CN117709508A - New energy grid-connected fault prediction method based on nonlinear data model - Google Patents
New energy grid-connected fault prediction method based on nonlinear data model Download PDFInfo
- Publication number
- CN117709508A CN117709508A CN202311488370.XA CN202311488370A CN117709508A CN 117709508 A CN117709508 A CN 117709508A CN 202311488370 A CN202311488370 A CN 202311488370A CN 117709508 A CN117709508 A CN 117709508A
- Authority
- CN
- China
- Prior art keywords
- new energy
- energy grid
- parameter set
- fault
- time
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000013499 data model Methods 0.000 title claims abstract description 33
- 238000012706 support-vector machine Methods 0.000 claims abstract description 19
- 238000012216 screening Methods 0.000 claims abstract description 10
- 238000012544 monitoring process Methods 0.000 claims abstract description 9
- 238000010219 correlation analysis Methods 0.000 claims abstract description 7
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 17
- 238000003066 decision tree Methods 0.000 claims description 10
- 238000004140 cleaning Methods 0.000 claims description 8
- 238000003064 k means clustering Methods 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 7
- 238000012549 training Methods 0.000 claims description 7
- 238000007781 pre-processing Methods 0.000 claims description 6
- 238000002790 cross-validation Methods 0.000 claims description 5
- 238000001514 detection method Methods 0.000 claims description 5
- 238000010248 power generation Methods 0.000 description 10
- 230000005855 radiation Effects 0.000 description 5
- 230000002159 abnormal effect Effects 0.000 description 4
- 238000000540 analysis of variance Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
Landscapes
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The application provides a new energy grid-connected fault prediction method based on a nonlinear data model, which comprises the following steps: based on the space-time information of new energy grid connection, extracting and classifying a corresponding new energy grid connection space-time change parameter set, wherein the new energy grid connection space-time change parameter set comprises geographic position, time, power and frequency; integrating the new energy grid-connected system change parameter set and the new energy grid-connected space-time change parameter set into a new energy grid-connected change parameter set; performing correlation analysis on the new energy grid-connected variation parameter set, and screening parameters which possibly cause faults in new energy grid connection; in the framework of the nonlinear data model, analyzing the new energy grid-connected variation parameter set by using a support vector machine algorithm to obtain a mode possibly causing new energy grid-connected faults; and monitoring the real-time data stream of the new energy grid connection, and searching whether a real-time mode matched with the mode possibly causing the new energy grid connection fault exists or not.
Description
Technical Field
The invention relates to the technical field of information, in particular to a new energy grid-connected fault prediction method based on a nonlinear data model.
Background
With the rapid development and wide application of new energy grid-connected systems, system faults become a serious problem. The new energy grid connection refers to a process of connecting renewable energy sources such as wind energy, solar energy, water energy and the like, or other clean energy sources with a traditional power system to supply power or supply power to a power grid. At present, the fault prediction method for the new energy grid-connected system has some problems. How to analyze the continuous and changing new energy grid-connected space-time information and correlation in the system, and identify the mode which may cause the fault or the problem to be spread. Firstly, because the new energy grid-connected system has changeable characteristics and a complex structure, the traditional fault prediction method cannot accurately predict the faults of the system. There is noise and uncertainty in the data. Whereas conventional fault prediction methods generally assume that the data is accurate and complete, there is no effective challenge to the uncertainty of the data. Secondly, the existing fault prediction method only focuses on a certain parameter or mode, and cannot comprehensively consider the interrelationship among a plurality of parameters of the system. If only certain parameters or modes are considered, the state of the system cannot be fully described, and important information may be ignored. In addition, the existing fault prediction method can only realize offline analysis, and cannot monitor and rapidly respond to faults of the system in real time.
Disclosure of Invention
The invention provides a new energy grid-connected fault prediction method based on a nonlinear data model, which mainly comprises the following steps:
based on wind energy, solar energy and water energy data of new energy grid connection, extracting a corresponding new energy grid connection system change parameter set, and constructing a nonlinear data model, wherein the new energy grid connection change parameter set comprises power, frequency, current and voltage parameters; based on the space-time information of new energy grid connection, extracting and classifying a corresponding new energy grid connection space-time change parameter set, wherein the new energy grid connection space-time change parameter set comprises geographic position, time, power and frequency; integrating the new energy grid-connected system change parameter set and the new energy grid-connected space-time change parameter set into a new energy grid-connected change parameter set; performing correlation analysis on the new energy grid-connected variation parameter set, and screening parameters which possibly cause faults in new energy grid connection; in the framework of the nonlinear data model, analyzing the new energy grid-connected variation parameter set by using a support vector machine algorithm to obtain a mode possibly causing new energy grid-connected faults; monitoring a real-time data stream of new energy grid connection, and searching whether a real-time mode matched with the mode possibly causing new energy grid connection fault exists or not; and if a real-time mode matched with the mode which possibly causes the new energy grid-connected fault is obtained in the real-time data stream, carrying out new energy grid-connected fault prediction through a decision tree algorithm, and obtaining a final fault prediction result.
Further optionally, the wind energy, solar energy and water energy data based on new energy grid connection are extracted to obtain corresponding new energy grid connection system change parameter sets, and a nonlinear data model is constructed, wherein the new energy grid connection change parameter sets comprise power, frequency, current and voltage parameters, and the method comprises the following steps:
acquiring wind energy, solar energy and water energy data of new energy grid connection from a new energy grid connection system; preprocessing wind energy, solar energy and water energy data of the new energy grid connection, including data cleaning, denoising and outlier processing, and extracting a corresponding new energy grid connection system change parameter set, wherein the new energy grid connection system change parameter set comprises power, frequency, current and voltage parameters; establishing a nonlinear data model according to the change parameter set of the new energy grid-connected system by utilizing polynomial regression to obtain the relationship among power, frequency, current and voltage in the new energy grid-connected system; and (5) fitting the nonlinear model through a least square method, and determining coefficients in the nonlinear data model.
Further optionally, the extracting and classifying the corresponding new energy grid-connected space-time variation parameter set based on the space-time information of the new energy grid-connected, where the new energy grid-connected space-time variation parameter set includes geographic position, time, power and frequency, and the method includes:
acquiring space-time information of new energy grid connection from a new energy grid connection system; preprocessing the space-time information of the new energy grid connection, including data cleaning, denoising and outlier processing, and extracting a corresponding new energy grid connection space-time change parameter set, wherein the new energy grid connection space-time change parameter set comprises geographic position, time, power and frequency; determining a corresponding cluster number K by using a K-means clustering algorithm, and classifying the geographic position; if the classification is successful, the Murmur Hash3 algorithm is used for carrying out unique identification coding on the time to obtain a geographic position and a time identification set; classifying power and frequency by adopting a K-means clustering algorithm to obtain a power identification set and a frequency identification set, and correlating the power identification set and the frequency identification set with the geographic position and the time identification set; and if the association is successful, combining the power identification set, the frequency identification set, the geographic position and the time identification set.
Further optionally, the integrating the new energy grid-connected system change parameter set and the new energy grid-connected space-time change parameter set into a new energy grid-connected change parameter set includes:
acquiring a change parameter set of the new energy grid-connected system and a new energy grid-connected space-time change parameter set; matching and aligning the data, and corresponding the new energy grid-connected system change parameter set and the new energy grid-connected space-time change parameter set; integrating the matched and aligned new energy grid-connected system change parameter set and the new energy grid-connected space-time change parameter set into a new energy grid-connected change parameter set according to a required format; comparing and checking consistency of the power and frequency data in the new energy grid-connected system change parameter set and the new energy grid-connected space-time change parameter set; how the differences exist, performing data adjustment and calibration through linear fitting; the data is updated periodically to reflect the latest state of system operation.
Further optionally, the performing correlation analysis on the new energy grid-connected variation parameter set, and screening parameters possibly causing faults in new energy grid connection includes:
acquiring the new energy grid-connected variation parameter set from a new energy grid-connected system, wherein the new energy grid-connected variation parameter set comprises power, frequency, current and voltage parameters; calculating pearson correlation coefficients between the parameters; judging the value range of the correlation coefficient, and determining the correlation degree between parameters, wherein the higher the correlation degree is, the larger the risk of the multiple collinearity problem is; screening out parameters with absolute values of correlation higher than a preset correlation threshold value as parameters possibly causing faults; and further verifying the screened parameters which possibly cause faults through binary variance analysis, and judging whether obvious differences exist between the parameters possibly causing the faults.
Further optionally, in the framework of the nonlinear data model, the analyzing the new energy grid-connected variation parameter set by using a support vector machine algorithm to obtain a mode possibly causing new energy grid-connected faults includes:
acquiring the power, frequency, current and voltage parameters with normal histories and a known fault parameter set from a new energy grid-connected system; marking the parameter set as six categories of voltage anomalies, frequency offsets, current overloads, unbalanced voltages or currents, overvoltage or overcurrent, and power quality problems, based on known fault samples; training a support vector machine by using the marked parameter set, and selecting to use a polynomial kernel function to process a nonlinear relation under the framework of a nonlinear data model; classifying and predicting the unlabeled parameter set by using a trained support vector machine model; using cross-validation, evaluating the performance and accuracy of the model; and identifying a mode which possibly causes grid-connected faults of the new energy according to a prediction result of the support vector machine.
Further optionally, the monitoring the real-time data stream of the new energy grid connection, searching whether there is a real-time mode matching the mode that may cause the new energy grid connection fault, includes:
acquiring real-time data flow including power, frequency, current and voltage parameters from a new energy grid-connected system; matching the acquired real-time data stream with the mode which possibly causes the grid-connected fault of the new energy; and triggering a corresponding fault alarm according to the pattern matching result, and performing fault detection.
Further optionally, if the real-time mode matching the mode that may cause the new energy grid-connected fault is obtained in the real-time data stream, the new energy grid-connected fault prediction is performed through a decision tree algorithm, so as to obtain a final fault prediction result, including:
if the pattern matching is successful, acquiring historical power, frequency, current and voltage parameters from a new energy grid-connected system; using historical fault data as a training set, and modeling the fault type and the fault probability by utilizing a decision tree algorithm to generate a preliminary fault prediction result; acquiring power grid environment parameters from a new energy grid-connected system, and integrating the power grid environment parameters into a prediction result to adjust, wherein the power grid environment parameters comprise air temperature and humidity; judging whether the fault probabilities corresponding to different fault types in the adjusted fault prediction result exceed a preset fault probability threshold value or not; and if the fault probability threshold value exceeds the preset fault probability threshold value, executing automatic power grid dispatching and optimizing.
The technical scheme provided by the embodiment of the invention can have the following beneficial effects:
the invention discloses a fault prediction method based on new energy grid connection. According to the method, a new energy grid-connected system change parameter set of wind energy, solar energy and water energy is extracted, wherein the new energy grid-connected system change parameter set comprises power, frequency, current and voltage parameters. And simultaneously, extracting a corresponding new energy grid-connected space-time variation parameter set comprising geographic position, time, power and frequency according to the space-time information of the new energy grid connection. And then integrating the new energy grid-connected system change parameter set and the new energy grid-connected space-time change parameter set into a new energy grid-connected change parameter set. And then, carrying out correlation analysis on the new energy grid-connected variable parameter set, and screening parameters possibly causing faults. And in the framework of the nonlinear data model, analyzing the new energy grid-connected change parameter set by using a support vector machine algorithm to obtain a mode which possibly causes new energy grid-connected faults. And then, monitoring the real-time data stream of the new energy grid connection, and searching whether a real-time mode matched with a mode possibly causing the new energy grid connection fault exists or not. If a real-time mode matched with a mode possibly causing new energy grid-connected faults is found in the real-time data stream, the new energy grid-connected faults are predicted through a decision tree algorithm, and a final fault prediction result is obtained. The method can effectively predict the faults of the new energy grid-connected system and improve the safety and reliability of the new energy grid-connected system.
Drawings
FIG. 1 is a flow chart of a new energy grid-connected fault prediction method based on a nonlinear data model.
Fig. 2 is a schematic diagram of a new energy grid-connected fault prediction method based on a nonlinear data model.
Fig. 3 is a schematic diagram of a new energy grid-connected fault prediction method based on a nonlinear data model according to the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The new energy grid-connected fault prediction method based on the nonlinear data model in the embodiment specifically comprises the following steps:
step S101, based on wind energy, solar energy and water energy data of new energy grid connection, a corresponding new energy grid connection system change parameter set is extracted, a nonlinear data model is constructed, and the new energy grid connection change parameter set comprises power, frequency, current and voltage parameters.
And acquiring wind energy, solar energy and water energy data of new energy grid connection from the new energy grid connection system. Preprocessing wind energy, solar energy and water energy data of the new energy grid connection, including data cleaning, denoising and outlier processing, and extracting a corresponding new energy grid connection system change parameter set, wherein the new energy grid connection system change parameter set comprises power, frequency, current and voltage parameters. And establishing a nonlinear data model according to the change parameter set of the new energy grid-connected system by utilizing polynomial regression, and obtaining the relation among power, frequency, current and voltage in the new energy grid-connected system. And (5) fitting the nonlinear model through a least square method, and determining coefficients in the nonlinear data model. For example, wind energy data obtained from a new energy grid-connected system includes wind speed data at different points in time, solar energy data includes illumination intensity data at different points in time, and water energy data includes water flow speed data at different points in time. First, data cleaning and denoising operations are performed to remove missing values, outliers, and noise data that may exist. If the wind speed data at a certain point in time exceeds the measuring range of the device, the data can be regarded as an abnormal value and processed. Next, a new energy grid-connected system variation parameter set is extracted, including power, frequency, current and voltage parameters. According to the wind energy, solar energy and water energy data, wind energy power generation, solar energy power generation and water energy power generation at corresponding time points can be calculated. Meanwhile, according to the power data, current and voltage parameters at corresponding time points can be calculated. And establishing a nonlinear data model according to the change parameter set of the new energy grid-connected system by utilizing polynomial regression. To study the relationship between power and frequency, a nonlinear function model can be built by fitting the data through polynomial regression. The following quadratic polynomial regression model can be used: power = a0+a1 frequency + a2 frequency ≡2, where a0, a1 and a2 are coefficients of the regression model, fitting by least square method is required. And (5) fitting a nonlinear model through a least square method, and determining coefficients in the model. The least squares method is a commonly used parameter estimation method, and determines parameters of a model by minimizing the sum of squares of residuals between actual observed values and model predicted values. For the quadratic polynomial regression model described above, the least squares method may be used to estimate the values of coefficients a0, a1, and a 2. And finally, predicting the relation among power, frequency, current and voltage in the new energy grid-connected system by using the determined nonlinear data model. By inputting different frequency values, corresponding power prediction values can be obtained. Coefficients of a quadratic polynomial regression model obtained by least square fitting are a0=10, a1=5, a2= -1.
Step S102, based on the space-time information of the new energy grid connection, extracting and classifying a corresponding new energy grid connection space-time change parameter set, wherein the new energy grid connection space-time change parameter set comprises geographic position, time, power and frequency.
And acquiring the space-time information of new energy grid connection from the new energy grid connection system. And preprocessing the space-time information of the new energy grid connection, including data cleaning, denoising and outlier processing, and extracting a corresponding new energy grid connection space-time change parameter set, wherein the new energy grid connection space-time change parameter set comprises geographic position, time, power and frequency. And determining a corresponding cluster number K by using a K-means clustering algorithm, and classifying the geographic positions. If the classification is successful, the time is coded by unique identification by using a Murmur Hash3 algorithm, and a geographic position and a time identification set are obtained. And adopting a K-means clustering algorithm to classify the power and the frequency respectively to obtain a power identification set and a frequency identification set, and correlating the power identification set and the frequency identification set with the geographic position and the time identification set. And if the association is successful, combining the power identification set, the frequency identification set, the geographic position and the time identification set. For example, the following data is obtained from a new energy grid-connected system, in geographical location (390,1140), time 2021-01-0112:00:00, power 1000kW, frequency 50Hz. First, data is subjected to cleaning and denoising. The detection of the presence of some outliers in the power data, e.g. 10000kW, may be replaced with a value within a reasonable range, e.g. 1000kW. Next, a classification number K for the geographic location is determined using a K-means clustering algorithm. Let k=3, the geographic locations (390,1140) are divided into three categories A, B and C according to the K-means algorithm. The time is then uniquely identification coded using the MurmurHash3 algorithm. After encoding, 2021-01-0112:00:00 is encoded as abc123. Next, the power and frequency are classified separately using a K-means clustering algorithm. Power is divided into two categories: low power and high power; frequencies are divided into two categories: normal frequency and abnormal frequency. Finally, combining the power identification set, the frequency identification set, and the geographic position and time identification set. The end result may be a geographic location a, time identified as abc123, power identified as high power, and frequency identified as normal frequency.
Step S103, integrating the new energy grid-connected system change parameter set and the new energy grid-connected space-time change parameter set into a new energy grid-connected change parameter set.
And acquiring the new energy grid-connected system change parameter set and the new energy grid-connected space-time change parameter set. And matching and aligning the data, and corresponding the new energy grid-connected system change parameter set and the new energy grid-connected space-time change parameter set. And integrating the matched and aligned new energy grid-connected system change parameter set and the new energy grid-connected space-time change parameter set into a new energy grid-connected change parameter set according to a required format. And comparing and checking consistency of the power and frequency data in the new energy grid-connected system change parameter set and the new energy grid-connected space-time change parameter set. How the differences exist, the data adjustment and calibration is performed by linear fitting. The data is updated periodically to reflect the latest state of system operation. For example, there is a new energy grid-connected system, and its variable parameter sets include wind speed, solar radiation intensity, load demand, etc. The space-time parameter set comprises system power generation efficiency, system power loss, system frequency stability and the like. First, variable parameter data and corresponding space-time parameter data for system operation over a period of time are collected. The wind speed per hour, solar radiation intensity and load demand data, as well as the power generation efficiency, power loss and frequency stability data of the system are recorded. Next, the change parameter set and the space-time parameter set are matched and aligned according to the time stamp. The wind speed, solar radiation intensity and load demand data of each hour are corresponding to the corresponding system power generation efficiency, power loss and frequency stability data through time stamps. And then integrating the matched and aligned change parameter set and the space-time parameter set into a new energy grid-connected change parameter set according to a required format. The wind speed, solar radiation intensity and load demand data for each hour and the corresponding system power generation efficiency, power loss and frequency stability data may be integrated into a table in rows. Next, power and frequency data in the change parameter set and the space-time parameter set are compared and checked for consistency. The system power generation efficiency and power loss data for each hour may be compared to corresponding wind speed, solar radiation intensity and load demand data. If there is a discrepancy, a linear fitting method can be used for data adjustment and calibration. If the system power generation efficiency data has a linear relation with the wind speed data, the system power generation efficiency data can be adjusted by using a linear fitting method so as to keep the same with the wind speed data. Finally, the data is updated periodically to reflect the latest state of system operation. The change parameter set and the space-time parameter set are updated once every hour or every day to ensure the accuracy and the real-time of the data.
And step S104, carrying out correlation analysis on the new energy grid-connected variation parameter set, and screening parameters which possibly cause faults in new energy grid connection.
And acquiring the new energy grid-connected variation parameter set from a new energy grid-connected system, wherein the new energy grid-connected variation parameter set comprises power, frequency, current and voltage parameters. The pearson correlation coefficient between the parameters is calculated. And judging the value range of the correlation coefficient, and determining the correlation degree between the parameters, wherein the higher the correlation degree is, the larger the risk of the multiple collinearity problem is. And screening out parameters with absolute values of correlation higher than a preset correlation threshold value as parameters possibly causing faults. And further verifying the screened parameters which possibly cause faults through binary variance analysis, and judging whether obvious differences exist between the parameters possibly causing the faults. For example, the parameter set acquired by the new energy grid-connected system includes power, frequency, current and voltage parameters. These parameters may be used to calculate pearson correlation coefficients between each pair of parameters to evaluate the relationship between them. When there is a high correlation between multiple parameters, multiple collinearity problems may arise. The following parameter values were obtained, power [100,120,110,90,95], frequency [50,52,51,53,49], current [10,12,11,9,8], voltage [220,230,225,215,210]. First, a correlation coefficient between power and other parameters may be calculated, a correlation coefficient between power and frequency is 0.981, a correlation coefficient between power and current is-0.79, and a correlation coefficient between power and voltage is 0.75. The range of values of these correlation coefficients may then be determined to determine the degree of correlation between the parameters. The correlation coefficient can take the value range from-1 to 1, wherein-1 represents a complete negative correlation, 0 represents no correlation, and 1 represents a complete positive correlation. From the above calculation results, the following can be concluded: the correlation coefficient between power and frequency is close to 1, indicating a high positive correlation between them. The correlation coefficient between power and current is close to-1, indicating a very high negative correlation between them. The correlation coefficient between power and voltage is also close to-1, indicating a very high negative correlation between them. Then, parameters with absolute values of correlation higher than a preset correlation threshold can be screened out according to the threshold. Setting the correlation threshold to 0.8, the absolute value of the correlation between power and frequency is higher than the threshold, so that power and frequency can be taken as parameters that may cause a fault. Finally, the screened parameters that may lead to failure may be further validated using binary analysis of variance. Binary analysis of variance may be used to compare differences between two or more groups to determine if there are significant differences.
And step 105, analyzing the new energy grid-connected variation parameter set by using a support vector machine algorithm in the framework of the nonlinear data model to obtain a mode possibly causing new energy grid-connected faults.
And acquiring the power, frequency, current and voltage parameters with normal histories and a known fault parameter set from the new energy grid-connected system. From known fault samples, parameter sets are labeled as six categories of voltage anomalies, frequency offsets, current overloads, unbalanced voltages or currents, over voltages or over currents, and power quality problems. The support vector machine is trained using the marked parameter set, and the nonlinear relation is processed by using a polynomial kernel function under the framework of a nonlinear data model. And using a trained support vector machine model to conduct classified prediction on the unlabeled parameter set. The performance and accuracy of the model were evaluated using cross-validation. And identifying a mode which possibly causes grid-connected faults of the new energy according to a prediction result of the support vector machine. For example, there is a new energy grid-connected system comprising 10 generators. The power, frequency, current and voltage parameters of each generator are recorded by the monitoring system. Some fault samples have also been identified and they are categorized into the following six categories: voltage anomalies, frequency offsets, current overloads, unbalanced voltages or currents, over voltages or currents, and power quality issues. Now, a support vector machine model is trained using these known failure samples, and is used to classify and predict unlabeled parameter sets. The polynomial kernel function is chosen to handle the nonlinear relationship. 100 samples have been collected, containing parameter sets for different fault categories. Each parameter set contains four parameters, power, frequency, current and voltage. These parameters are converted into numerical values for processing. Cross-validation may be used to evaluate the performance and accuracy of the model. The dataset may be divided into 5 subsets, then trained sequentially using 4 subsets, tested using the remaining 1 subset, and then accuracy calculated. This procedure was repeated 5 times, and finally the average accuracy was taken as the evaluation result. Through training and cross validation, a support vector machine model with the accuracy of 90% is obtained. The model can now be used to classify predictions of unlabeled parameter sets. There is a new set of parameters, where the power is 1000kW, the frequency is 50Hz, the current is 50A, and the voltage is 220V. This parameter set is input into a support vector machine model that predicts that it belongs to the class of current overload.
And step S106, monitoring the real-time data stream of the new energy grid connection, and searching whether a real-time mode matched with the mode possibly causing the new energy grid connection fault exists or not.
And acquiring real-time data flow including power, frequency, current and voltage parameters from the new energy grid-connected system. And matching the acquired real-time data stream with the mode which possibly causes the grid-connected fault of the new energy. And triggering a corresponding fault alarm according to the pattern matching result, and performing fault detection. For example, there is a new energy grid-tie system that includes a plurality of solar and wind generators. Possible faults are detected by monitoring the real-time data flow. First, real-time power, frequency, current and voltage parameters may be obtained from the system. The following data were obtained with a power of 100kW, a frequency of 60Hz, a current of 150A and a voltage of 480V. Next, these real-time data need to be matched to the patterns that may cause the failure. It is known that failure modes may occur that have too high a voltage, i.e. a voltage exceeding 500V, a current overload, i.e. a current exceeding 200A, and an abnormal frequency, i.e. a frequency less than 50Hz or greater than 70Hz. According to the above mode, the acquired real-time data can be evaluated and analyzed. Both current and voltage parameters can be seen to be within normal ranges. Based on the pattern matching result, it can be determined that the system is not currently suffering any failure. Therefore, there is no need to trigger any fault alarms. If the parameters in the real-time data stream are matched with the fault mode, the voltage exceeds 500V or the current exceeds 200A, and a corresponding fault alarm is triggered for fault detection.
Step S107, if a real-time mode matched with the mode possibly causing the new energy grid-connected fault is obtained in the real-time data stream, the new energy grid-connected fault is predicted through a decision tree algorithm, and a final fault prediction result is obtained.
And if the pattern matching is successful, acquiring the historical power, frequency, current and voltage parameters from the new energy grid-connected system. And using the historical fault data as a training set, and modeling the fault type and the fault probability by using a decision tree algorithm to generate a preliminary fault prediction result. And acquiring power grid environment parameters from the new energy grid-connected system, and integrating the power grid environment parameters into a prediction result for adjustment, wherein the power grid environment parameters comprise air temperature and humidity. Judging whether the fault probabilities corresponding to different fault types in the adjusted fault prediction result exceed a preset fault probability threshold value or not. And if the fault probability threshold value exceeds the preset fault probability threshold value, executing automatic power grid dispatching and optimizing. For example, a solar power plant is being subjected to fault prediction and grid dispatching optimization. Historical power, frequency, current and voltage parameter data are obtained from the new energy grid-connected system. Firstly, using historical fault data as a training set, and modeling fault types and fault probabilities by utilizing a decision tree algorithm. The fault type can be predicted from the abnormal change of the power and the probability of different fault types can be calculated. Next, grid environmental parameters, such as air temperature and humidity, are obtained from the new energy grid-connected system. The current air temperature is 30 ℃ and the humidity is 60%. And integrating the power grid environment parameters into the prediction result for adjustment. The probability of different fault types can be corrected according to the air temperature and the humidity. If high temperature and high humidity may lead to an increased probability of certain fault types, the prediction results will be adjusted accordingly. And finally, judging whether the fault probabilities corresponding to different fault types in the adjusted fault prediction result exceed a preset fault probability threshold value. If the probability threshold value of a certain fault type is preset to be 8 and the probability of the predicted result showing the fault type is 9, it can be judged that the fault probability has exceeded the preset threshold value. If the judgment exceeds the preset fault probability threshold, automatic power grid dispatching and optimization are performed. The fault can be handled by adjusting the running state of the solar generator set, increasing the investment of the standby generator, or adjusting the load of the power grid and the like.
The foregoing disclosure is illustrative of the preferred embodiments of the present invention, and is not to be construed as limiting the scope of the invention, as it is understood by those skilled in the art that all or part of the above-described embodiments may be practiced with equivalents thereof, which fall within the scope of the invention as defined by the appended claims.
Claims (8)
1. The new energy grid-connected fault prediction method based on the nonlinear data model is characterized by comprising the following steps of:
based on wind energy, solar energy and water energy data of new energy grid connection, extracting a corresponding new energy grid connection system change parameter set, and constructing a nonlinear data model, wherein the new energy grid connection change parameter set comprises power, frequency, current and voltage parameters; based on the space-time information of new energy grid connection, extracting and classifying a corresponding new energy grid connection space-time change parameter set, wherein the new energy grid connection space-time change parameter set comprises geographic position, time, power and frequency; integrating the new energy grid-connected system change parameter set and the new energy grid-connected space-time change parameter set into a new energy grid-connected change parameter set; performing correlation analysis on the new energy grid-connected variation parameter set, and screening parameters which possibly cause faults in new energy grid connection; in the framework of the nonlinear data model, analyzing the new energy grid-connected variation parameter set by using a support vector machine algorithm to obtain a mode possibly causing new energy grid-connected faults; monitoring a real-time data stream of new energy grid connection, and searching whether a real-time mode matched with the mode possibly causing new energy grid connection fault exists or not; and if a real-time mode matched with the mode which possibly causes the new energy grid-connected fault is obtained in the real-time data stream, carrying out new energy grid-connected fault prediction through a decision tree algorithm, and obtaining a final fault prediction result.
2. The method of claim 1, wherein the extracting a corresponding new energy grid-connected system variation parameter set based on new energy grid-connected wind, solar and water energy data, and constructing a nonlinear data model, wherein the new energy grid-connected variation parameter set includes power, frequency, current and voltage parameters, includes:
acquiring wind energy, solar energy and water energy data of new energy grid connection from a new energy grid connection system; preprocessing wind energy, solar energy and water energy data of the new energy grid connection, including data cleaning, denoising and outlier processing, and extracting a corresponding new energy grid connection system change parameter set, wherein the new energy grid connection system change parameter set comprises power, frequency, current and voltage parameters; establishing a nonlinear data model according to the change parameter set of the new energy grid-connected system by utilizing polynomial regression to obtain the relationship among power, frequency, current and voltage in the new energy grid-connected system; and (5) fitting the nonlinear model through a least square method, and determining coefficients in the nonlinear data model.
3. The method of claim 1, wherein the extracting and classifying the corresponding new energy grid-connected spatiotemporal variation parameter sets based on the new energy grid-connected spatiotemporal information includes geographic location, time, power and frequency, and includes:
acquiring space-time information of new energy grid connection from a new energy grid connection system; preprocessing the space-time information of the new energy grid connection, including data cleaning, denoising and outlier processing, and extracting a corresponding new energy grid connection space-time change parameter set, wherein the new energy grid connection space-time change parameter set comprises geographic position, time, power and frequency; determining a corresponding cluster number K by using a K-means clustering algorithm, and classifying the geographic position; if the classification is successful, the Murmur Hash3 algorithm is used for carrying out unique identification coding on the time to obtain a geographic position and a time identification set; classifying power and frequency by adopting a K-means clustering algorithm to obtain a power identification set and a frequency identification set, and correlating the power identification set and the frequency identification set with the geographic position and the time identification set; and if the association is successful, combining the power identification set with the frequency identification set, and combining the geographic position with the time identification set.
4. The method of claim 1, wherein the integrating the new energy grid-tie system variation parameter set and the new energy grid-tie spatio-temporal variation parameter set into one new energy grid-tie variation parameter set comprises:
acquiring a change parameter set of the new energy grid-connected system and a new energy grid-connected space-time change parameter set; matching and aligning the data, and corresponding the new energy grid-connected system change parameter set and the new energy grid-connected space-time change parameter set; integrating the matched and aligned new energy grid-connected system change parameter set and the new energy grid-connected space-time change parameter set into a new energy grid-connected change parameter set according to a required format; comparing and checking consistency of the power and frequency data in the new energy grid-connected system change parameter set and the new energy grid-connected space-time change parameter set; how the differences exist, performing data adjustment and calibration through linear fitting; the data is updated periodically to reflect the latest state of system operation.
5. The method of claim 1, wherein the performing correlation analysis on the new energy grid-connected variation parameter set, and screening parameters that may cause a fault in new energy grid connection, includes:
acquiring the new energy grid-connected variation parameter set from a new energy grid-connected system, wherein the new energy grid-connected variation parameter set comprises power, frequency, current and voltage parameters; calculating pearson correlation coefficients between the parameters; judging the value range of the correlation coefficient, and determining the correlation degree between parameters, wherein the higher the correlation degree is, the larger the risk of the multiple collinearity problem is; screening out parameters with absolute values of correlation higher than a preset correlation threshold value as parameters possibly causing faults; and further verifying the screened parameters which possibly cause faults through binary variance analysis, and judging whether obvious differences exist between the parameters possibly causing the faults.
6. The method of claim 1, wherein the analyzing the new energy grid-tie change parameter set using a support vector machine algorithm within the framework of the nonlinear data model to obtain a pattern that may cause a new energy grid-tie fault comprises:
acquiring the power, frequency, current and voltage parameters with normal histories and a known fault parameter set from a new energy grid-connected system; marking the parameter sets as six categories of voltage anomalies, frequency offsets, current overloads, unbalanced voltages or currents and over-voltages or over-currents, and power quality problems, based on known fault samples; training a support vector machine by using the marked parameter set, and selecting to use a polynomial kernel function to process a nonlinear relation under the framework of a nonlinear data model; classifying and predicting the unlabeled parameter set by using a trained support vector machine model; using cross-validation, evaluating the performance and accuracy of the model; and identifying a mode which possibly causes grid-connected faults of the new energy according to a prediction result of the support vector machine.
7. The method of claim 1, wherein the monitoring the new energy grid-tie real-time data stream for the presence of a real-time pattern matching the pattern that may cause the new energy grid-tie fault comprises:
acquiring real-time data flow including power, frequency, current and voltage parameters from a new energy grid-connected system; matching the acquired real-time data stream with the mode which possibly causes the grid-connected fault of the new energy; and triggering a corresponding fault alarm according to the pattern matching result, and performing fault detection.
8. The method according to claim 1, wherein if the real-time data stream obtains a real-time pattern matching the pattern that may cause the new energy grid-connected fault, performing new energy grid-connected fault prediction through a decision tree algorithm to obtain a final fault prediction result, including:
if the pattern matching is successful, acquiring historical power, frequency, current and voltage parameters from a new energy grid-connected system; using historical fault data as a training set, and modeling the fault type and the fault probability by utilizing a decision tree algorithm to generate a preliminary fault prediction result; acquiring power grid environment parameters from a new energy grid-connected system, and integrating the power grid environment parameters into a prediction result to adjust, wherein the power grid environment parameters comprise air temperature and humidity; judging whether the fault probabilities corresponding to different fault types in the adjusted fault prediction result exceed a preset fault probability threshold value or not; and if the fault probability threshold value exceeds the preset fault probability threshold value, executing automatic power grid dispatching and optimizing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311488370.XA CN117709508A (en) | 2023-11-09 | 2023-11-09 | New energy grid-connected fault prediction method based on nonlinear data model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311488370.XA CN117709508A (en) | 2023-11-09 | 2023-11-09 | New energy grid-connected fault prediction method based on nonlinear data model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117709508A true CN117709508A (en) | 2024-03-15 |
Family
ID=90146832
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311488370.XA Pending CN117709508A (en) | 2023-11-09 | 2023-11-09 | New energy grid-connected fault prediction method based on nonlinear data model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117709508A (en) |
-
2023
- 2023-11-09 CN CN202311488370.XA patent/CN117709508A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111539550B (en) | Method, device, equipment and storage medium for determining working state of photovoltaic array | |
| CN111047082A (en) | Early warning method and device for equipment, storage medium and electronic device | |
| Li et al. | Fault diagnosis of photovoltaic panels using full I–V characteristics and machine learning techniques | |
| CN109660206B (en) | Wasserstein GAN-based photovoltaic array fault diagnosis method | |
| Gonzalez et al. | On the use of high-frequency SCADA data for improved wind turbine performance monitoring | |
| CN108627720B (en) | Power equipment state monitoring method based on Bayesian algorithm | |
| CN112611936A (en) | Distribution network transformer fault dynamic detection and classification system based on edge calculation | |
| Yesilbudak | Partitional clustering-based outlier detection for power curve optimization of wind turbines | |
| CN106529786A (en) | Power distribution network load calibration method and power distribution network load calibration device based on big data analysis | |
| CN116739829B (en) | Big data-based power data analysis method, system and medium | |
| Tao et al. | Abnormal detection of wind turbine based on SCADA data mining | |
| CN111275295A (en) | Distributed photovoltaic fault diagnosis method based on inverse distance weight interpolation | |
| Huang et al. | Wind turbine health assessment framework based on power analysis using machine learning method | |
| CN110096723B (en) | High-voltage switch cabinet insulation state analysis method based on operation and maintenance detection big data | |
| CN117494009A (en) | Electrical equipment state evaluation method based on insulating material pyrolysis analysis and cloud platform | |
| CN117235617A (en) | ML-RFKNN-based photovoltaic array fault diagnosis method in sand and dust weather | |
| CN117131321A (en) | Electric energy data acquisition method | |
| CN117709508A (en) | New energy grid-connected fault prediction method based on nonlinear data model | |
| CN111506636A (en) | System and method for analyzing residential electricity consumption behavior based on autoregressive and neighbor algorithm | |
| Platon et al. | Modelling the power production of a photovoltaic system: Comparison of sugeno-type fuzzy logic and PVSAT-2 models | |
| Heres et al. | Machine learning methods for the health-indexing and ranking of underground distribution cables and joints | |
| CN114236308A (en) | Feeder line information-based power distribution network fault detection and positioning method | |
| CN112682269A (en) | Wind turbine generator state monitoring method based on OC-RKELM | |
| Talayero et al. | Diagnosis of failures in solar plants based on performance monitoring | |
| Sarquis Filho et al. | Analysis of automatic fault detection methods for commercially operated pv power plants |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |