A virtual power plant equipment health state evaluation method and device

Abstract

The application provides a virtual power plant equipment health state evaluation method and device, and belongs to the technical field of data processing and prediction, and comprises the following steps: establishing an equipment normal operation state data Gaussian mixture model A; using an EM algorithm to iterate A until parameters converge; determining whether the to-be-tested data belongs to a working condition C j , and satisfying that the to-be-tested data and the normal data set mean in the mode are minimum; calculating the Kullback-Leibler divergence of the to-be-tested data and the working condition distribution function, and then calculating the Jensen-Shannon divergence; and verifying whether the state of the equipment corresponding to the data with a larger Jensen-Shannon divergence is abnormal. The application judges the equipment health degree by using the divergence of historical normal operation data and current equipment operation data, has strong practicability, and can greatly reduce operation and maintenance costs.

Description

Technical Field

[0001] This invention belongs to the field of data processing and prediction technology, and more specifically, relates to a method and apparatus for assessing the health status of virtual power plant equipment. Background Technology

[0002] With the development of distributed power sources, energy storage, and diversified loads, virtual power plants, as a special type of power plant, participate in the power market and grid operation through aggregation and coordination optimization, which greatly helps to alleviate power shortages.

[0003] Because virtual power plants integrate diverse source and load devices such as distributed power sources, energy storage systems, controllable loads, and electric vehicles, and have varied operating conditions, the operation and maintenance of virtual power plant equipment present significant challenges. Reducing operating and maintenance costs is a problem that needs to be solved. Summary of the Invention

[0004] The purpose of this invention is to provide a method and apparatus for assessing the health status of virtual power plant equipment, aiming to solve the problem of reducing operating and maintenance costs.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is: to provide a method for assessing the health status of virtual power plant equipment, comprising:

[0006] Step 1: Collect real-time and historical data of the equipment during normal operation;

[0007] Step 2: Establish a Gaussian mixture model A for the normal operating status data of the equipment:

[0008] A = [B m C n α n y i ]

[0009] Among them, B m For a device within the system, C n A certain normal operating state of the equipment, α n Let y be the probability of this state occurring. i Provides multi-dimensional working data for this type of equipment operation; α n satisfy:

[0010]

[0011] y i Satisfy: Equipment B m Normal working state C n Download device operation data y i It obeys the parameter θ n (μ n , σ n2 The normal distribution of μ, where μ n Let σ be the mean of this normal distribution. n 2 Let be the standard deviation of the normal distribution; and the probability density function of this data set is as follows:

[0012]

[0013] Step 3: Iterate the parameters of Gaussian mixture model A using the EM algorithm until the parameters converge;

[0014] Step 4: Determine the data to be tested Z = {z1, z2, ... z} i This belongs to operating condition C. j The condition is met by minimizing the mean of the test data and the normal dataset under this model, as shown in the following formula:

[0015]

[0016]

[0017] Step 5: Determine the number of data to be tested and the operating condition C j Then, calculate the test data Z = {z1, z2, ... z}. i The Kullback-Leibler divergence of the distribution function for this operating condition is calculated, and then the Jensen-Shannon divergence is calculated:

[0018] Step 6: Verify whether there are any abnormalities in the status of the devices corresponding to data with large Jensen-Shannon divergence.

[0019] Preferably, step 3 includes the following steps:

[0020] Step 3.1: Calculate the sub-model n for each pair of running data y. i response parameter δ i,n

[0021]

[0022] Step 3.2: Use the response parameter δ obtained in Step 3.1 i,n Iterative model parameters

[0023]

[0024]

[0025]

[0026] Step 3.3: Repeat steps 3.1 and 3.2 until the parameters converge.

[0027] Preferably, in step 5, the formula for calculating the Kullback-Leibler divergence between the measured data and the operating condition distribution function is:

[0028]

[0029] Preferably, in step 5, the formula for calculating the Jensen-Shannon divergence between the measured data and the operating condition distribution function is:

[0030]

[0031]

[0032] The Jensen-Shannon divergence value ranges from 0 to 1.

[0033] Preferably, step 6 further includes: perfecting the Gaussian mixture model A.

[0034] The present invention also provides a virtual power plant equipment health status assessment device, characterized in that,

[0035] The data acquisition component is used to collect real-time and historical data during normal equipment operation.

[0036] Gaussian mixture model algorithm component, used to build Gaussian mixture model A of equipment normal operation status data based on real-time and historical data during normal operation:

[0037] A = [B m C n α n y i ]

[0038] Among them, B m For a device within the system, C n A certain normal operating state of the equipment, α n Let y be the probability of this state occurring. i Provides multi-dimensional working data for this type of equipment operation; α n satisfy:

[0039]

[0040] The EM algorithm iteration component is used to iterate the parameters of Gaussian mixture model A using the EM algorithm until the parameters converge.

[0041] The test data determination component is used to determine the test data Z = {z1, z2, ... z}. i This belongs to operating condition C. j The condition is met by minimizing the mean of the test data and the normal dataset under this model, as shown in the following formula:

[0042]

[0043]

[0044] The divergence calculation component is used to determine the relationship between the number of data points to be measured and the operating condition C. j Then, calculate the test data Z = {z1, z2, ... z}. i The Kullback-Leibler divergence of the distribution function for this operating condition is calculated, and then the Jensen-Shannon divergence is calculated.

[0045] The verification component is used to verify whether there are any anomalies in the status of devices corresponding to data with large Jensen-Shannon divergences.

[0046] Preferably, the Gaussian mixture model algorithm component includes a probability density function module, which is used to perform the following calculations:

[0047]

[0048] Among them, equipment B m Normal working state C n Download device operation data y i It obeys the parameter θ n (μ n , σ n 2 The normal distribution of μ, where μ n Let σ be the mean of this normal distribution. n 2 denoted as the standard deviation of the normal distribution.

[0049] Preferably, the EM algorithm iteration component includes:

[0050] The E-step module is used to calculate the sub-model n pairs of running data y. i response parameter δ i,n ;

[0051] The M-step module is used to obtain the response parameter δ from the E-step module. i,n Iterative model parameters;

[0052] The parameter convergence module is used to repeat the tasks performed by the E-step module and the M-step module until the parameters converge.

[0053] Preferably, the divergence calculation component includes:

[0054] The KL divergence module is used to calculate the Kullback-Leibler divergence between the measured data and the load distribution function.

[0055] The JS divergence module is used to calculate the Jensen-Shannon divergence between the measured data and the load distribution function.

[0056] Preferably, the verification component includes: a model improvement module for improving the Gaussian mixture model.

[0057] The beneficial effects of the virtual power plant equipment health status assessment method and device provided by the present invention are as follows: Compared with the prior art, the virtual power plant equipment health status assessment method and device of the present invention uses the divergence of historical normal operation data and current equipment operation data to judge the equipment health status, and continuously improves the normal operation data model through manual verification. The reliability of the Gaussian mixture model of normal operation data gradually increases, which has strong practicality and can greatly reduce operation and maintenance costs. Attached Figure Description

[0058] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0059] Figure 1 This is a flowchart of an anomaly determination process in a virtual power plant equipment health status assessment method provided in an embodiment of the present invention;

[0060] Figure 2 This is a flowchart illustrating a method for assessing the health status of virtual power plant equipment, as provided in an embodiment of the present invention. Detailed Implementation

[0061] To make the technical problems, solutions, and beneficial effects of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0062] Please refer to the following: Figure 1 and Figure 2 The present invention will now describe a method for assessing the health status of virtual power plant equipment. The method includes the following steps:

[0063] Step S1: Collect real-time and historical data of the equipment during normal operation;

[0064] In this embodiment of the invention, the real-time and historical data of normal equipment operation include: distributed photovoltaic: photovoltaic bracket tilt angle and azimuth angle; tracking system elevation angle and azimuth angle; operating status; automatic / manual status; DC cabinet input and output current and voltage; inverter system DC input current and voltage; AC output current, voltage, power factor, frequency, active power, reactive power; power consumption; inverter efficiency, etc.

[0065] Step-up transformer: transformer oil temperature; transformer oil level; transformer oil pressure; high and low voltage circuit breaker position; high and low voltage voltage, current, active power, reactive power, power factor, frequency; transformer bushing temperature, etc.

[0066] Distributed energy storage system: system status (charging / discharging / standby / disabled); individual battery voltage, current, temperature, and maximum voltage; battery cluster voltage, current, temperature, average voltage, minimum voltage of individual cells, maximum temperature of individual cells, minimum temperature of individual cells, and average temperature of individual cells; battery array individual maximum voltage, minimum voltage, average voltage, maximum temperature, minimum temperature, average temperature, charge / discharge voltage, and current; cycle count; depth of discharge, etc.

[0067] Step S2: Establish Gaussian mixture model A for normal equipment operation status data:

[0068] A = [B m C n α n y i ]

[0069] Among them, B m For a device within the system, C n A certain normal operating state of the equipment, α n Let y be the probability of this state occurring. i Provide multi-dimensional working data for this type of equipment operation;

[0070] For example, a distributed energy storage system within the system may have four operating states: C1 is charging, C2 is discharging, C3 is standby, and C4 is disabled. The probabilities of these states, α1, α2, α3, and α4, are related to the actual operating conditions of the system and are obtained from existing and historical data. This includes system state (charging / discharging / standby / disabled); individual battery cell voltage, current, temperature, and maximum voltage; battery cluster voltage, current, temperature, average voltage, individual cell minimum voltage, individual cell maximum temperature, individual cell minimum temperature, and individual cell average temperature; battery array individual cell maximum voltage, minimum voltage, average voltage, maximum temperature, minimum temperature, average temperature, charge / discharge voltage, and current; cycle count; and depth of discharge.

[0071] α n satisfy:

[0072]

[0073] y i Satisfy: Equipment B m Normal working state C n Download device operation data y i It obeys the parameter θ n (μ n , σ n 2 The normal distribution of μ, where μ n Let σ be the mean of this normal distribution. n 2 Let be the standard deviation of the normal distribution; and the probability density function of this data set is as follows:

[0074]

[0075] Step S3: Iterate the parameters of Gaussian mixture model A using the EM algorithm until the parameters converge;

[0076] Step S3 can be specifically described as follows:

[0077] Step S3.1: Calculate the nth distribution pair running data y i response parameter δ i,n Complete the parameter α n μ n and σ n 2 The information is used for subsequent iteration parameters, where φ represents the preset completion parameter.

[0078]

[0079] Step S3.2: Use the response parameter δ obtained in step S3.1. i,n Iterative model parameters

[0080]

[0081]

[0082]

[0083] Step 3.3: Repeat steps S3.1 and S3.2 until parameter α is reached. n μ n , Convergence (maximum) yields the required model parameters.

[0084] Step S4: Determine the data to be tested Z = {z1, z2, ... z} i This belongs to operating condition C. jThe condition is met by minimizing the mean of the test data and the normal dataset under this model, as shown in the following formula:

[0085]

[0086]

[0087] In the formula, z i It is the i-th data point in the distribution function Z(x) of the data to be tested, μ i It is the mean of the normal data distribution obtained from step 3.3 iteration.

[0088] Step S5: Determine the number of data to be tested and the operating condition C j Then, calculate the test data Z = {z1, z2, ... z}. i The distribution function Z(x) of the normal data distribution function P(x) under this working condition is calculated using the Kullback-Leibler divergence, and then the Jensen-Shannon divergence is calculated:

[0089] Step S5 can be specifically described as follows:

[0090] Step 5.1: Determine the number of data to be tested and the operating condition C. j Then, the Kullback-Leibler divergence D between the measured data Z(x) and the normal data distribution function P(x) under this working condition is calculated. KL (P||Z):

[0091]

[0092] Step 5.2: Calculate the Jensen-Shannon divergence between the measured data and the distribution function of the operating condition:

[0093] Since the KL divergence does not satisfy symmetry and the triangle inequality, the JS divergence D is calculated based on the KL divergence. JS (P||Z):

[0094]

[0095]

[0096] Where R(x) is the average distribution of the two distribution functions P(x) and Z(x), and D KL (PR) is the KL divergence measure of the distribution functions P(x) and R(x), D KL(Z||R) are the KL divergence measures of the two distribution functions Z(x) and R(x), respectively. The range of JS divergence is between 0 and 1. JS divergence of 0 indicates that the probability distribution of the data to be tested is completely consistent with that of the data under normal operating conditions, and the equipment is in a healthy state. JS divergence of 1 indicates that the two probability distributions are very different, and the equipment is in an unhealthy state.

[0097] Step S6: Verify whether the status of the devices corresponding to data with large Jensen-Shannon divergence is abnormal. Verification can be performed manually.

[0098] Step S6 also includes perfecting the normal data model A.

[0099] The present invention provides a virtual power plant equipment health status assessment method. Compared with the prior art, it uses the divergence between historical normal operation data and current equipment operation data to judge the equipment health status. The normal operation data model is continuously improved through manual verification, and the reliability of the Gaussian mixture model of normal operation data gradually increases. It has strong practicality and can greatly reduce operation and maintenance costs.

[0100] This invention also provides a virtual power plant equipment health status assessment device; please refer to [link / reference]. Figures 1 to 2 It includes: a data acquisition component, a Gaussian mixture model algorithm component, an EM algorithm iteration component, a test data determination component, a divergence calculation component, and a verification component. The data acquisition component is used to collect real-time and historical data of the equipment's normal operation; the Gaussian mixture model algorithm component is used to establish a Gaussian mixture model A of the equipment's normal operation status data based on the real-time and historical data of normal operation.

[0101] A = [B m C n α n y i ]

[0102] Among them, B m For a device within the system, C n A certain normal operating state of the equipment, α n Let y be the probability of this state occurring. i Provides multi-dimensional working data for this type of equipment operation; α n satisfy:

[0103]

[0104] The EM algorithm iteration component is used to iterate the parameters of Gaussian mixture model A using the EM algorithm until the parameters converge.

[0105] The test data determination component is used to determine the test data Z = {z1, z2, ... z}. i This belongs to operating condition C.j The condition is met by minimizing the mean of the test data and the normal dataset under this model, as shown in the following formula:

[0106]

[0107]

[0108] The divergence calculation component is used to determine the relationship between the number of data to be measured and the operating condition C. j Then, calculate the test data Z = {z1, z2, ... z}. i The Kullback-Leibler divergence of the distribution function for this operating condition is calculated, and then the Jensen-Shannon divergence is calculated.

[0109] The verification component is used to verify whether there are any anomalies in the status of devices corresponding to data with large Jensen-Shannon divergence.

[0110] As some embodiments provided in this invention, please refer to Figures 1 to 2 The Gaussian mixture model algorithm component includes a probability density function module, which is used to perform the following calculations:

[0111]

[0112] Among them, equipment B m Normal working state C n Download device operation data y i It obeys the parameter θ n (μ n , σ n 2 The normal distribution of μ, where μ n Let σ be the mean of this normal distribution. n 2 denoted as the standard deviation of the normal distribution.

[0113] As some implementation methods provided in this invention, the EM algorithm iteration component includes: an E-step module, an M-step module, and a parameter convergence module. The E-step module is used to calculate the sub-model n pairs of running data y. i response parameter δ i,n The M-step module is used to obtain the response parameter δ from the E-step module. i,n Iterative model parameters; the parameter convergence module is used to repeat the tasks performed by the E-step module and the M-step module until the parameters converge.

[0114] As one of the embodiments provided in this invention, the verification component includes a model improvement module for improving the Gaussian mixture model.

[0115] The present invention provides a virtual power plant equipment health status assessment device, which can be used to implement a virtual power plant equipment health status assessment method.

[0116] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for assessing the health status of virtual power plant equipment, characterized in that, include: Step 1: Collect real-time and historical data of the equipment during normal operation; Step 2: Establish a Gaussian mixture model A for the normal operating status data of the equipment: ； in, B m For a device within the system, C n A certain normal operating state of the equipment, α n This represents the probability of this state occurring. y i Provide multi-dimensional working data for this type of equipment operation; α n satisfy: ； y i Satisfy: Equipment B m Normal working status C n Download device operation data y i It obeys parameters θ n ( μ n , σ n 2 The normal distribution of ), where, μ n Let be the mean of this normal distribution. σ n 2 Let be the standard deviation of the normal distribution; and the probability density function of this data set is as follows: ； Step 3: Iterate the parameters of Gaussian mixture model A using the EM algorithm until the parameters converge; Step 4: Determine the data to be tested Does it belong to operating condition C? j The condition is met by minimizing the mean of the test data and the normal dataset under this model, as shown in the following formula: ； ； in, It is the mean of the normal data distribution obtained from step 3 iteration; Step 5: Calculate the test data The Kullback-Leibler divergence of the distribution function for this operating condition is calculated, and then the Jensen-Shannon divergence is calculated. In step 5, the formula for calculating the Kullback-Leibler divergence between the measured data and the load distribution function is: ； in, P ( x () is the operating condition distribution function. Z ( x () represents the distribution function of the data to be tested; Step 6: Verify whether there are any abnormalities in the status of the devices corresponding to data with large Jensen-Shannon divergence.

2. The method for assessing the health status of virtual power plant equipment as described in claim 1, characterized in that, Step 3 includes the following steps: Step 3.1: Calculate the nth distribution pair running data. y i Response parameters ； Step 3.2: Use the response parameters obtained in Step 3.1 Iterative model parameters ； ； ； Step 3.3: Repeat steps 3.1 and 3.2 until the parameters converge.

3. The method for assessing the health status of virtual power plant equipment as described in claim 1, characterized in that, In step 5, the formula for calculating the Jensen-Shannon divergence between the measured data and the load distribution function is as follows: ； ； in, R ( x )yes P ( x )and Z ( x The average distribution of the two distribution functions, yes P ( x )and R ( x The KL divergence measure distribution of two distribution functions. They are Z ( x )and R ( x The KL divergence measure distribution of the two distribution functions, with the Jensen-Shannon divergence value range between 0 and 1.

4. The method for assessing the health status of virtual power plant equipment as described in claim 1, characterized in that, Step 6 also includes: refining Gaussian mixture model A.

5. A virtual power plant equipment health status assessment device, characterized in that, The data acquisition component is used to collect real-time and historical data during normal equipment operation. Gaussian mixture model algorithm component, used to build Gaussian mixture model A of equipment normal operation status data based on real-time and historical data during normal operation: ； in, B m For a device within the system, C n A certain normal operating state of the equipment, α n This represents the probability of this state occurring. y i Provide multi-dimensional working data for this type of equipment operation; α n satisfy: ； y i Satisfy: Equipment B m Normal working status C n Download device operation data y i It obeys parameters θ n ( μ n , σ n 2 The normal distribution of ), where, μ n Let be the mean of this normal distribution. σ n 2 Let be the standard deviation of the normal distribution; and the probability density function of this data set is as follows: ； The EM algorithm iteration component is used to iterate the parameters of Gaussian mixture model A using the EM algorithm until the parameters converge. The test data determination component is used to determine the test data. Does it belong to operating condition C? j The condition is met by minimizing the mean of the test data and the normal dataset under this model, as shown in the following formula: ； ； in, It is the mean of the normal data distribution obtained by the above iterative steps; The divergence calculation component calculates the divergence of the data to be tested. The Kullback-Leibler divergence of the distribution function for this operating condition is calculated, and then the Jensen-Shannon divergence is calculated. The formula for calculating the Kullback-Leibler divergence between the measured data and the operating condition distribution function is as follows: ； in, P ( x () is the operating condition distribution function. Z ( x () represents the distribution function of the data to be tested; The verification component is used to verify whether there are any anomalies in the status of devices corresponding to data with large Jensen-Shannon divergences.

6. The apparatus for assessing the health status of virtual power plant equipment as described in claim 5, characterized in that, The EM algorithm iteration components include: E-step module, used for calculating sub-models n For running data y i Response parameters ; The M-step module is used to obtain the response parameters from the E-step module. Iterative model parameters; The parameter convergence module is used to repeat the tasks performed by the E-step module and the M-step module until the parameters converge.

7. The virtual power plant equipment health status assessment device as described in claim 6, characterized in that, The divergence calculation component includes: The KL divergence module is used to calculate the Kullback-Leibler divergence between the measured data and the load distribution function. The JS divergence module is used to calculate the Jensen-Shannon divergence between the measured data and the load distribution function.

8. The virtual power plant equipment health status assessment device as described in claim 7, characterized in that, The verification component includes a model improvement module for improving the Gaussian mixture model.

Images