Intelligent load regulation and control system for medium-voltage AC / DC hybrid engineering

Abstract

The invention relates to the technical field of load regulation and control, in particular to an intelligent load regulation and control system for medium-voltage alternating current and direct current hybrid engineering, the system comprises a processor and a memory, and the processor executes a computer program stored in the memory to realize the following steps: obtaining a stability degree characterization value of random load data to be analyzed; according to the stability degree representation value of the random to-be-analyzed load data, adjusting the initial coefficient of the to-be-analyzed load data to obtain a target coefficient of the random to-be-analyzed load data, and taking the initial coefficient of the non-random to-be-analyzed load data as the target coefficient of the corresponding non-random to-be-analyzed load data; and according to the ARIMA model and the target coefficient of the to-be-analyzed load data, obtaining predicted load data at the current monitoring moment, and according to the predicted load data, regulating and controlling the load of the medium-voltage AC / DC hybrid engineering power system. According to the invention, the reliability and accuracy of regulating and controlling the load of the medium-voltage AC / DC hybrid engineering power system can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of load regulation, and particularly to an intelligent load regulation system for a medium-voltage AC / DC hybrid project. Background Art

[0002] Currently, in order to optimize the overall performance and reliability of the power system, a medium-voltage AC / DC hybrid project is applied to the power system, and the power system applying the medium-voltage AC / DC hybrid project is called a medium-voltage AC / DC hybrid project power system, and the medium-voltage AC / DC hybrid project power system also refers to a power system that uses both alternating current and direct current.

[0003] Moreover, since the load regulation of the medium-voltage AC / DC hybrid project power system is crucial for aspects such as the safe and stable operation of the power system, power system faults, and the economy and reliability of the power system, currently, in order to ensure the safe and stable operation of the medium-voltage AC / DC hybrid project power system, prevent power system faults of the medium-voltage AC / DC hybrid project power system caused by load changes, and improve the economy and reliability of the medium-voltage AC / DC hybrid project power system, it is necessary to regulate the load of the medium-voltage AC / DC hybrid project power system.

[0004] In the prior art, generally, prediction means are used to achieve the purpose of load regulation, that is, in the prior art, generally, based on the historical load data collected on the demand side of the medium-voltage AC / DC hybrid project power system, the ARIMA model is used to predict the load data, and then based on the obtained prediction results, the load of the medium-voltage AC / DC hybrid project power system is regulated; however, the demand-side load data is affected by user human factors and external environmental factors, and these effects will cause random load data to appear in the collected load data time series, and the appearance of random load data will affect the periodicity and trend, and further lead to low reliability and accuracy of the prediction results obtained by using the ARIMA model. When the reliability and accuracy of the prediction results are low, it may cause problems such as inaccurate or unreliable regulation when subsequently regulating the load of the medium-voltage AC / DC hybrid project power system based on the prediction results. Therefore, how to obtain accurate and reliable prediction results to improve the reliability and accuracy when regulating the load of the medium-voltage AC / DC hybrid project power system has become an urgent problem to be solved. Summary of the Invention

[0005] In order to solve the above problems, the present invention provides an intelligent load regulation system for a medium-voltage AC / DC hybrid project, and the specific technical solution adopted is as follows: An embodiment of the present invention provides an intelligent load regulation system for a medium-voltage AC / DC hybrid project, including a processor and a memory, and the processor executes the computer program stored in the memory to implement the following steps: Obtain the load data sequence to be analyzed, where the load data to be analyzed in the load data sequence to be analyzed is the load data on the demand side of the medium-voltage AC / DC hybrid power system; According to the ARIMA model, obtain the coefficients of each load data to be analyzed in the load data sequence to be analyzed, and denote them as the initial coefficients; According to the adjacent load data to be analyzed of each load data to be analyzed in the load data sequence to be analyzed, obtain the random load data to be analyzed and the non-random load data to be analyzed in the load data sequence to be analyzed; Obtain the random historical load data, the random data points corresponding to the random load data to be analyzed and the random historical load data, and use the clustering algorithm to cluster the random data points to obtain each cluster. According to the minimum circumscribed circle of the cluster and the nearest neighbor random data points of the random data points in the cluster, obtain the stability degree characterization value of the random load data to be analyzed; According to the stability degree characterization value of the random load data to be analyzed, adjust the initial coefficients of the load data to be analyzed to obtain the target coefficients of the random load data to be analyzed, and use the initial coefficients of the non-random load data to be analyzed as the target coefficients of the corresponding non-random load data to be analyzed; According to the ARIMA model and the target coefficients of the load data to be analyzed, obtain the predicted load data at the current monitoring moment, and regulate the load of the medium-voltage AC / DC hybrid power system according to the predicted load data.

[0006] Beneficial effect: The present invention first obtains a sequence of load data to be analyzed; then, according to the ARIMA model, obtains the coefficients of each load data to be analyzed in the sequence of load data to be analyzed, and records them as initial coefficients; then, according to the adjacent load data to be analyzed of each load data to be analyzed in the sequence of load data to be analyzed, obtains random load data to be analyzed and non-random load data to be analyzed in the sequence of load data to be analyzed; then, obtains random historical load data and random data points corresponding to the random load data to be analyzed and the random historical load data, and clusters the random data points using a clustering algorithm to obtain various cluster clusters, and according to the minimum The nearest neighbor random data points of the random data points in the circumscribed circle and the clusters are used to obtain the stability characterization value of the random load data to be analyzed; then, according to the stability characterization value of the random load data to be analyzed, the initial coefficient of the load data to be analyzed is adjusted to obtain the target coefficient of the random load data to be analyzed, and the initial coefficient of the non-random load data to be analyzed is used as the target coefficient of the corresponding non-random load data to be analyzed; finally, according to the ARIMA model and the target coefficient of the load data to be analyzed, the predicted load data at the current monitoring time is obtained, and the load of the medium-voltage AC / DC hybrid engineering power system is regulated according to the predicted load data. The present invention can make the prediction result more accurate and reliable based on the target coefficient obtained by adjusting the initial coefficient, and based on the more accurate and reliable prediction result, it can improve the reliability and accuracy of the load regulation of the medium-voltage AC / DC hybrid engineering power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0008] Figure 1 The present invention is a flow chart of an intelligent load control method for a medium voltage AC / DC hybrid project. DETAILED DESCRIPTION

[0009] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the embodiments of the present invention.

[0010] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0011] This embodiment provides an intelligent load regulation system for a medium-voltage AC / DC hybrid project, including a processor and a memory. The processor executes the computer program stored in the memory to implement an intelligent load regulation method for a medium-voltage AC / DC hybrid project, as Figure 1 shown. The intelligent load regulation method for a medium-voltage AC / DC hybrid project includes the following steps: Step S001: Obtain the load data sequence to be analyzed.

[0012] Currently, the ARIMA model is generally used to predict the load data on the demand side of the medium-voltage AC / DC hybrid project power system, and then the load of the medium-voltage AC / DC hybrid project power system is regulated based on the prediction result. However, since the load data on the demand side is affected by user human factors and external environmental factors, these effects will cause random load data to appear in the collected load data time series. The appearance of random load data will affect the periodicity and trend. Also, since the reliability and accuracy of the prediction result mainly depend on the periodicity and trend in the data, when the influence of random load data on the data periodicity and trend is large, it may lead to low reliability and accuracy of the prediction result. When the reliability and accuracy of the prediction result are low, it will lead to inaccurate or unreliable regulation when the subsequent load of the medium-voltage AC / DC hybrid project power system is regulated based on the prediction result, such as an incorrect selection of the regulation mode.

[0013] In order to ensure the reliability and accuracy of load regulation, in the following embodiments, the coefficients of the data will be optimized or adaptively adjusted to improve the reliability and accuracy of the prediction results. That is, in the process of using the ARIMA model for prediction, it is necessary to obtain the coefficients of each data in the input data sequence. The main purpose of this embodiment is to optimize or adaptively adjust the coefficients of the obtained data, and then complete the prediction based on the optimized or adaptively adjusted coefficients, and complete the load regulation of the medium-voltage AC-DC hybrid power system based on the prediction results. The characteristics of the random load data are that there is no obvious pattern, and each random load data is independent and irrelevant, that is, the distribution or change of the random load data in the sequence is irregular and unpredictable. It should be noted that in the process of using the ARIMA model for prediction, only the coefficients of the data are adjusted, and the adjusted coefficients are used to complete the prediction later, without changing or adjusting other links or process steps in the prediction. In addition, for the convenience of analysis and understanding, in the following embodiments, the load regulation process of any already operating medium-voltage AC-DC hybrid power system will be analyzed and described. Therefore, all load data that appears in the following embodiments are the load data on the demand side of the medium-voltage AC-DC hybrid power system, and all power systems that appear later are the medium-voltage AC-DC hybrid power system. And the demand side of the power system refers to the demand and use of electricity by consumers in the power supply system. It is an important part of the power market and covers the needs of different types of users such as households, businesses, industries, and agriculture.

[0014] Since, for the effectiveness of load regulation, the load of the medium-voltage AC-DC hybrid power system is generally regulated in real time, this embodiment needs to obtain the predicted load data at the current monitoring moment. If you want to obtain the predicted load data at the current monitoring moment, you need to use the historical data time series to complete the prediction. Therefore, this embodiment will next obtain the load data sequence to be analyzed, and the load data sequence to be analyzed is mainly used to obtain the predicted load data at the current monitoring moment. Then the specific acquisition process of the load data sequence to be analyzed is as follows: First, obtain the current monitoring time interval, that is, the current monitoring time period. Then, obtain the load data of the demand side of the medium-voltage AC-DC hybrid power system collected by the data acquisition module in the power system at each monitoring moment within the current monitoring time interval. Denote the time series sequence composed of all the load data of the demand side of the medium-voltage AC-DC hybrid power system collected within the current monitoring time interval as the load data sequence to be analyzed, and denote all the load data in the load data sequence to be analyzed as the load data to be analyzed. Moreover, the collected load data will be stored in the terminal database for subsequent data retrieval. Additionally, since it is necessary to use the load data sequence to be analyzed to predict the predicted load data at the current monitoring moment, it is required that the current monitoring time interval is adjacent to the current monitoring moment, but the current monitoring time interval does not include the current monitoring moment.

[0015] And in specific applications, the implementer also needs to set the acquisition interval between the current monitoring time interval and the adjacent monitoring moments according to the actual situation. For example, in this embodiment, the time length of the current monitoring time interval can be set to 1 hour, and the acquisition interval between adjacent monitoring moments can be set to 1 second.

[0016] Therefore, through the above process, this embodiment obtains the load data sequence to be analyzed.

[0017] Step S002: According to the ARIMA model, obtain the coefficients of each load data to be analyzed in the load data sequence to be analyzed, and denote them as the initial coefficients.

[0018] Since this embodiment mainly optimizes or adaptively adjusts the obtained coefficients, this embodiment needs to first obtain the coefficients of each load data to be analyzed in the load data sequence to be analyzed. The specific obtaining process is as follows: First, construct an ARIMA model, and import the load data sequence to be analyzed into the constructed ARIMA model. Obtain the coefficients of each load data to be analyzed in the load data sequence to be analyzed, and denote the coefficients of the load data to be analyzed obtained at this time as the initial coefficients corresponding to the load data to be analyzed. Moreover, the initial coefficients of the load data to be analyzed are determined by the traditional coefficient obtaining method in the ARIMA model. For example, the maximum likelihood (MLE) method can be used to determine the initial coefficients of the load data to be analyzed. And since the coefficients of the load data to be analyzed include the autoregressive coefficients in the AR model and the moving average coefficients in the MA, the initial coefficient of each load data to be analyzed includes the initial autoregressive coefficient and the initial moving average coefficient. Additionally, it should be noted that the obtaining method of the initial coefficients of the load data to be analyzed is a well-known technology, so this embodiment will not describe it in detail.

[0019] Therefore, in this embodiment, the initial coefficients of each load data to be analyzed in the obtained load data sequence to be analyzed are obtained through the above process, that is, the initial autoregressive coefficient and the initial moving average coefficient of the load data to be analyzed.

[0020] Step S003: Obtain the random load data to be analyzed and the non-random load data to be analyzed in the load data sequence to be analyzed according to the adjacent load data to be analyzed of each load data to be analyzed in the load data sequence to be analyzed.

[0021] Based on the above analysis, it can be seen that random load data is an important factor leading to unreliable and inaccurate prediction results. That is, the more random load data there is, the greater the impact on periodicity and trend, and thus the lower the accuracy and reliability of the obtained prediction results. Therefore, in this embodiment, the random load data to be analyzed and the non-random load data to be analyzed in the load data sequence to be analyzed will be obtained according to the adjacent load data to be analyzed of each load data to be analyzed in the load data sequence to be analyzed, and the specific acquisition process is as follows: First, obtain the outlier characterization value of each load data to be analyzed according to the adjacent load data to be analyzed of each load data to be analyzed in the load data sequence to be analyzed; and the specific acquisition process of the outlier characterization value of each load data to be analyzed is as follows: Perform STL time series decomposition on the load data sequence to be analyzed to obtain the residuals of each load data to be analyzed in the load data sequence to be analyzed, also known as residual terms or remainder terms; and the process of performing STL time series decomposition on the data sequence to obtain the trend term, periodic term, and residual term is a well-known technology, so this embodiment will not describe it in detail. After obtaining the residuals of the load data to be analyzed, for the convenience of understanding in this embodiment, the acquisition process of the outlier characterization value of the a-th load data to be analyzed in the load data sequence to be analyzed will be used as an example for description. Then, the acquisition process of the outlier characterization value of the a-th load data to be analyzed is as follows: First, obtain the absolute value of the difference between the residual of the a-th load data to be analyzed and the preset residual, and perform normalization processing on the absolute value of the difference between the residual of the a-th load data to be analyzed and the preset residual, and record the normalized value obtained by the normalization processing as the first index value corresponding to the a-th load data to be analyzed; and in specific applications, the implementer needs to set the preset residual according to the actual situation. For example, in this embodiment, the preset residual can be set to 0.

[0022] Then obtain the remaining set corresponding to the ath load data to be analyzed, and denote the set constructed by the residuals of all the load data to be analyzed in the remaining set as the residual set. The remaining set corresponding to the ath load data to be analyzed refers to the set constructed by all the load data to be analyzed in the load data sequence to be analyzed except the ath load data to be analyzed. Then obtain the mean value of the residual set, and denote it as the remaining residual mean value of the ath load data to be analyzed. Then obtain the absolute value of the difference between the residual of the ath load data to be analyzed and the remaining residual mean value, and denote it as the characteristic difference value of the ath load data to be analyzed. After obtaining the characteristic difference value, obtain the adjacent difference value of the ath load data to be analyzed. Then multiply the adjacent difference value of the ath load data to be analyzed by the characteristic difference value of the ath load data to be analyzed, and perform normalization processing on the obtained result. Denote the result obtained by the normalization processing as the second index value corresponding to the ath load data to be analyzed.

[0023] Finally, obtain the result of adding the first index value corresponding to the ath load data to be analyzed and its corresponding second index value, and perform normalization processing on the obtained result of the addition. Denote the result obtained by the normalization processing as the outlier characterization value of the ath load data to be analyzed.

[0024] In addition, in this embodiment, the process of obtaining the adjacent difference value of the ath load data to be analyzed is as follows: First, determine whether the value of a is 1. If it is determined that the value of a is 1, then denote the right difference value of the ath load data to be analyzed as the adjacent difference value of the ath load data to be analyzed. Otherwise, continue to determine whether the value of a is A. If it is determined that the value of a is A, then denote the left difference value of the ath load data to be analyzed as the adjacent difference value of the ath load data to be analyzed. If it is determined that the value of a is neither 1 nor A, then denote the sum of the right difference value and the left difference value of the ath load data to be analyzed as the adjacent difference value of the ath load data to be analyzed. The right difference value of the ath load data to be analyzed is the absolute value of the difference between the residual of the ath load data to be analyzed and the residual value of the (a + 1)th load data to be analyzed in the load data sequence to be analyzed. The left difference value of the ath load data to be analyzed is the absolute value of the difference between the residual of the ath load data to be analyzed and the residual value of the (a - 1)th load data to be analyzed in the load data sequence to be analyzed. A is the total number of data in the load data sequence to be analyzed.

[0025] In addition, in this embodiment, the specific expression for obtaining the outlier characterization value of the ath load data to be analyzed is: Among them, is the outlier characterization value of the ath load data to be analyzed, and norm() is a normalization function. is the absolute value of the difference between the residual of the ath load data to be analyzed and a preset residual. is the adjacent difference value of the ath load data to be analyzed. is the residual of the ath load data to be analyzed. is the mean value of the remaining residuals of the ath load data to be analyzed. is the characteristic difference value of the ath load data to be analyzed, C1 is a preset first constant, and C1 is used to perform normalization. And in specific applications, the implementer needs to set the value of C1 according to the actual situation, but it is required that C1 is greater than 2. For example, in this embodiment, C1 can be set to 3. Additionally, as another real-time method, the normalization function norm() can also be used to perform normalization processing.

[0026] And when is larger, it indicates that the periodicity and trend of the ath load data to be analyzed are worse, and it also indicates that the outlier degree of the ath load data to be analyzed is greater. When is larger, it indicates that the difference in trend and periodicity between the ath load data to be analyzed and its adjacent data is greater, and it also indicates that the outlier degree of the ath load data to be analyzed is greater. When is larger, it indicates that the difference between the residual of the ath load data to be analyzed and the residuals of other data is greater, and it also indicates that the outlier degree of the ath load data to be analyzed is greater. And when the outlier degree of the ath load data to be analyzed is greater, it more indicates that the probability that the ath load data to be analyzed is a random load data to be analyzed is greater. Also, because when , and are larger, is larger. Therefore, when is larger, it more indicates that the probability that the ath load data to be analyzed is a random load data to be analyzed is greater. On the contrary, when is smaller, it more indicates that the probability that the ath load data to be analyzed is a random load data to be analyzed is smaller.

[0027] Therefore, through the above process, this embodiment can obtain the outlier characterization value of each load data to be analyzed. After obtaining the outlier characterization value, the random load data to be analyzed and the non-random load data to be analyzed in the load data sequence to be analyzed are obtained according to the outlier characterization values of each load data to be analyzed. And the specific process is as follows: For any load data to be analyzed in the load data sequence to be analyzed, determine whether the outlier characterization value of the load data to be analyzed is not less than a preset outlier threshold. If so, mark the load data to be analyzed as random load data to be analyzed; otherwise, mark the load data to be analyzed as non-random load data to be analyzed. And in specific applications, the implementer needs to set the preset outlier threshold according to the actual situation. For example, in this embodiment, the preset outlier threshold can be set to 0.7.

[0028] Therefore, through the above process, this embodiment obtains the random load data to be analyzed and the non-random load data to be analyzed in the load data sequence to be analyzed. And in this embodiment, it is not necessary to adjust or optimize the initial coefficients of the non-random load data to be analyzed.

[0029] Step S004: Obtain the random historical load data, the random data points corresponding to the random load data to be analyzed and the random historical load data, and use a clustering algorithm to cluster the random data points to obtain each clustering cluster. According to the minimum circumscribed circle of the clustering cluster and the nearest neighbor random data points of the random data points in the clustering cluster, obtain the stability degree characterization value of the random load data to be analyzed.

[0030] After obtaining the random load data to be analyzed in this embodiment, it is necessary to analyze and obtain the stability degree of each random load data to be analyzed. And when the stability degree of the random load data to be analyzed is higher, it indicates that it can have a positive impact on the prediction in the ARIMA model. On the contrary, when the stability degree of the random load data to be analyzed is lower, it indicates that it can have a negative impact on the prediction in the ARIMA model. A positive impact can improve the reliability and accuracy of the prediction result, while a negative impact may reduce the reliability and accuracy of the prediction result. Therefore, this embodiment next needs to obtain the stability degree characterization value of the random load data to be analyzed, and subsequently, the initial coefficients are adjusted based on the obtained stability degree characterization value. However, in the process of obtaining the stability degree characterization value of the random load data to be analyzed, not only the random load data to be analyzed is referred to, but also the random historical load data needs to be referred to. Then, combined with the clustering results of the data points corresponding to the random historical load data and the random load data to be analyzed, the stability degree characterization value of the random load data to be analyzed is determined. Therefore, this embodiment will first obtain the random historical load data next. That is, the specific process of obtaining the random historical load data is as follows: First, in the historical operation time period of the medium-voltage AC / DC hybrid power system, obtain a preset number of historical time intervals that are the same as and closest to the current monitoring time interval, and all are recorded as historical monitoring time intervals; then obtain the load data sequence corresponding to each historical monitoring time interval, and all are recorded as the historical load data sequence corresponding to the corresponding historical monitoring time interval. The historical load data in the historical load data sequence also belongs to the load data on the demand side of the medium-voltage AC / DC hybrid power system, and all historical monitoring time intervals do not overlap with the current monitoring time interval, and the time interval between two adjacent historical monitoring time intervals in time is 24 hours, that is, the historical monitoring time interval closest to the current monitoring time interval in time also has a 24-hour interval from the current monitoring time interval; for example, if the current monitoring time interval is from 7:00 to 8:00 in the morning of the current day, then the historical monitoring time interval also refers to 7:00 to 8:00 in the morning, but on different days.

[0031] In addition, in specific applications, the implementer needs to set the value of the preset number according to the actual situation. For example, in this embodiment, the preset number can be set to 5; in addition, the acquisition time interval between adjacent load data in the historical monitoring time interval is the same as the acquisition time interval between adjacent load data in the current monitoring time interval.

[0032] After obtaining the historical load data sequence corresponding to each historical monitoring time interval, the method of obtaining the random to-be-analyzed load data in the to-be-analyzed load data sequence is used to obtain the random historical load data in the historical load data sequence. Since the acquisition methods are the same, this embodiment will not be described in detail.

[0033] Since the subsequent data points participate in clustering, this embodiment then obtains the random data points corresponding to the random to-be-analyzed load data and the random data points corresponding to the random historical load data. The specific acquisition process of the random data points is as follows: First, a two-dimensional space is constructed, where the horizontal axis of the two-dimensional space is the time marker value and the vertical axis is the load data. Then, all the random load data to be analyzed, all the random historical load data, the time marker values of the random load data to be analyzed, and the time marker values of the random historical load data are mapped into the two-dimensional space to obtain the data points corresponding to the random load data to be analyzed and the data points corresponding to the random historical load data, and all the obtained data points are denoted as random data points. The abscissa value of the random data point corresponding to any random load data to be analyzed is the time marker value of the random load data to be analyzed, and the ordinate value is the random load data to be analyzed. The coordinate meaning of the random data point corresponding to the random historical load data is the same as that of the random data point corresponding to the random load data to be analyzed. In addition, the time marker values of the f-th historical load data in the historical load data sequence and the f-th load data to be analyzed in the load data to be analyzed sequence are both f, that is, in the historical load data sequence and the load data to be analyzed sequence, the time marker values of the load data at the same position are the same.

[0034] After obtaining the random data points corresponding to the random load data to be analyzed and the random data points corresponding to the random historical load data, a clustering algorithm is used to cluster all the random data points in the two-dimensional space to obtain each clustering cluster, and subsequently, the clustering result will be analyzed to obtain the stability degree characterization value of the random load data to be analyzed. In addition, in specific applications, the implementer needs to select a clustering algorithm according to the actual situation. For example, in this embodiment, the k-means clustering algorithm is used to cluster the random data points, and the value of K during clustering, that is, the number of clustering centers, is determined by the elbow method. Since the process of clustering using the k-means clustering algorithm and the process of determining the number of clustering centers using the elbow method are well-known technologies, they will not be described in detail in this embodiment.

[0035] After obtaining the clustering clusters, the stability degree characterization value of each random load data to be analyzed is obtained according to the minimum circumscribed circle of each clustering cluster and the nearest neighbor random data points in each clustering cluster. For the convenience of understanding, in this embodiment, the process of obtaining the stability degree characterization value of any random load data g in the load data to be analyzed sequence will be used as an example for description, that is, the process of obtaining the stability degree characterization value of the load data g to be analyzed is as follows: First, obtain the clustering cluster containing the random data points corresponding to the random load data g to be analyzed, and denote it as clustering cluster G; then obtain the total number of random data points in clustering cluster G, and perform normalization processing on the total number of random data points in clustering cluster G, and denote the result of the normalization processing as the quantity characterization value; immediately afterwards, obtain the area of the minimum circumscribed circle of clustering cluster G, and perform a negative correlation mapping on the area of the minimum circumscribed circle of clustering cluster G, and denote the mapping result as the area characterization value; afterwards, obtain the sum of the quantity characterization value and the area characterization value, and denote it as the first distribution characteristic value of the random load data g to be analyzed.

[0036] After obtaining the first distribution characteristic value, obtain the nearest neighbor random data point of each random data point in clustering cluster G, and the nearest neighbor random data points of the random data points in clustering cluster G also all belong to clustering cluster G; then obtain the Euclidean distance between each random data point in clustering cluster G and its corresponding nearest neighbor random data point, and denote it as the nearest neighbor distance of the corresponding random data point, and the nearest neighbor distance of any random data point d in clustering cluster G refers to the Euclidean distance between random data point d and the nearest neighbor random data point of random data point d; afterwards, obtain the set composed of the nearest neighbor distances of all random data points in clustering cluster G, and denote it as the nearest neighbor distance set.

[0037] Immediately afterwards, obtain the product of the variance of the nearest neighbor distance set and the nearest neighbor distance of the random data points corresponding to the random load data g to be analyzed, and denote it as the first product value; then obtain the reciprocal of the result obtained by adding the first product value and a preset second constant as the second distribution characteristic value of the random load data g to be analyzed; and in specific applications, the implementer needs to set the preset second constant according to the actual situation. For example, in this embodiment, the preset second constant can be set to 1.

[0038] Finally, obtain the product of the first distribution characteristic value of the random load data g to be analyzed and the second distribution characteristic value of the random load data g to be analyzed, and denote the result of the multiplication as the stability degree characterization value of the random load data g to be analyzed.

[0039] In addition, the specific expression for obtaining the stability degree characterization value of the random load data g to be analyzed is: And is the stability degree characterization value of the random load data g to be analyzed, exp() is the exponential function with the constant e as the base, is the area of the minimum circumscribed circle of clustering cluster G, is the total number of random data points in clustering cluster G, is the variance of the nearest neighbor distance set, is the nearest neighbor distance of the random data point corresponding to the random load data g to be analyzed, C2 is a preset second constant, and the reason for adding the preset second constant to the denominator is to prevent the denominator from being 0.

[0040] In addition, when the area of the minimum circumscribed circle of the clustering cluster G is smaller and the total number of random data points in the clustering cluster G is larger, it indicates that the aggregation of random data points in the clustering cluster G is better. And when the aggregation of random data points in the clustering cluster G is better, it indicates that the stability degree of the random load data g to be analyzed is higher; when the variance of the nearest neighbor distance set is smaller and the nearest neighbor distance of the random data point corresponding to the random load data g to be analyzed is smaller, it indicates that on the basis that the random data points in the clustering cluster G are more evenly distributed, the distribution between the random load data g and its neighborhood data points is closer. And when on the basis that the random data points in the clustering cluster G are more evenly distributed, if the distribution between the random load data g and its neighborhood data points is closer, it more indicates that the stability degree of the random load data g to be analyzed is higher; thus, it can be known that when is smaller, is larger, and is smaller, it indicates that the stability degree of the random load data g to be analyzed is higher, otherwise it indicates that the stability degree of the random load data g to be analyzed is lower.

[0041] Therefore, through the above process, this embodiment can obtain the stability degree characterization value of each random load data to be analyzed.

[0042] Step S005, according to the stability degree characterization value of the random load data to be analyzed, adjust the initial coefficient of the load data to be analyzed to obtain the target coefficient of the random load data to be analyzed, and use the initial coefficient of the non-random load data to be analyzed as the target coefficient of the corresponding non-random load data to be analyzed.

[0043] After obtaining the stability degree characterization value of the random load data to be analyzed, this embodiment adjusts the initial coefficient of the load data to be analyzed according to the stability degree characterization value of the random load data to be analyzed, so as to obtain the target coefficient of the random load data to be analyzed. And the specific obtaining process of the target coefficient of the random load data to be analyzed is as follows: For the random load data g to be analyzed: First, multiply the random load data g by a preset adjustment control factor and then add the value of a preset second constant, and record it as the adjustment degree characterization value of the random load data g, that is, the adjustment degree characterization value of the random load data g is , is a preset adjustment control factor; then, the product of the adjustment degree characterization value and the initial coefficient of the random load data g to be analyzed is denoted as the target coefficient of the random load data g to be analyzed. Based on step S002, the initial coefficient of each random load data to be analyzed includes the initial autoregressive coefficient and the initial moving average coefficient. Therefore, the target coefficient of the load data to be analyzed obtained should also include the target autoregressive coefficient and the target moving average coefficient. That is, the target autoregressive coefficient of the random load data g to be analyzed is the product of the adjustment degree characterization value of the random load data g to be analyzed and the initial autoregressive coefficient of the random load data g to be analyzed, and the target moving average coefficient of the random load data g to be analyzed is the product of the adjustment degree characterization value of the random load data g to be analyzed and the initial moving average coefficient of the random load data g to be analyzed.

[0044] In specific applications, the implementer needs to determine the preset adjustment control factor according to the adjustment rate or the magnitude of the adjustment degree. For example, in this embodiment, if the adjustment degree is to be made smaller, then the preset adjustment control factor in this embodiment can be set to 0.2.

[0045] In addition, it should be noted that in this embodiment, the initial coefficient of the non-random load data to be analyzed is used as the target coefficient of the corresponding non-random load data to be analyzed.

[0046] Therefore, through the above process, this embodiment can obtain the target coefficients of each load data to be analyzed in the load data sequence to be analyzed.

[0047] Step S006: According to the ARIMA model and the target coefficients of the load data to be analyzed, obtain the predicted load data at the current monitoring moment, and regulate the load of the medium-voltage AC / DC hybrid power system according to the predicted load data.

[0048] Immediately afterwards, this embodiment obtains the predicted load data at the current monitoring moment according to the ARIMA model and the target coefficients of the load data to be analyzed. When using the ARIMA model to obtain the predicted load data at the current monitoring moment, only the initial coefficient of the load data to be analyzed is replaced with the target coefficient of the load data to be analyzed, and other steps or processes remain unchanged.

[0049] After obtaining the predicted load data at the current monitoring moment, the load data of the demand side of the medium-voltage AC / DC hybrid power system at the current monitoring moment collected by the data acquisition module is obtained and recorded as the current load data at the current monitoring moment; then, according to the current load data and the predicted load data at the current monitoring moment, the load of the medium-voltage AC / DC hybrid power system is regulated at the current monitoring moment, and on the basis of knowing the current load data and the predicted load data at the current monitoring moment, this embodiment also uses the traditional load regulation method to complete the load regulation process of the medium-voltage AC / DC hybrid power system at the current monitoring moment.

[0050] In addition, as another implementation manner, other methods can also be used to determine the method for regulating the load of the medium-voltage AC / DC hybrid power system. For example, first, obtain the absolute value of the difference between the current load data and the predicted load data, and record it as the target determination index value corresponding to the current monitoring moment. Then, determine whether the determination index value is greater than the preset determination index value. If so, use the AC mode to regulate the load of the medium-voltage AC / DC hybrid power system; otherwise, use the DC mode to regulate the load of the medium-voltage AC / DC hybrid power system; and the step-up / step-down performance of the regulation in the AC mode is better, and the power transmission performance of the regulation in the DC mode is better.

[0051] In addition, in specific applications, the implementer also needs to set the preset determination index value according to the actual situation. For example, this embodiment can also use the confidence interval method to determine the preset determination index value, that is, the preset determination index value can be determined with a 95% confidence level. The value corresponding to the 95% confidence level is 1.96, and the specific determination process is as follows: First, according to the method of obtaining the predicted load data at the current monitoring moment described above, obtain the predicted load data at the monitoring moment corresponding to each analyzed load data in the analyzed load data sequence, and record it as the predicted load data corresponding to the corresponding analyzed load data. Then, obtain the difference sequence, and the v-th difference in the difference sequence is the absolute value of the difference between the v-th analyzed load data in the analyzed load data sequence and the predicted load data corresponding to the v-th analyzed load data. Then, multiply the standard deviation of the difference sequence by 1.96 and add the mean value of the difference sequence, and the obtained result is used as the preset determination index value.

[0052] So far, this embodiment has completed the load regulation of the medium-voltage AC / DC hybrid power system.

[0053] In summary, in this embodiment, the load data sequence to be analyzed is first obtained; then, according to the ARIMA model, the coefficients of each load data to be analyzed in the load data sequence to be analyzed are obtained and recorded as the initial coefficients; then, according to the adjacent load data to be analyzed of each load data to be analyzed in the load data sequence to be analyzed, the random load data to be analyzed and the non-random load data to be analyzed in the load data sequence to be analyzed are obtained; then, the random historical load data, the random load data to be analyzed, and the corresponding random data points of the random historical load data are obtained, and the clustering algorithm is used to cluster the random data points to obtain each clustering cluster. According to the minimum circumscribed circle of the clustering cluster and the nearest neighbor random data points of the random data points in the clustering cluster, the stability characterization value of the random load data to be analyzed is obtained; then, according to the stability characterization value of the random load data to be analyzed, the initial coefficients of the load data to be analyzed are adjusted to obtain the target coefficients of the random load data to be analyzed, and the initial coefficients of the non-random load data to be analyzed are used as the target coefficients of the corresponding non-random load data to be analyzed; finally, according to the ARIMA model and the target coefficients of the load data to be analyzed, the predicted load data at the current monitoring moment is obtained, and the load of the medium-voltage AC / DC hybrid project power system is regulated according to the predicted load data. In this embodiment, the target coefficients obtained by adjusting the initial coefficients can make the prediction results more accurate and reliable, and based on the relatively accurate and reliable prediction results, the reliability and accuracy of regulating the load of the medium-voltage AC / DC hybrid project power system can be improved.

[0054] The above-described embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. An intelligent load control system for a medium voltage AC / DC hybrid project, comprising a processor and a memory, characterized in that: The processor executes the computer program stored in the memory to implement the following steps: Acquire a load data sequence to be analyzed, wherein the load data to be analyzed in the load data sequence to be analyzed is load data on the demand side of a medium voltage AC / DC hybrid power system; According to the ARIMA model, the coefficients of each load data to be analyzed in the load data sequence to be analyzed are obtained and recorded as initial coefficients; Obtain random load data to be analyzed and non-random load data to be analyzed in the load data sequence to be analyzed according to adjacent load data to be analyzed of each load data to be analyzed in the load data sequence to be analyzed; Obtain random historical load data and random data points corresponding to the random load data to be analyzed and the random historical load data, and cluster the random data points using a clustering algorithm to obtain various cluster clusters, and obtain a stability characterization value of the random load data to be analyzed according to the minimum circumscribed circle of the cluster cluster and the nearest neighbor random data points of the random data points in the cluster cluster; According to the stability characterization value of the random load data to be analyzed, the initial coefficient of the load data to be analyzed is adjusted to obtain the target coefficient of the random load data to be analyzed, and the initial coefficient of the non-random load data to be analyzed is used as the target coefficient of the corresponding non-random load data to be analyzed; According to the ARIMA model and the target coefficient of the load data to be analyzed, the predicted load data at the current monitoring time is obtained, and the load of the medium-voltage AC / DC hybrid power system is regulated according to the predicted load data.

2. The intelligent load control system for medium voltage AC / DC hybrid engineering according to claim 1, characterized in that: The method for acquiring random load data to be analyzed and non-random load data to be analyzed in the load data sequence to be analyzed comprises: Obtaining an outlier characterization value of each load data to be analyzed according to adjacent load data to be analyzed of each load data to be analyzed in the load data sequence to be analyzed; For any load data to be analyzed in the load data sequence to be analyzed, determine whether the outlier characterization value of the load data to be analyzed is not less than a preset outlier threshold; if so, record the load data to be analyzed as random load data to be analyzed; otherwise, record the load data to be analyzed as non-random load data to be analyzed.

3. The intelligent load control system for a medium voltage AC / DC hybrid project according to claim 2, characterized in that: The method for obtaining the outlier characterization value of each load data to be analyzed includes: Performing STL time series decomposition on the load data sequence to be analyzed to obtain the residual of each load data to be analyzed in the load data sequence to be analyzed; For the ath load data to be analyzed in the load data sequence to be analyzed: the normalized value of the absolute value of the difference between the residual of the ath load data to be analyzed and the preset residual is recorded as the first index value; the set constructed by all the load data to be analyzed except the ath load data to be analyzed in the load data sequence to be analyzed is recorded as the residual set, the mean of the residuals of all the load data to be analyzed in the residual set is recorded as the residual residual mean, and the absolute value of the difference between the residual of the ath load data to be analyzed and the residual residual mean is recorded as the characteristic difference value; the adjacent difference value of the ath load data to be analyzed is obtained; the normalized value of the result obtained by multiplying the adjacent difference value and the characteristic difference value is recorded as the second index value; A normalized value of a result obtained by adding the first index value and the second index value is recorded as the outlier characterization value of the ath load data to be analyzed.

4. The intelligent load control system for medium voltage AC / DC hybrid engineering according to claim 3, characterized in that: The method for obtaining adjacent difference values ​​of the ath load data to be analyzed includes: If the value of a is 1, the right difference value of the a-th load data to be analyzed is recorded as the adjacent difference value of the a-th load data to be analyzed; if the value of a is A, the left difference value of the a-th load data to be analyzed is recorded as the adjacent difference value of the a-th load data to be analyzed; if the value of a is not 1 and A, the sum of the right difference value and the left difference value is recorded as the adjacent difference value of the a-th load data to be analyzed; the right difference value is the absolute value of the difference between the residual of the a-th load data to be analyzed and the residual value of the a+1-th load data to be analyzed in the load data sequence to be analyzed; the left difference value is the absolute value of the difference between the residual of the a-th load data to be analyzed and the residual value of the a-1-th load data to be analyzed in the load data sequence to be analyzed; and A is the total number of data in the load data sequence to be analyzed.

5. The intelligent load control system for medium voltage AC / DC hybrid engineering according to claim 1, characterized in that: The method for acquiring random historical load data includes: Recording the acquisition time interval of the load data sequence to be analyzed as the current monitoring time interval; obtaining a preset number of historical time intervals that are the same as the current monitoring time interval and closest to the current monitoring time interval in the historical operation time period of the medium-voltage AC / DC hybrid power system, and recording them as historical monitoring time intervals; Obtaining a load data sequence corresponding to the historical monitoring time interval, and recording it as a historical load data sequence corresponding to the historical monitoring time interval; According to the method for obtaining the random load data to be analyzed in the load data sequence to be analyzed, the random historical load data in the historical load data sequence is obtained.

6. The intelligent load control system for medium voltage AC / DC hybrid engineering according to claim 1, characterized in that: The method for obtaining random data points includes: Constructing a two-dimensional space, wherein the horizontal axis of the two-dimensional space is the time mark value and the vertical axis is the load data; All of the random load data to be analyzed, all of the random historical load data, the time stamp values ​​of the random load data to be analyzed, and the time stamp values ​​of the random historical load data are mapped into the two-dimensional space to obtain data points corresponding to the random load data to be analyzed and data points corresponding to the random historical load data, and are all recorded as random data points. The time stamp value of the fth historical load data in the historical load data sequence and the time stamp value of the fth load data to be analyzed in the load data sequence to be analyzed are both f.

7. The intelligent load control system for medium voltage AC / DC hybrid engineering according to claim 1, characterized in that: The method for obtaining the stability characterization value of the random load data to be analyzed includes: For any random load data to be analyzed g in the load data sequence to be analyzed: The cluster containing the random data points corresponding to the random load data g to be analyzed is recorded as cluster G; Recording the normalized value of the total number of random data points in the cluster G as the quantity characterization value; performing negative correlation mapping on the minimum circumscribed circle area of ​​the cluster G, and recording the mapping result as the area characterization value; recording the sum of the quantity characterization value and the quantity characterization value as the first distribution characteristic value of the random load data to be analyzed g; Obtain the nearest neighbor distance of each random data point in the cluster G, and record the set of nearest neighbor distances of all random data points in the cluster G as the nearest neighbor distance set; the nearest neighbor distance of any random data point d in the cluster G refers to the Euclidean distance between the random data point d and the nearest neighbor random data point of the random data point d, and the nearest neighbor random data point of the random data point d belongs to the cluster G; The product of the variance of the nearest neighbor distance set and the nearest neighbor distance of the random data point corresponding to the random load data to be analyzed g is recorded as a first product value; the reciprocal of the result obtained by adding the first product value and a preset second constant is used as the second distribution characteristic value of the random load data to be analyzed g; The product of the first distribution characteristic value and the second distribution characteristic value is recorded as the stability representation value of the random load data g to be analyzed.

8. The intelligent load control system for medium voltage AC / DC hybrid engineering according to claim 1, characterized in that: The method of adjusting the initial coefficient of the load data to be analyzed according to the stability characterization value of the random load data to be analyzed to obtain the target coefficient of the random load data to be analyzed includes: For any random load data to be analyzed g in the load data sequence to be analyzed: The value obtained by multiplying the random load data g to be analyzed by the preset adjustment control factor and then adding the value obtained by adding the preset second constant is recorded as the adjustment degree representation value of the random load data g to be analyzed; the product of the adjustment degree representation value and the initial coefficient of the random load data g to be analyzed is recorded as the target coefficient of the random load data g to be analyzed.

Images