation can be achieved, and the reliability and coordination of the distribution network's power supply performance to the load end can be improved. Through automated differential analysis and calculation, the effects of changes in power supply scenarios and the reliability of the load end itself on the distribution network's power supply process to the load end are effectively introduced, and small abnormalities in the values ​​are detected, so that the distribution network's power supply to the load end can provide feedback on small abnormal values ​​and configure control modes to ensure continuous improvement of power supply performance. Based on the response learning and control mode improvement of past power supply values, it can match power supply changes and continuously improve and improve the distribution network's power supply process to the load end, thereby improving the distribution network's power supply performance to the load end. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a partial flow chart of the power supply reliability evaluation method of the source-grid-load-storage distribution network described in the present invention;

[0050] Figure 2 It is a partial structural schematic diagram of the power supply reliability assessment device containing source, grid, load, storage and distribution network described in the present invention. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely expressed in conjunction with the drawings in the embodiments of the present invention. The embodiments expressed in this application are only some embodiments of the present invention, not all embodiments. According to the spirit of the present invention, other embodiments obtained by ordinary technicians in this field without creative work are all within the protection scope of the present invention.

[0052] like Figure 1 As shown, the power supply reliability evaluation method of a source-grid-load-storage distribution network described in the present invention comprises:

[0053] Collect the power usage data at the load end and construct a model to perform load power prediction Pload;

[0054] The power supply reliability evaluation method of the source-grid-load-storage distribution network also includes:

[0055] Step 1: Obtain the target power supply value, target operation parameter value and power supply scenario value 1 of the load end, and construct an import parameter vector 1 and a detection parameter vector 1 according to the operation parameter value;

[0056] In a preferred but non-limiting embodiment of the present invention, in step 1, the target power supply value is used to represent the power supply specification value of the distribution network currently supplying power to the load end. The target power supply value includes the type of device at the load end (such as the load end contains electrical devices such as refrigerators, washing machines, motors, etc.), the power consumption capacity of the load end, the rated parameters of the load end (such as rated power, rated voltage or rated current), the distribution network's load end control specification value (such as the output power, output voltage, and output current of the transformer or transformer station of the distribution network that supplies power to the load end) and other values. The target operation parameter value includes an operation input parameter 1 and an operation detection parameter 1, which respectively represent the load end being fed into the distribution network during the actual distribution network supplying power to the load end. Parameters (the parameters may be the power value, voltage value and current value sent by the distribution network to the load end) and the working values ​​of the load end obtained through specific detection by devices such as power transmitters, voltage transmitters, current transmitters, etc. (the working values ​​include the working power, working voltage and working current when the load end is working). The power supply scenario value 1 includes the scenario parameters of the load end during operation, such as the temperature, wind speed, humidity and other values ​​of the scene where the load end is located. The operation imported parameter 1 and the operation detection parameter 1 are respectively extracted to obtain the imported parameter vector 1 and the detection parameter vector 1. Each value in the parameter vector represents a parameter, such as the power value, voltage value, current value sent by the distribution network to the load end, the working power, working voltage and working current and other parameters when the load end is working.

[0057] Step 2: extracting a number of groups of past power supply reference values ​​from the power supply values ​​used to register the power supply of the distribution network to the load end according to the target power supply value and constructing a power supply reference value group, extracting the corresponding input parameter vector 2 and detection parameter vector 2 of each group of past power supply reference values;

[0058] In a preferred but non-restrictive embodiment of the present invention, in step 2, the power supply value used to register the power supply of the distribution network to the load end records the past power supply values ​​of the distribution network to the load end under different power supply modes of the distribution network to the load end (that is, different power supply values ​​of the distribution network to the load end), which is used to provide a lot of numerical evidence for the analysis of the power supply mode of the distribution network to the load end. In this application, the past power supply values ​​that are the same as the target power supply values ​​are selected from the power supply values ​​used to register the power supply of the distribution network to the load end, and they are several clusters of past power supply values ​​without abnormalities during the period of power supply of the distribution network to the load end, so as to obtain several clusters of past power supply benchmark values ​​and construct a power supply benchmark value group.

[0059] In a preferred but non-limiting embodiment of the present invention, in step 2, a random cluster of past power supply reference values ​​includes past target power supply values, past operation parameter values ​​and power supply scenario value 2, and past operation parameter values ​​include operation import parameter 2 and operation detection parameter 2. Value extraction is performed on the past operation parameter values ​​to obtain import parameter vector 2 corresponding to operation import parameter 2 and detection parameter vector 2 corresponding to operation detection parameter 2.

[0060] Step 3: Perform a discriminative analysis 1 on the target operating parameter value and a plurality of clusters of past power supply reference values, extract a parameter discriminative vector 1 of the target operating parameter value, and a parameter discriminative vector 2 of the past power supply reference values ​​of each cluster;

[0061] In a preferred but non-restrictive embodiment of the present invention, in step 3, the key to differential analysis is to analyze the parameter differences involved in the power supply of the primary distribution network to the load end, that is, to analyze the difference between the imported parameter vector and the detected parameter vector, just as the difference between each parameter between the imported parameter vector one and the detected parameter vector is calculated (the difference between the parameters is the absolute value of the quantity obtained by subtracting the parameters), and the differences are combined to construct a parameter difference vector one for obtaining the target operating parameter value, and the differences between each parameter between the imported parameter vector two and the detected parameter vector two of the past power supply benchmark values ​​of each cluster are calculated respectively (the difference between the parameters is the absolute value of the quantity obtained by subtracting the parameters), and the differences are combined to construct a parameter difference vector two for obtaining the past power supply benchmark values ​​of each cluster.

[0062] Step 4: Performing a second distinguishing analysis on a plurality of clusters of past power supply reference values ​​according to the target operating parameter value, and extracting detection parameter distinguishing vectors corresponding to the target operating parameter value and the plurality of clusters of past power supply reference values;

[0063] In a preferred but non-limiting embodiment of the present invention, in step 4, the key to the distinguishing analysis is to analyze the difference between the current power supply value and the previous power supply value, including analyzing the difference between the detection parameter vector 1 of the target operating parameter value and the detection parameter vector 2 corresponding to the respective clusters of the previous power supply reference values ​​(the difference is the value obtained by subtracting the detection parameter vector 2 corresponding to the respective clusters of the previous power supply reference values ​​from the parameter vector 1), and according to the imported parameter vector 1 of the target operating parameter value and the imported parameter vector 2 corresponding to the respective clusters of the previous power supply reference values The difference between the two (the difference is the value obtained by subtracting the imported parameter vector 1 of the target operating parameter value from the imported parameter vector 2 corresponding to several clusters of past power supply benchmark values) is used as the calibration reference to obtain the detection parameter difference vectors corresponding to the target operating parameter value and several clusters of past power supply benchmark values. The detection parameter difference vector is used to represent the difference between the detection parameter vector 1 of the target operating parameter value and the detection parameter vector 2 corresponding to an arbitrary cluster of past power supply benchmark values, and calibration is performed based on the parameter differences involved in the power supply operation of the primary distribution network to the load end.

[0064] Step 5: Determine the difference reference coefficient of each cluster's past power supply reference value based on parameter difference vector 1 and several clusters' parameter difference vector 2, form an abnormal detection value of the target operating parameter value based on several clusters' detection parameter difference vectors and the difference reference coefficient of each cluster's past power supply reference value, and configure the load-side control mode based on the abnormal detection value.

[0065] In a preferred but non-limiting embodiment of the present invention, in step 5, the reference coefficient is differentiated through the parameter differentiation involved in the period of power supply from the primary distribution network to the load end (the period of power supply from the primary distribution network to the load end is the period from the start time of power supply to the load end to the termination of power supply to the load end), and the correlation between the parameter differentiation of the primary distribution network power supply process from the primary distribution network to the load end between several clusters of past power supply reference values ​​is integrated. The detection parameter differentiation vector is the correlation between the parameter differentiation involved in the period of power supply from the primary distribution network to the load end and the specific detection of the operation detection parameters between the target operating parameter value and several clusters of past power supply reference values ​​after numerical calibration. The target differentiation values ​​corresponding to several clusters of past power supply reference values ​​in the power supply reference value group are obtained by calculation, and the abnormal detection values ​​including the target differentiation values ​​corresponding to several clusters of past power supply reference values ​​in the power supply reference value group are obtained. The control mode of the load end can be configured according to the abnormal detection value.

[0066] Determine whether the target difference values ​​corresponding to the previous power supply benchmark values ​​of each cluster within the abnormal detection value are lower than the preset difference threshold, if so, the control mode of the load end is not configured;

[0067] On the contrary, the target power supply value and the power supply scenario value of the load end are sent into the control mode formation mode to form the corresponding operating parameter benchmark value of the load end, and the control mode of the load end is configured according to the operating parameter benchmark value.

[0068] Several clusters of past power supply benchmark values ​​are all values ​​within the past power supply values ​​during which the distribution network has no abnormal power supply to the load end. The analytical calculation of the target difference value uses the past power supply values ​​as the benchmark, and performs multi-dimensional difference analysis on the past power supply values ​​and the current values ​​to obtain the result. The higher the target difference value, the higher the deviation of the target operating parameter value and the past power supply value. The deviation is numerically set through the pre-set difference critical value. The target difference values ​​corresponding to each cluster of past power supply benchmark values ​​are all lower than the pre-set difference critical value, which means that the target operating parameter value and the past power supply value have not deviated, confirming that the value is not abnormal. On the contrary, if the target area corresponding to any cluster of past power supply benchmark values ​​is If the difference value is not lower than the preset difference critical amount, it means that the target operating parameter value deviates from the previous power supply value, which can be regarded as the target operating parameter value often being abnormal. Under such conditions, in order to ensure the power supply performance of the distribution network at the load end to the load end, the control mode formation model obtained in advance is used to process the values ​​involved in the power supply task of the distribution network to the load end, including the target power supply value and the power supply scenario value of the load end, which are sent to the control mode formation model to form the corresponding operating parameter benchmark value of the load end, and then the control mode of the load end is configured according to the formed operating parameter benchmark value, so as to achieve active improvement and precise control of the power supply process of the distribution network to the load end, and improve the reliability of the power supply performance of the distribution network to the load end.

[0069] In a preferred but non-limiting embodiment of the present invention, in step 4, a method of performing a second distinguishing analysis on a plurality of clusters of past power supply reference values ​​according to a target operating parameter value to extract detection parameter distinguishing vectors corresponding to the target operating parameter value and the plurality of clusters of past power supply reference values ​​includes:

[0070] The differences of each parameter between a detection parameter vector one and several clusters of detection parameter vector twos are calculated separately (the difference of parameters is the absolute value of the quantity obtained by subtracting the parameters), and the differences are combined to obtain several detection difference vectors; the differences of each parameter between an input parameter vector one and several input parameter vector twos are calculated separately (the difference of parameters is the absolute value of the quantity obtained by subtracting the parameters), and the differences are combined to obtain multiple input difference vectors.

[0071] The detection difference vector represents the difference between the parameters specifically detected and obtained between the target operating parameter value and several clusters of past power supply benchmark values, and the input difference vector represents the difference between the parameters specifically inputted between the target operating parameter value and several clusters of past power supply benchmark values. The detection difference vector corresponding to each cluster of past power supply benchmark values ​​is aggregated with the input difference vector to calibrate the detection difference vector according to the difference between the input parameter vector 1 of the target operating parameter value and the input parameter vector 2 corresponding to the several clusters of past power supply benchmark values, so as to obtain the detection parameter difference vectors corresponding to the target operating parameter value and the several clusters of past power supply benchmark values.

[0072] In a preferred but non-limiting embodiment of the present invention, in step 5, the method of identifying the difference reference coefficient of the previous power supply reference value of each cluster according to the parameter difference vector 1 and the plurality of cluster parameter difference vectors 2 includes:

[0073] Determine the scenario effect coefficient of the previous power supply benchmark value of each cluster according to the power supply scenario value 1 of the load end and the power supply scenario value 2 corresponding to the previous power supply benchmark value of each cluster;

[0074] In a preferred but non-restrictive embodiment of the present invention, in step 5, the method for calculating the scene effect coefficient includes: respectively calculating the KL divergence between the power supply scene value 1 of the load end and the power supply scene value 2 corresponding to the previous power supply benchmark value of each cluster. The KL divergence is suitable for situations where the difference in scene detection values ​​is very important. Scene elements such as scene temperature, wind speed, and humidity will affect the function of the load end and the performance of the distribution network in supplying power to the load end. The difference in scene elements between different values ​​analyzed according to the KL divergence can well represent the scene difference characteristics. In the face of several KL divergences obtained by calculation, it is necessary to be suitable for subsequent numerical analysis, and the Z-score method is used to perform normalization processing on several KL divergences, which is suitable for mutual comparison between the values, so as to obtain the scene effect coefficient of the previous power supply benchmark value of each cluster.

[0075] Performing approximation analysis on a plurality of clusters of parameter distinguishing vectors 2 according to the parameter distinguishing vector 1, and determining the distinguishing correlation coefficient of the previous power supply benchmark values ​​of each cluster according to the approximation between the parameter distinguishing vector 1 and the plurality of clusters of parameter distinguishing vectors 2;

[0076] In a preferred but non-limiting embodiment of the present invention, in step 5, the method for calculating the difference correlation coefficient includes respectively calculating the L2 norm between parameter difference vector one and several clusters of parameter difference vector two to evaluate the correlation between the parameter difference vectors, and similarly performing standardization processing on several L2 norms using the Z-score method to obtain the difference correlation coefficient of the past power supply benchmark values ​​of each cluster.

[0077] After the scene effect coefficient and the difference correlation coefficient of the previous power supply benchmark values ​​of each cluster are obtained by calculation, the difference benchmark coefficient of the previous power supply benchmark values ​​of each cluster is determined according to the scene effect coefficient and the difference correlation coefficient, as shown in the following equation:

[0078]

[0079] In the equation, Esf j is the difference benchmark coefficient of the previous power supply benchmark value of the jth cluster, Fp j is the scenario effect coefficient of the jth cluster’s previous power supply benchmark value, St j is the difference correlation coefficient of the previous power supply benchmark values ​​of the jth cluster, θ, They are respectively the pre-set importance parameters of the scene effect coefficient and the distinction correlation coefficient.

[0080] The difference benchmark coefficient obtained by the above method is used to identify the scenario effect coefficient, which effectively locks the load end in the external effects of temperature, humidity and other external factors of the scenario and sends them to the configuration of the power supply mode of the load end through the distribution network, thereby improving the matching and reliability of the power supply process of the distribution network to the load end. The difference benchmark coefficient integrates the multi-dimensional differences between the target operating parameter values ​​and the previous power supply benchmark values. Through numerical processing and analysis, a multi-dimensional integrated evaluation system is formed, which can fully and comprehensively reflect the hidden dangers of the distribution network during the power supply to the load end.

[0081] In a preferred but non-limiting embodiment of the present invention, in step 5, the method of forming an abnormal detection value of a target operating parameter value according to a plurality of cluster detection parameter difference vectors and a difference reference coefficient of a previous power supply reference value of each cluster includes:

[0082] According to the detection parameter distinction vectors respectively corresponding to the target operation parameter value and a plurality of clusters of past power supply reference values, respectively, the distinction reference values ​​of the past power supply reference values ​​of each cluster corresponding to the target operation parameter value are determined;

[0083] In a preferred but non-limiting embodiment of the present invention, in step 5, the variance of the detection parameter difference vector of each cluster is calculated, and the variance of the detection parameter difference vector of each cluster is used as the difference reference value of each cluster's previous power supply reference value corresponding to the target operation parameter value.

[0084] Calibrate the difference reference values ​​of the previous power supply reference values ​​of each cluster according to the difference reference coefficients of the previous power supply reference values ​​of each cluster to the difference reference values ​​of the target operation parameter values, and obtain the target difference values ​​of the previous power supply reference values ​​of each cluster in the power supply reference value group corresponding to the target operation parameter values. Here, the calculation equation of the target difference value is:

[0085] H k =v k W k

[0086] In the equation, H k is the target difference value of the kth cluster's previous power supply benchmark value, W k is the difference reference value of the previous power supply reference value of the kth cluster, ν k It is the distinguishing benchmark coefficient of the kth cluster of previous power supply benchmark values ​​set in advance.

[0087] Through the above method, an abnormal detection value including several target difference values ​​can be obtained. Here, the target difference value is obtained by performing a calibration operation on the difference reference value of the previous power supply reference value of each cluster corresponding to the target operation parameter value, which can accurately reflect the error status of the distribution network power supply performance to the load end between the previous power supply value of each cluster and the current target operation parameter value. The target difference value represents the higher the error. The calculation of the target difference value comprehensively evaluates the effect of the scene effect coefficient and the difference correlation coefficient. According to a number of past power supply values ​​and real-time operation parameter values, analytical calculation is performed. This kind of comprehensive evaluation can comprehensively analyze the interaction of various factors during the power supply of the distribution network to the load end, and provides a better numerical basis for the improvement of power supply performance. The target difference value can instantly reflect the error status of the power supply performance of the distribution network to the load end, and is combined with the integrated evaluation of past benchmark data. It has a good cyclic improvement function in practical application. By continuously collecting and analyzing the new power supply values ​​of the distribution network to the load end, the difference reference coefficient and the pre-set difference critical value can be flexibly configured, which better improves the control of the power supply process of the distribution network to the load end.

[0088] In a preferred but non-limiting embodiment of the present invention, the control mode is formed into a modality, further comprising:

[0089] The control mode formation mode is obtained by inputting sample learning. The sample contains several clusters of operation parameter values, and the target power supply values ​​and power supply scenario values ​​corresponding to each cluster of operation parameter values ​​are as follows:

[0090] The restricted Boltzmann machine is used to construct the control mode formation model to grasp the relationship between several clusters of operating parameter values ​​in the sample, and the target power supply values ​​and power supply scenario values ​​corresponding to each cluster of operating parameter values. During the model learning, the target power supply values ​​and power supply scenario values ​​corresponding to each cluster of operating parameter values ​​are used as the input values ​​of the control mode formation model and fed into the restricted Boltzmann machine, and several clusters of operating parameter values ​​are used as the learning objectives of the control mode formation model. The control mode formation model obtained by learning can be used to evaluate the corresponding operating parameter values ​​based on the input target power supply values ​​and power supply scenario values, so as to obtain an accurate distribution network control mode for the load-end power supply machine.

[0091] The present application is suitable for performing abnormal detection on the target operating parameter values ​​during the operation of the load end, and can be used to detect hidden abnormalities in the values. The load end is affected by factors such as scene factors and its own consumption during operation. There are often slight differences between the parameters sent during operation and the parameters of the actual operation of the load end. Various factors cause the value fluctuations to often cover the slight abnormalities in the values. Therefore, the values ​​currently collected can be combined with the past power supply values ​​to perform abnormal detection. According to the target power supply values, several clusters of past power supply reference values ​​with benchmarks are selected from the information table. It is involved that the load end is affected by the load end's own factors and external scene factors during operation. When the load end is powered on according to the same target power supply value, it is necessary to calibrate the errors between the actual parameters and the theoretical parameters sent under several types of factors, and often slight changes are made to the theoretical parameters sent.

[0092] In order to find out the hidden abnormal values ​​that are often covered in the values ​​and prevent the power supply performance of the distribution network to the load end from being adversely affected, the present invention can achieve active improvement and precise control during the operation of the load end through multi-faceted analysis of the current operating parameter values ​​and the previous power supply benchmark values, and improve the reliability and coordination of the power supply performance of the distribution network to the load end. Through automated differential analysis and calculation, the effects of changes in the power supply scene and the reliability of the load end itself on the power supply process of the distribution network to the load end are efficiently introduced, and small abnormalities in the values ​​are detected, so that the power supply of the distribution network to the load end can perform feedback on small abnormal values ​​and configure control modes to ensure continuous improvement of power supply performance. According to the response learning and control mode improvement of the previous power supply values, the power supply changes can be matched and the power supply process of the distribution network to the load end can be continuously improved and improved, thereby improving the power supply performance of the distribution network to the load end.

[0093] like Figure 2 As shown, the power supply reliability evaluation device of the source-grid-load-storage distribution network described in the present invention comprises:

[0094] A construction module, which is used to obtain a target power supply value, a target operation parameter value and a power supply scenario value 1 of the load end, and to obtain an input parameter vector 1 and a detection parameter vector 1 according to the operation parameter value construction;

[0095] An extraction module is used to extract a number of clusters of past power supply reference values ​​from the power supply values ​​used to register the power supply of the distribution network to the load end according to the target power supply value and construct a power supply reference value group, and extract the corresponding input parameter vector 2 and detection parameter vector 2 of each cluster of past power supply reference values;

[0096] The first analysis module is used to perform a first distinguishing analysis on the target operation parameter value and a plurality of clusters of past power supply reference values, and extract a first parameter distinguishing vector of the target operation parameter value and a second parameter distinguishing vector of the past power supply reference values ​​of each cluster;

[0097] The second analysis module is used to perform a second distinguishing analysis on a plurality of clusters of past power supply reference values ​​according to the target operation parameter value, and extract detection parameter distinguishing vectors corresponding to the target operation parameter value and the plurality of clusters of past power supply reference values;

[0098] The control module is used to determine the difference reference coefficient of the previous power supply reference value of each cluster based on the parameter difference vector 1 and the parameter difference vector 2 of several clusters, form the abnormal detection value of the target operation parameter value based on the detection parameter difference vectors of several clusters and the difference reference coefficient of the previous power supply reference value of each cluster, and configure the control mode of the load end according to the abnormal detection value.

[0099] The beneficial effects of the present invention are that, compared with the prior art, the technical effects of the present invention include:

[0100] Through comprehensive analysis of real-time operating parameter values ​​and past power supply benchmark values, active improvement and precise control during load operation can be achieved, and the reliability and coordination of the distribution network's power supply performance to the load end can be improved. Through automated differential analysis and calculation, the effects of changes in power supply scenarios and the reliability of the load end itself on the distribution network's power supply process to the load end are effectively introduced, and small abnormalities in the values ​​are detected, so that the distribution network's power supply to the load end can provide feedback on small abnormal values ​​and configure control modes to ensure continuous improvement of power supply performance. Based on the response learning and control mode improvement of past power supply values, it can match power supply changes and continuously improve and improve the distribution network's power supply process to the load end, thereby improving the distribution network's power supply performance to the load end.

[0101] The present disclosure can be a system, method and / or computer program product. The computer program product can include a computer-readable backup medium carrying computer-readable program instructions for causing a processor to implement various aspects of the disclosure.

[0102] The computer readable backup medium can be a tangible network capable of retaining and backing up the instructions used by the instruction execution network. The computer readable backup medium can be, but is not limited to, an electrical backup network, a magnetic backup network, an optical backup network, an electromagnetic backup network, a semiconductor backup network, or any suitable combination of the above. Further examples of computer readable backup media (non-exhaustive list) include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disk read-only memory (HD-ROM), digital versatile disk (DXD), memory stick, floppy disk, mechanically encoded network, such as punch cards or protrusions in grooves with instructions backed up thereon, and any suitable combination of the above. The computer readable backup medium used herein is not to be construed as a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a power line cable), or an electrical signal transmitted through wires.

[0103] The computer readable program instructions expressed herein can be downloaded from the computer readable backup medium to each inference / processing power grid line, or downloaded to an external computer or external backup power grid line through a wireless network, such as the Internet, a local area network, a wide area network and / or a wireless network. The wireless network can include copper transmission cables, power transmission lines, wireless transmission, routers, firewalls, switches, gateway computers and / or edge service devices. The wireless network adapter card or wireless network interface in each inference / processing power grid line receives the computer readable program instructions from the wireless network and forwards the computer readable program instructions for storage in the computer readable backup medium in each inference / processing power grid line.

[0104] The computer program instructions for performing the operations of the present disclosure can be assembler instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, conditional setting values, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Sdaletal A, H++, etc., and conventional procedural programming languages ​​such as "H" language or similar programming languages. The computer-readable program instructions can be executed entirely on the client computer, partially on the client computer, as a separate software package, partially on the client computer and partially on the remote computer, or completely on the remote computer or server. In the case of a remote computer, the remote computer can be connected to the client computer through any other wireless network, including a local area network (LAb) or a wide area network (WAb), or can be connected to an external computer (just like using an Internet service provider to connect through the Internet). In some embodiments, the electronic circuit is customized by using the status values ​​of computer-readable program instructions, such as a programmable logic circuit, a field programmable gate array (FPGA) or a programmable logic array (PLA), and the electronic circuit can execute computer-readable program instructions to achieve various aspects of the overhead disclosure.

[0105] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not deviate from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for evaluating the reliability of power supply in a source-grid-load-storage distribution network, characterized in that: include: Collect the power usage data at the load end and construct a model to perform load power prediction Pload; The power supply reliability evaluation method of the source-grid-load-storage distribution network also includes: Step 1: Obtain the target power supply value, target operation parameter value and power supply scenario value 1 of the load end, and construct an import parameter vector 1 and a detection parameter vector 1 according to the operation parameter value; Step 2: extracting a number of groups of past power supply reference values ​​from the power supply values ​​used to register the power supply of the distribution network to the load end according to the target power supply value and constructing a power supply reference value group, extracting the corresponding input parameter vector 2 and detection parameter vector 2 of each group of past power supply reference values; Step 3: Perform a discriminative analysis 1 on the target operating parameter value and a plurality of clusters of past power supply reference values, extract a parameter discriminative vector 1 of the target operating parameter value, and a parameter discriminative vector 2 of the past power supply reference values ​​of each cluster; Step 4: Performing a second distinguishing analysis on a plurality of clusters of past power supply reference values ​​according to the target operating parameter value, and extracting detection parameter distinguishing vectors corresponding to the target operating parameter value and the plurality of clusters of past power supply reference values; Step 5: Determine the difference reference coefficient of each cluster's past power supply reference value based on parameter difference vector 1 and several clusters' parameter difference vector 2, form an abnormal detection value of the target operating parameter value based on several clusters' detection parameter difference vectors and the difference reference coefficient of each cluster's past power supply reference value, and configure the load-side control mode based on the abnormal detection value.

2. The power supply reliability evaluation method of the source-grid-load-storage distribution network according to claim 1 is characterized in that: In step 1, the target power supply value is used to represent the power supply specification value of the distribution network currently supplying power to the load end. The target power supply value includes the type of device at the load end), the power consumption capacity of the load end, the rated parameters of the load end, and the control specification value of the distribution network for the load end. The target operation parameter value includes an operation import parameter 1 and an operation detection parameter 1, which respectively represent the parameters sent to the load end by the distribution network and the working value of the load end obtained through detection during the actual distribution network supplying power to the load end. The power supply scenario value 1 includes the scenario parameters of the load end during the operation period. The operation import parameter 1 and the operation detection parameter 1 are respectively extracted to obtain an import parameter vector 1 and a detection parameter vector 1. Each value in the parameter vector represents a parameter.

3. The power supply reliability evaluation method of the source-grid-load-storage distribution network according to claim 2 is characterized in that: In step 2, the power supply values ​​for registering the power supply of the distribution network to the load end under different power supply modes of the distribution network to the load end are registered, and the power supply values ​​of the past that are the same as the target power supply value are selected from the power supply values ​​for registering the power supply of the distribution network to the load end, and they are several groups of past power supply values ​​without abnormalities during the power supply of the distribution network to the load end, so as to obtain several groups of past power supply reference values ​​and construct a power supply reference value group; In step 2, a random cluster of past power supply reference values ​​includes past target power supply values, past operating parameter values ​​and power supply scenario value 2, and past operating parameter values ​​include operation import parameter 2 and operation detection parameter 2. Value extraction is performed on the past operating parameter values ​​to obtain import parameter vector 2 corresponding to operation import parameter 2 and detection parameter vector 2 corresponding to operation detection parameter 2.

4. The power supply reliability evaluation method of the source-grid-load-storage distribution network according to claim 3 is characterized in that: In step 3, the difference analysis 1 analyzes the difference between the input parameter vector and the detection parameter vector, just as the difference between each parameter between the input parameter vector 1 and the detection parameter vector is calculated, and the difference is combined to construct a parameter difference vector 1 for obtaining the target operating parameter value, and the difference between each parameter between the input parameter vector 2 and the detection parameter vector 2 of the past power supply reference value of each cluster is calculated respectively, and the difference is combined to construct a parameter difference vector 2 for obtaining the past power supply reference value of each cluster.

5. The power supply reliability evaluation method of the source-grid-load-storage distribution network according to claim 4 is characterized in that: In step 4, the second distinguishing analysis includes analyzing the difference between a detection parameter vector 1 of the target operating parameter value and a detection parameter vector 2 respectively corresponding to a number of clusters of past power supply reference values, and using the difference between an imported parameter vector 1 of the target operating parameter value and an imported parameter vector 2 respectively corresponding to a number of clusters of past power supply reference values ​​as a calibration reference to obtain detection parameter difference vectors respectively corresponding to the target operating parameter value and a number of clusters of past power supply reference values.

6. The power supply reliability evaluation method of the source-grid-load-storage distribution network according to claim 5 is characterized in that: In step 5, it is determined whether the target difference values ​​corresponding to the previous power supply reference values ​​of each cluster in the abnormal detection value are all lower than the preset difference threshold. If so, the control mode of the load end is not configured; On the contrary, the target power supply value and the power supply scenario value of the load end are sent into the control mode formation mode to form the corresponding operating parameter benchmark value of the load end, and the control mode of the load end is configured according to the operating parameter benchmark value.

7. The power supply reliability evaluation method of the source-grid-load-storage distribution network according to claim 6 is characterized in that: In step 4, a method of performing a second distinguishing analysis on a plurality of clusters of past power supply reference values ​​according to the target operating parameter value to extract detection parameter distinguishing vectors corresponding to the target operating parameter value and the plurality of clusters of past power supply reference values ​​includes: The differences of each parameter between a detection parameter vector one and several clusters of detection parameter vector twos are calculated respectively, and the differences are combined to obtain several detection difference vectors; the differences of each parameter between an input parameter vector one and several input parameter vector twos are calculated respectively, and the differences are combined to obtain multiple input difference vectors.

8. The power supply reliability evaluation method of the source-grid-load-storage distribution network according to claim 7 is characterized in that: In step 5, the method of determining the difference reference coefficient of the previous power supply reference value of each cluster according to the parameter difference vector 1 and the plurality of cluster parameter difference vectors 2 includes: Determine the scenario effect coefficient of the previous power supply benchmark value of each cluster according to the power supply scenario value 1 of the load end and the power supply scenario value 2 corresponding to the previous power supply benchmark value of each cluster; Performing approximation analysis on a plurality of clusters of parameter distinguishing vectors 2 according to the parameter distinguishing vector 1, and determining the distinguishing correlation coefficient of the previous power supply benchmark values ​​of each cluster according to the approximation between the parameter distinguishing vector 1 and the plurality of clusters of parameter distinguishing vectors 2; After the scene effect coefficient and the difference correlation coefficient of the previous power supply benchmark values ​​of each cluster are obtained by calculation, the difference benchmark coefficient of the previous power supply benchmark values ​​of each cluster is determined according to the scene effect coefficient and the difference correlation coefficient, as shown in the following equation: In the equation, Esf j is the difference benchmark coefficient of the previous power supply benchmark value of the jth cluster, Fp j is the scenario effect coefficient of the jth cluster’s previous power supply benchmark value, St j is the difference correlation coefficient of the previous power supply benchmark values ​​of the jth cluster, θ, They are the pre-set importance parameters of the scene effect coefficient and the distinction correlation coefficient; In step 5, the calculation method of the scenario effect coefficient includes: respectively calculating the KL divergence between the power supply scenario value 1 of the load end and the power supply scenario value 2 corresponding to the previous power supply benchmark value of each cluster, and performing standardization processing on a plurality of KL divergences using the Z-score method; In step 5, the calculation method of the difference correlation coefficient includes respectively calculating the L2 norm between the parameter difference vector 1 and the parameter difference vectors 2 of the plurality of clusters to evaluate the correlation between the parameter difference vectors, performing standardization processing on the plurality of L2 norms using the Z-score method, and obtaining the difference correlation coefficient of the previous power supply benchmark values ​​of each cluster; In step 5, the method of forming an abnormal detection value of a target operation parameter value according to a plurality of cluster detection parameter difference vectors and a difference reference coefficient of a previous power supply reference value of each cluster includes: According to the detection parameter distinction vectors respectively corresponding to the target operation parameter value and a plurality of clusters of past power supply reference values, respectively, the distinction reference values ​​of the past power supply reference values ​​of each cluster corresponding to the target operation parameter value are determined; Calibrate the difference reference values ​​of the previous power supply reference values ​​of each cluster according to the difference reference coefficients of the previous power supply reference values ​​of each cluster to the difference reference values ​​of the target operation parameter values, and obtain the target difference values ​​of the previous power supply reference values ​​of each cluster in the power supply reference value group corresponding to the target operation parameter values. Here, the calculation equation of the target difference value is: H k =n k W k In the equation, H k is the target difference value of the kth cluster's previous power supply benchmark value, W k is the difference reference value of the previous power supply reference value of the kth cluster, ν k is the difference reference coefficient of the previous power supply reference value of the kth cluster set in advance; In step 5, the variance of the detection parameter difference vector of each cluster is calculated, and the variance of the detection parameter difference vector of each cluster is used as the difference reference value of the previous power supply reference value of each cluster corresponding to the target operation parameter value.

9. The power supply reliability evaluation method of the source-grid-load-storage distribution network according to claim 8 is characterized in that: Facing the control mode, the modality is formed, including: The control mode formation mode is obtained by inputting sample learning. The sample contains several clusters of operation parameter values, and the target power supply values ​​and power supply scenario values ​​corresponding to each cluster of operation parameter values ​​are as follows: The restricted Boltzmann machine is used to construct the control mode formation model to grasp the relationship between several clusters of operating parameter values ​​in the sample, the target power supply values ​​and the power supply scenario values ​​corresponding to each cluster of operating parameter values. During the model learning, the target power supply values ​​and the power supply scenario values ​​corresponding to each cluster of operating parameter values ​​are used as the input values ​​of the control mode formation model and fed into the restricted Boltzmann machine, and several clusters of operating parameter values ​​are used as the learning objectives of the control mode formation model to learn and obtain the control mode formation model.

10. A power supply reliability evaluation device for a source-grid-load-storage distribution network, characterized in that: include: A construction module, which is used to obtain a target power supply value, a target operation parameter value and a power supply scenario value 1 of the load end, and to obtain an input parameter vector 1 and a detection parameter vector 1 according to the operation parameter value construction; An extraction module is used to extract a number of clusters of past power supply reference values ​​from the power supply values ​​used to register the power supply of the distribution network to the load end according to the target power supply value and construct a power supply reference value group, and extract the corresponding input parameter vector 2 and detection parameter vector 2 of each cluster of past power supply reference values; The first analysis module is used to perform a first distinguishing analysis on the target operation parameter value and a plurality of clusters of past power supply reference values, and extract a first parameter distinguishing vector of the target operation parameter value and a second parameter distinguishing vector of the past power supply reference values ​​of each cluster; The second analysis module is used to perform a second distinguishing analysis on a plurality of clusters of past power supply reference values ​​according to the target operation parameter value, and extract detection parameter distinguishing vectors corresponding to the target operation parameter value and the plurality of clusters of past power supply reference values; The control module is used to determine the difference reference coefficient of the previous power supply reference value of each cluster based on the parameter difference vector 1 and the parameter difference vector 2 of several clusters, form the abnormal detection value of the target operation parameter value based on the detection parameter difference vectors of several clusters and the difference reference coefficient of the previous power supply reference value of each cluster, and configure the control mode of the load end according to the abnormal detection value.

Images