A virtual power plant load forecasting system based on big data

Abstract

The present invention discloses a virtual power plant load forecasting system based on big data, which relates to the technical field of power load forecasting. The virtual power plant load forecasting system based on big data includes: a time series evaluation module, a numerical characteristic evaluation module, a data feature evaluation module and a multi-source data evaluation module. The present invention obtains a time series anomaly index through the acquired time series data and determines whether to take time adjustment measures, then obtains a numerical characteristic anomaly index through the acquired numerical characteristic data to determine whether to take data adjustment measures, then obtains a data feature deviation index through the acquired data feature data to determine whether to regulate the intensity of data adjustment measures, and finally obtains a multi-source data consistency evaluation index through the acquired multi-source feature data to determine whether to perform equipment operation and maintenance, thereby achieving the effect of obtaining more accurate data for power load forecasting and solving the problem of unstable data source when performing power load forecasting in the prior art.

Description

Technical Field

[0001] The present invention relates to the technical field of power load forecasting, and in particular to a virtual power plant load forecasting system based on big data. Background Art

[0002] With the transformation of the energy structure and the continuous increase in renewable energy penetration, the operational complexity of the power system has increased significantly. Virtual power plants (VPPs) are gaining increasing attention as an innovative model for integrating distributed energy resources. Leveraging advanced digital and information technologies, VPPs aggregate and optimize the scheduling of distributed energy resources, electric vehicles, and load-side resources, improving resource utilization efficiency and enhancing grid stability. Against this backdrop, the rapid development of big data technology has provided strong support for VPPs. Through the real-time collection, processing, and analysis of massive amounts of data, accurate load forecasting and dynamic control can be achieved. The big data-based VPP load forecasting system, combined with machine learning, the Internet of Things, and cloud computing, can effectively predict user load characteristics, optimize scheduling strategies, and promote the intelligent and low-carbon transformation of the energy system.

[0003] Existing virtual power plant load forecasting technologies primarily rely on historical load data and simple time series analysis models. However, these traditional methods lack accuracy and adaptability in complex scenarios such as the large-scale integration of distributed energy resources and increased grid load volatility. Some research is beginning to introduce big data-based analysis methods, incorporating massive multi-source data such as meteorological information, user electricity usage behavior, and electricity market prices into forecasting models to enhance forecasting capabilities. Simultaneously, the application of artificial intelligence technologies, particularly deep learning and machine learning algorithms, has significantly enhanced the nonlinear modeling capabilities of load forecasting. For example, methods such as long-short-term memory networks and support vector machines can effectively capture the temporal characteristics of load and external influencing factors. However, existing technologies still have limitations in data quality management, real-time performance, and cross-platform compatibility, making them unable to fully meet the needs of complex and dynamic power grids.

[0004] For example, the invention patent announcement with announcement number: CN114037179B discloses a power load forecasting system and method based on big data, including: analyzing deviations through a training unit, training the data model, optimizing the model and algorithm, the data model outputs analysis data closest to the real data to the correlation value analysis module, the correlation value analysis module retrieves the real load data from the real load database, and then compares the analysis data, analyzes the correlation of the big data, and obtains the data with the highest correlation, the information integration module obtains the analysis value of the correlation of the correlation value analysis module, and optimizes the information correlation integration.

[0005] For example, the invention patent announcement with announcement number: CN104881706B discloses a method for short-term load forecasting of power systems based on big data technology, including: cluster analysis of load curves, classifying load curves with similar shape characteristics into one category; establishing key influencing factors to simplify classification rules and simplify the prediction model; establishing classification rules and using the CART (Chimeric Antigen Receptor T-Cell Therapy) decision tree algorithm to obtain agglomerative hierarchical clustering analysis results; classifying the days to be predicted; training the prediction model and predicting, and selecting the corresponding support vector machine model to complete the prediction based on the classification results of the days to be predicted; and calculating the system load. This step is completed on the Hadoop big data computing platform.

[0006] However, in the process of implementing the technical solutions of the invention in the embodiments of the present application, the present application found that the above technology has at least the following technical problems:

[0007] In the existing technology, since the load forecasting of virtual power plants relies on multi-source data, and the power load data used for load forecasting may be affected by multiple factors during the collection process, there is a problem of unstable data source when performing power load forecasting. Summary of the Invention

[0008] The embodiment of the present application solves the problem of unstable data source in power load forecasting in the prior art by providing a virtual power plant load forecasting system based on big data, and achieves the goal of obtaining more accurate data for power load forecasting.

[0009] The embodiment of the present application provides a virtual power plant load forecasting system based on big data, including: a time series evaluation module, a numerical characteristic evaluation module, a data feature evaluation module and a multi-source data evaluation module: wherein the time series evaluation module is used to obtain a time series anomaly index through the acquired time series data, and judge whether to take time adjustment measures based on the time series anomaly index, and the time series anomaly index is used to quantify the probability of abnormalities in the time aspect of the power load data during the acquisition process; the numerical characteristic evaluation module is used to obtain a numerical characteristic anomaly index through the acquired numerical characteristic data, and judge whether to take data adjustment measures based on the numerical characteristic anomaly index, and the numerical characteristic anomaly index is used to quantify The probability of numerical anomalies in the power load data during the collection process; the data feature evaluation module is used to obtain the data feature deviation index through the acquired data feature data, and judge whether to regulate the intensity of data adjustment measures based on the data feature deviation index. The data feature deviation index is used to quantify the probability of deviation in the feature data during the collection process of the power load data; the multi-source data evaluation module is used to obtain the multi-source data consistency evaluation index through the multi-source feature data acquired within a preset time period, and judge whether to perform equipment operation and maintenance based on the multi-source data consistency evaluation index. The multi-source data consistency evaluation index is used to quantify the probability of conflicts between multiple data sources during the collection process of the power load data.

[0010] Furthermore, the time series data includes collection frequency, collection data volume, data missing rate, data delay duration and timestamp deviation; the numerical characteristic data includes short-term fluctuation amplitude, saturation value duration, proportion of out-of-range abnormal points and invalid value ratio; the data characteristic data includes load stability index, fluctuation frequency, fluctuation amplitude, peak-valley deviation and mean deviation; the multi-source characteristic data includes data merging error, communication interruption rate and number of equipment warnings.

[0011] Furthermore, the specific steps for obtaining the time series anomaly index are as follows: number the preset time periods, and obtain reference time series data from a preset database, wherein the reference time series data includes the maximum value of the data delay duration and the maximum value of the timestamp deviation; obtain the corresponding preset duration according to the preset time period; if the time series data meets the first judgment condition, process the time series data and the corresponding reference time series data to obtain the time series anomaly index, wherein the first judgment condition indicates that the data delay duration and the timestamp deviation are both less than the corresponding maximum value of the data delay duration and the maximum value of the timestamp deviation.

[0012] Furthermore, the specific steps for obtaining the numerical characteristic anomaly index are: obtaining reference numerical characteristic data from a preset database, the reference numerical characteristic data including the maximum short-time fluctuation amplitude and the maximum saturation value duration; comparing the numerical characteristic data with the reference numerical characteristic data: if the numerical characteristic data does not meet the second judgment condition, directly taking data adjustment measures, the second judgment condition indicating that the short-time fluctuation amplitude and the saturation value duration are both less than the corresponding maximum short-time fluctuation amplitude and maximum saturation value duration; if the numerical characteristic data meets the second judgment condition, performing hyperbolic cotangent processing on the ratio of the short-time fluctuation amplitude to the maximum short-time fluctuation amplitude and the ratio of the saturation value duration to the maximum saturation value duration to obtain a first characteristic index, and performing inverse cotangent processing on the sum of the proportion of out-of-range abnormal points and the proportion of invalid values to obtain a second characteristic index; obtaining the numerical characteristic anomaly index based on the first characteristic index and the second characteristic index.

[0013] Furthermore, the specific acquisition process of the data feature deviation index is as follows: obtain reference feature data and deviation assessment weights from a preset database, the reference feature data including the maximum stability index, the maximum fluctuation amplitude, the maximum fluctuation frequency, the maximum peak-to-valley deviation and the maximum mean deviation, and the deviation assessment weights including the load weight, the fluctuation weight and the deviation weight; obtain deviation indicators based on the data feature data and the corresponding reference feature data, the deviation indicators including the first deviation indicator, the second deviation indicator, the third deviation indicator, the fourth deviation indicator and the fifth deviation indicator; process the deviation indicators and the deviation assessment weights to obtain the data feature deviation index.

[0014] Furthermore, the specific acquisition process of the multi-source data consistency evaluation index is as follows: obtain reference consistency data from a preset database, the reference consistency data including the maximum merge error, the maximum communication interruption rate, and the maximum number of device warnings; obtain the multi-source data consistency evaluation index based on the multi-source feature data and the reference consistency data, and the specific restricted expression of the multi-source data consistency evaluation index is as follows:

[0015]

[0016] Where n represents the number of the preset time period, , Indicates the number of the preset time period, Indicates the data merging error of the nth preset time period, Indicates the communication interruption rate of the nth preset time period, Indicates the number of device warnings in the nth preset time period, represents the maximum value of the combined error, Indicates the maximum value of communication interruption rate, Indicates the maximum number of device warnings. Indicates the multi-source data consistency evaluation index for the nth preset time period.

[0017] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0018] 1. The time series anomaly index is obtained by the acquired time series data to determine whether time adjustment measures should be taken. Then, the numerical characteristic anomaly index is obtained by the acquired numerical characteristic data to determine whether data adjustment measures should be taken. Then, the data characteristic deviation index is obtained by the acquired data characteristic data to determine whether the intensity of data adjustment measures should be adjusted. Finally, the multi-source data consistency assessment index is obtained by the acquired multi-source characteristic data to determine whether equipment operation and maintenance should be performed, thereby making the power load data more timely, and thus achieving more accurate data for power load forecasting, effectively solving the problem of unstable data source when conducting power load forecasting in the existing technology.

[0019] 2. By numbering the preset time periods and obtaining reference time series data from a preset database, and then obtaining the corresponding preset duration according to the preset time period, if the time series data meets the first judgment condition, the time series data and the corresponding reference time series data are processed to obtain a time series anomaly index, thereby timely detecting problems in the data collection time series, thereby improving the stability of the data source related to the collected power load data.

[0020] 3. By obtaining reference feature data and deviation assessment weights from a preset database, and then obtaining a deviation index based on the data feature data and the corresponding reference feature data, and then processing the deviation index and deviation assessment weights to obtain a data feature deviation index, the degree of deviation of the collected data can be more accurately assessed, thereby achieving more accurate power load-related data. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A schematic diagram of the structure of a virtual power plant load forecasting system based on big data provided in an embodiment of the present application;

[0022] Figure 2 It is a schematic diagram of the change of numerical characteristic anomaly index with short-term fluctuation amplitude;

[0023] Figure 3 It is a schematic diagram of the change of the numerical characteristic anomaly index with the duration of saturation value;

[0024] Figure 4 This is a schematic diagram showing the change in the numerical characteristic anomaly index with the proportion of out-of-range anomaly points;

[0025] Figure 5Schematic diagram of the change of numerical characteristic anomaly index with the proportion of invalid values. DETAILED DESCRIPTION

[0026] The embodiment of the present application solves the problem of unstable data source in power load forecasting in the prior art by providing a virtual power plant load forecasting system based on big data. The system numbers the preset time periods and obtains reference time series data from a preset database. Then, the corresponding preset duration is obtained according to the preset time period. If the time series data meets the first judgment condition, the time series data and the corresponding reference time series data are processed to obtain a time series anomaly index and determine whether to take time adjustment measures. Then, the numerical characteristic anomaly index is obtained from the obtained numerical characteristic data to determine whether to take data adjustment measures. Then, reference feature data and deviation assessment weights are obtained from the preset database. Then, a deviation index is obtained based on the data feature data and the corresponding reference feature data. The deviation index and the deviation assessment weight are processed to obtain a data feature deviation index to determine whether to regulate the intensity of the data adjustment measures. Finally, a multi-source data consistency assessment index is obtained from the obtained multi-source feature data to determine whether to perform equipment operation and maintenance, thereby achieving more accurate data for power load forecasting.

[0027] The technical solution in the embodiment of the present application is to solve the problem of unstable data source when performing power load forecasting. The overall idea is as follows:

[0028] The obtained time series anomaly index is used to determine whether time adjustment measures should be taken, and then the obtained numerical characteristic anomaly index is used to determine whether data adjustment measures should be taken. Then, the obtained data feature deviation index is used to determine whether the intensity of data adjustment measures should be adjusted. Finally, the obtained multi-source data consistency assessment index is used to determine whether equipment operation and maintenance should be carried out, thereby achieving the effect of obtaining more accurate data for power load forecasting.

[0029] In order to better understand the above technical solution, the above technical solution will be described in detail below with reference to the accompanying drawings and specific implementation methods.

[0030] like Figure 1As shown, it is a structural diagram of a virtual power plant load forecasting system based on big data provided by an embodiment of the present application. The virtual power plant load forecasting system based on big data provided by an embodiment of the present application includes: a time series evaluation module, a numerical characteristic evaluation module, a data feature evaluation module and a multi-source data evaluation module: wherein the time series evaluation module is used to obtain a time series anomaly index by acquiring time series data within a preset time period, and judge whether to take time adjustment measures based on the time series anomaly index, and the time series anomaly index is used to quantify the probability of abnormalities in time of power load data during the collection process; the numerical characteristic evaluation module is used to obtain a numerical characteristic anomaly index by acquiring numerical characteristic data within a preset time period, and judge whether to take time adjustment measures based on the numerical characteristic anomaly index. Whether to take data adjustment measures, the numerical characteristic anomaly index is used to quantify the probability of numerical anomalies in the power load data during the collection process; the data feature evaluation module is used to obtain the data feature deviation index by obtaining the data feature data within a preset time period, and judge whether to adjust the intensity of data adjustment measures based on the data feature deviation index. The data feature deviation index is used to quantify the probability of feature data deviation during the collection process of power load data; the multi-source data evaluation module is used to obtain the multi-source data consistency evaluation index by obtaining the multi-source feature data within a preset time period, and judge whether to perform equipment operation and maintenance based on the multi-source data consistency evaluation index. The multi-source data consistency evaluation index is used to quantify the probability of conflicts between multiple data sources during the collection process of power load data.

[0031] In this embodiment, the big data-based virtual power plant load forecasting system not only improves the accuracy of load forecasting, but also enhances the robustness and stability of the system, improves data quality and utilization efficiency, and achieves the effect of obtaining more accurate data for power load forecasting through the comprehensive application of multiple evaluation modules; at the same time, it also adopts advanced, mature, reliable, and mainstream general technologies with successful application cases, with high safety, economy and operability, ensuring the safety of personnel and equipment, and facilitating operation, observation, monitoring and maintenance, with high availability; among them, big data not only provides a sufficient data foundation, but also improves the accuracy, real-time and reliability of load forecasting through efficient data processing and analysis capabilities, while providing strong support for the scheduling optimization and flexible operation of virtual power plants.

[0032] It should be added that time series data includes collection frequency, collection data volume, data missing rate, data delay duration and timestamp deviation; numerical characteristic data includes short-term fluctuation amplitude, saturation value duration, proportion of out-of-range abnormal points and proportion of invalid values; data characteristic data includes load stability index, fluctuation frequency, fluctuation amplitude, peak-valley deviation and mean deviation; multi-source characteristic data includes data merging error, communication interruption rate and number of equipment warnings.

[0033] Among them, the collection frequency is directly obtained through the collection program of the data collection equipment (such as smart meters, power detection instruments, distributed energy management system equipment and distribution automation terminal equipment, etc.), the collected data volume is directly obtained through the system log, and the data missing rate represents the ratio of the amount of missing data to the preset amount of collected data. The data verification mechanism (such as serial number) is used to find missing data. The data delay duration is directly obtained through the network monitoring tool. The timestamp is directly obtained through the deviation time synchronization protocol (such as NTP, Network Time Protocol, PTP, Precision Time Protocol).

[0034] The short-term fluctuation amplitude indicates the ratio of the difference between the maximum and minimum values of the data collected within the preset time period to the average value of the data. The saturation value duration indicates the duration for which the collected data reaches the maximum or minimum value of the reference load range. The reference load range is set according to the data collection requirements. The proportion of out-of-range anomalies indicates the ratio of the number of data that are not within the reference load range to the amount of collected data. The invalid value ratio indicates the ratio of the amount of missing data to the amount of collected data.

[0035] The load stability index is the ratio of the difference between the peak and valley values of the data collected within a preset time period to the data average. The fluctuation frequency represents the ratio of the number of times the set fluctuation amplitude is exceeded to the preset duration. The set fluctuation amplitude is specified by the data collection requirements. The fluctuation amplitude represents the ratio of the difference between the maximum and minimum values in the currently collected data to the data average. The peak-to-valley deviation represents the absolute value of the difference between the peak and valley values of the data collected within the preset time period and the peak and valley values of the corresponding data in the historical data. The mean deviation represents the deviation between the average value of the data collected within the preset time period and the mean of the corresponding historical data. The data merging error represents the absolute deviation between the data automatically merged by the system from multiple data sources (such as substations, distribution rooms, distribution network automation systems, and power equipment monitoring terminals) and the historical mean of the corresponding data. The communication interruption rate represents the ratio of the number of communication failures to the total number of communications. The number of equipment warnings is directly obtained from the operation log of the data collection equipment. The collection and acquisition of the above data facilitates the subsequent analysis of the accuracy of power load data acquisition and provides a data foundation for subsequent analysis.

[0036] Furthermore, the specific steps for obtaining the time series anomaly index are as follows: number the preset time periods, and obtain reference time series data from a preset database, the reference time series data including the maximum value of the data delay duration and the maximum value of the timestamp deviation; obtain the corresponding preset duration according to the preset time period, and judge the time series data and the reference time series data: if the time series data meets the first judgment condition, the time series data and the corresponding reference time series data are processed to obtain the time series anomaly index, and the first judgment condition indicates that the data delay duration and the timestamp deviation are both less than the corresponding maximum value of the data delay duration and the maximum value of the timestamp deviation.

[0037] The specific restriction expression of the time series anomaly index is as follows:

[0038] ;

[0039] Where n represents the number of the preset time period, , Indicates the number of the preset time period, Indicates the acquisition frequency of the nth preset time period, Indicates the amount of collected data in the nth preset time period, Indicates the data missing rate of the nth preset time period, Indicates the data delay duration of the nth preset time period, Indicates the timestamp deviation of the nth preset time period, Indicates the maximum data delay time. Indicates the maximum value of timestamp deviation, Indicates the preset duration. Indicates the time series anomaly index for the nth preset time period.

[0040] In this embodiment, the algorithm combines the time series data and the corresponding reference time series data for comprehensive analysis to obtain the time series anomaly index, where the product of the acquisition frequency and the preset duration should be the ratio of the amount of collected data (i.e. ) is analyzed as a whole. When the ratio of the product of the acquisition frequency and the preset time length to the amount of collected data approaches 1, the time series anomaly index is smaller, indicating that the collected data is more comprehensive and the possibility of time series anomaly is smaller. Similarly, when the time series data meets the first judgment condition, as the data delay time is less than the maximum data delay time, the timestamp deviation is less than the maximum timestamp deviation, and the data missing rate is smaller, the time series anomaly index is smaller, indicating that the collected data is less missing and the possibility of time series anomaly is smaller. Among them, the acquisition frequency determines the amount of data collected per unit time. A higher acquisition frequency may lead to an increase in the amount of collected data, thereby increasing the data missing rate, increasing the data delay time, and expanding the timestamp deviation. In addition, the data delay time is affected by the amount of collected data and the acquisition frequency. The longer the data delay time, the larger the timestamp deviation, and even affects the continuity and real-time performance of the data. Therefore, the analysis of time series data cannot be ignored. Through the analysis of time series data, it is helpful to timely observe the time problems of data collection, thereby improving the efficiency of data collection and improving the accuracy of data acquisition.

[0041] Specifically, the reference time series data is obtained from a preset database. In one specific embodiment, the maximum data delay duration and the maximum timestamp deviation are set by professionals related to the power load system. For example, in an industrial environment, it can be assumed that the maximum data delay duration is 10ms and the maximum timestamp deviation is 1ms.

[0042] Furthermore, the specific process of determining whether to take time adjustment measures is as follows: if the time series data does not meet the first judgment condition, then directly take time adjustment measures; otherwise, obtain the time threshold from the preset database, and the time threshold is used to determine whether to take time adjustment measures; compare the time series anomaly index with the time threshold: if the time series anomaly index is not less than the time threshold, then take time adjustment measures in the next preset time period, and the time adjustment measures include increasing bandwidth and using time synchronization protocol, and the time synchronization protocol is used to reduce timestamp differences; if the time series anomaly index is less than the time threshold, then execute the function of the numerical characteristic evaluation module.

[0043] In this embodiment, the system can quickly respond to anomalies in time series data, dynamically adjust, and improve data quality and system performance. It also reduces resource waste through hierarchical processing. This approach not only ensures the accuracy of time series data but also supports efficient system operation, making it suitable for a variety of high-precision, low-latency application scenarios.

[0044] Specifically, the time threshold is obtained from a preset database. In one specific embodiment, the time series data corresponding to the time errors in the power load data collection in the historical data is substituted into the specific restriction expression of the time series anomaly index to obtain the corresponding data set, and the result of the mean operation on the data set is recorded as the time threshold.

[0045] Furthermore, the specific steps for obtaining the numerical characteristic anomaly index are as follows: obtaining reference numerical characteristic data from a preset database, the reference numerical characteristic data including the maximum short-time fluctuation amplitude and the maximum saturation value duration; comparing the numerical characteristic data with the reference numerical characteristic data: if the numerical characteristic data does not meet the second judgment condition, directly taking data adjustment measures, the second judgment condition indicating that the short-time fluctuation amplitude and the saturation value duration are both less than the corresponding maximum short-time fluctuation amplitude and maximum saturation value duration; if the numerical characteristic data meets the second judgment condition, performing hyperbolic cotangent processing on the ratio of the short-time fluctuation amplitude to the maximum short-time fluctuation amplitude and the ratio of the saturation value duration to the maximum saturation value duration to obtain a first characteristic index, and performing inverse cotangent processing on the sum of the proportion of out-of-range abnormal points and the proportion of invalid values to obtain a second characteristic index; obtaining the numerical characteristic anomaly index based on the first characteristic index and the second characteristic index, the specific restriction expression of the numerical characteristic anomaly index is as follows:

[0046] ;

[0047] Where n represents the number of the preset time period, , Indicates the number of the preset time period, Indicates the short-term fluctuation amplitude of the nth preset time period, Indicates the duration of the saturation value of the nth preset time period, Indicates the proportion of abnormal points exceeding the range in the nth preset time period, Indicates the ratio of invalid values in the nth preset time period, Indicates the maximum fluctuation amplitude, Indicates the maximum duration of saturation value. It represents the numerical characteristic abnormality index of the nth preset time period, and e represents a natural constant.

[0048] In this embodiment, the algorithm combines the short-term fluctuation amplitude, saturation value duration, out-of-range anomaly point ratio and invalid value ratio with the corresponding reference numerical characteristic data for comprehensive analysis to obtain the numerical characteristic anomaly index. In the formula, as the short-term fluctuation amplitude, saturation value duration, out-of-range anomaly point ratio and invalid value ratio increase, the numerical characteristic anomaly index also increases accordingly, and the numerical characteristic anomaly index is only needed to be obtained when the time series anomaly index is less than the time threshold.

[0049] like Figure 2-Figure 5 As shown, it is a schematic diagram of the change of the numerical characteristic abnormality index provided by the embodiment of the present application, and it is assumed that the maximum fluctuation amplitude is 1 and the saturation value duration is 1, wherein Figure 2 This is a diagram showing the change of the numerical characteristic anomaly index with the short-term fluctuation amplitude, assuming that the saturation value duration is 0.5, the proportion of out-of-range anomalies is 0.2, and the proportion of invalid values is 0.1. Figure 3 This is a diagram showing the change of the numerical characteristic anomaly index with the duration of the saturation value, and assuming that the short-term fluctuation amplitude is 0.5, the proportion of out-of-range anomalies is 0.2, and the proportion of invalid values is 0.1. Figure 4 This is a diagram showing the change in the numerical characteristic anomaly index with the proportion of out-of-range anomaly points, assuming that the saturation value duration is 0.5, the short-term fluctuation amplitude is 0.5, and the invalid value ratio is 0.1. Figure 5 This is a diagram showing the change of the numerical characteristic anomaly index with the proportion of invalid values. Assuming that the short-term fluctuation amplitude is 0.5, the saturation value duration is 0.5, and the proportion of out-of-range abnormal points is 0.2, then Figure 2-Figure 5 It can be seen that with the increase of short-term fluctuation amplitude, saturation value duration, proportion of out-of-range anomaly points and invalid value ratio, the image shows an upward trend, and the numerical characteristic anomaly index increases accordingly, indicating that the short-term fluctuation amplitude, saturation value duration, proportion of out-of-range anomaly points and invalid value ratio are positively correlated with the numerical characteristic anomaly index, indicating that when the values of short-term fluctuation amplitude, saturation value duration, proportion of out-of-range anomaly points and invalid value ratio are larger, the probability of numerical characteristic anomaly is greater, and the possibility of unstable data source is also greater.

[0050] In this algorithm, changes in the short-term fluctuation amplitude and the saturation value duration will directly affect the generation of out-of-range points, thereby affecting the proportion of out-of-range abnormal points. Abnormalities in the short-term fluctuation amplitude, saturation value duration, and the proportion of out-of-range abnormal points will lead to a decline in data quality, which will ultimately manifest as an increase in invalid values and an increase in the proportion of invalid values. By analyzing the numerical characteristic anomaly index, it is helpful to detect data collection errors in a timely manner according to the fluctuation changes of the collected power load, thereby promptly discovering whether the data source of the collected data is stable, and taking corresponding adjustment measures to obtain more accurate data.

[0051] Specifically, the reference numerical characteristic data is obtained from a preset database. In one specific embodiment, the maximum short-term fluctuation amplitude and the maximum saturation value duration are set by professionals related to the power load system. For example, in an industrial environment, the maximum short-term fluctuation amplitude can be assumed to be 5%, and the maximum saturation value duration can be assumed to be 10 seconds.

[0052] Furthermore, the specific process of determining whether to take data adjustment measures is as follows; if the numerical characteristic data does not meet the second judgment condition, then directly take data adjustment measures, otherwise obtain the numerical threshold from the preset database, and the numerical threshold is used to determine whether to take data adjustment measures; compare the numerical characteristic anomaly index with the numerical threshold: if the numerical characteristic anomaly index is not less than the numerical threshold, then take data adjustment measures in the next preset time period, and the data adjustment measures include adjusting the data collection frequency and anomaly marking, and the anomaly marking indicates marking the saturated value data and ignoring the marked saturated value in subsequent analysis; if the numerical characteristic anomaly index is less than the numerical threshold, then execute the function of the data feature evaluation module.

[0053] In this embodiment, the data collection frequency can be adjusted according to a preset adjustment ratio based on the current data collection frequency. For example, if the preset adjustment ratio set by the preset staff is 10% and the current data collection frequency is 60 times / minute, then the data collection frequency can be adjusted to 66 times / minute; the abnormal mark marks the data in the collected data that reaches the maximum or minimum value of the reference load range to indicate abnormality; through data adjustment measures and optimized processing procedures, data quality is improved, data collection and processing efficiency is optimized, system robustness is enhanced, and ultimately the accuracy of load forecasting is improved; and accurate data is the basis of load forecasting, and this process ensures the accuracy and reliability of data through data adjustment measures, thereby providing a solid foundation for load forecasting.

[0054] Specifically, the numerical threshold is obtained from a preset database. In one specific embodiment, the numerical characteristic data corresponding to the numerical errors in the power load data collection in the historical data is substituted into the specific limiting expression of the numerical characteristic anomaly index to obtain the corresponding data set, and the result of the mean operation on the data set is recorded as the numerical threshold.

[0055] Furthermore, the specific process of obtaining the data feature deviation index is as follows: obtaining reference feature data and deviation assessment weights from a preset database, the reference feature data including the maximum stability index, the maximum fluctuation amplitude, the maximum fluctuation frequency, the maximum peak-to-valley deviation, and the maximum mean deviation, and the deviation assessment weights including the load weight, the fluctuation weight, and the deviation weight; obtaining deviation indicators based on the data feature data and the corresponding reference feature data, the deviation indicators including the first deviation indicator, the second deviation indicator, the third deviation indicator, the fourth deviation indicator, and the fifth deviation indicator; processing the deviation indicators and the deviation assessment weights to obtain the data feature deviation index, the specific restricted expression of the data feature deviation index is as follows:

[0056] ;

[0057] ;

[0058] ;

[0059] ;

[0060] ;

[0061] ;

[0062] Where n represents the number of the preset time period, , Indicates the number of the preset time period, Indicates the load stability index of the nth preset time period, Indicates the fluctuation frequency of the nth preset time period, Indicates the fluctuation amplitude of the nth preset time period, Indicates the peak-to-valley deviation of the nth preset time period, Indicates the mean deviation of the nth preset time period, represents the maximum stability index, Indicates the maximum fluctuation range, represents the maximum fluctuation frequency, represents the maximum peak-to-valley deviation, represents the maximum mean deviation, represents the load weight, represents the volatility weight, represents the bias weight, represents the first deviation index of the nth preset time period, represents the second deviation index of the nth preset time period, represents the third deviation index of the nth preset time period, represents the fourth deviation index of the nth preset time period, represents the fifth deviation index of the nth preset time period, Indicates the data feature deviation index of the nth preset time period.

[0063] In this embodiment, the algorithm combines the deviation index and the deviation evaluation weight for comprehensive analysis to obtain the data feature deviation index, wherein, when the deviation evaluation weight is fixed, as the first deviation index, the second deviation index, the third deviation index, the fourth deviation index and the fifth deviation index increase (these data are not less than 0), the data feature deviation index decreases accordingly, indicating that the degree of data feature deviation is lower. When the load stability index is smaller than the maximum stability index, the smaller the fluctuation degree of the collected power load data is, the larger the corresponding first deviation index is. The second deviation index, the third deviation index, the fourth deviation index and the fifth deviation index are analyzed in the same way. When the fluctuation amplitude, fluctuation frequency, peak-to-valley deviation and mean deviation are smaller than the corresponding maximum fluctuation amplitude, maximum fluctuation frequency, peak-to-valley deviation and mean deviation, the fluctuation amplitude, fluctuation frequency, peak-to-valley deviation and mean deviation are smaller than the corresponding maximum fluctuation amplitude, maximum fluctuation frequency and peak-to-valley deviation. When the frequency, maximum peak-to-valley deviation and maximum mean deviation are large, the second deviation index, the third deviation index, the fourth deviation index and the fifth deviation index are larger, indicating that the fluctuation degree of the collected power load data is smaller; among them, when the fluctuation amplitude is smaller, the fluctuation frequency is lower and the peak-to-valley deviation is smaller, the load is usually more stable and the load stability index is higher. At the same time, the larger the mean deviation may mean the stronger the fluctuation, and the smaller the load stability index will be. The data feature deviation index is only needed to be obtained when the numerical characteristic anomaly index is less than the numerical threshold. The analysis of the data feature deviation index helps to understand whether the collected data range does not meet the data collection requirements, so as to take corresponding measures in time to reduce the risk of data errors and ensure the stability of power load data collection.

[0064] Specifically, the load weight is a weight value pre-set in the preset database based on the load stability index, which reflects the influence of the load stability index on the data feature deviation index. In practical applications, the weight corresponding to the load stability index can be directly found from the preset database. This correspondence is a pre-set mapping relationship. For example, when performing a data feature deviation assessment, a mapping set will be established. The mapping set contains the weights corresponding to the load stability index and the preset data feature deviation index. By inputting the real-time load stability index into this mapping set, the corresponding weight can be found. The mapping relationship here is that each load stability index uniquely corresponds to a weight value, and the weight value range is limited to between 0 and 1.

[0065] Specifically, the fluctuation weight is the weight value set for the fluctuation-related data (specifically including the fluctuation amplitude and fluctuation frequency) in the preset database. These weight values reflect the degree of influence of the fluctuation-related data on the data feature deviation index. In actual operation, the weight corresponding to the fluctuation-related data can be directly extracted from the preset database. This correspondence is achieved through a pre-set mapping relationship. For example, during the process of evaluating data feature deviation, a mapping set is constructed, which associates the fluctuation-related data with the weight corresponding to the data feature deviation index preset in the preset database. When real-time fluctuation-related data is input into this mapping set, the corresponding weight value can be retrieved. It should be emphasized that the mapping relationship here is a one-to-one correspondence, that is, each fluctuation-related data uniquely corresponds to a weight value. In this example, the value range of the fluctuation weight is limited to between 0 and 1.

[0066] Specifically, in this example, the sum of the deviation weight, the load weight, and the fluctuation weight is 1.

[0067] Specifically, the reference characteristic data is obtained from a preset database. In one specific embodiment, the reference consistency data is set by professionals related to the power load system based on specific industrial conditions. For example, generally, it can be assumed that the maximum stability index is 0.6, the maximum fluctuation amplitude is 20%, the maximum fluctuation frequency is 10 Hz, the maximum peak-to-valley deviation is 50%, and the maximum mean deviation is 30%.

[0068] Furthermore, the specific process for determining whether to regulate the intensity of data adjustment measures is as follows: obtaining a deviation threshold from a preset database, and the deviation threshold is used to determine whether to regulate the intensity of data adjustment measures; comparing the data feature deviation index with the deviation threshold: if the data feature deviation index is not less than the deviation threshold, increasing the data collection frequency; if the data feature deviation index is less than the deviation threshold, executing the function of the multi-source data evaluation module.

[0069] In this embodiment, the dynamic comparison of deviation thresholds and deviation indices is used to regulate the intensity of data adjustment measures, improving data collection accuracy and quality while avoiding unnecessary resource waste. A multi-source data evaluation module ensures data consistency and reliability, enabling intelligent, precise, and efficient data processing to meet the needs of various scenarios.

[0070] Specifically, the deviation threshold is obtained from a preset database. In one specific embodiment, the data feature data corresponding to the data feature errors in the power load data collection in the historical data is substituted into the specific restriction expression of the data feature deviation index to obtain the corresponding data set, and the result of the mean operation on the data set is recorded as the deviation threshold.

[0071] Furthermore, the specific process for obtaining the multi-source data consistency evaluation index is as follows: obtain reference consistency data from a preset database, the reference consistency data including the maximum merge error, the maximum communication interruption rate, and the maximum number of device warnings; obtain the multi-source data consistency evaluation index based on the multi-source feature data and the reference consistency data. The specific restricted expression of the multi-source data consistency evaluation index is as follows:

[0072]

[0073] Where n represents the number of the preset time period, , Indicates the number of the preset time period, Indicates the data merging error of the nth preset time period, Indicates the communication interruption rate of the nth preset time period, Indicates the number of device warnings in the nth preset time period, represents the maximum value of the combined error, Indicates the maximum value of communication interruption rate, Indicates the maximum number of device warnings. Indicates the multi-source data consistency evaluation index for the nth preset time period.

[0074] In this embodiment, the algorithm combines multi-source feature data and corresponding reference consistency data for comprehensive analysis to obtain a multi-source data consistency evaluation index. When the multi-source feature data are all smaller than the corresponding reference consistency data, as the data merging error, communication interruption rate, and number of equipment warnings increase, the multi-source data consistency evaluation index decreases, indicating that the possibility of conflict between power load data collected by different data sources is greater. Assuming that the maximum merging error and the maximum communication interruption rate are 1, and the maximum number of equipment warnings are 8, a data change table of the multi-source data consistency evaluation index is obtained, as shown in Table 1:

[0075] Table 1 Data changes of multi-source data consistency evaluation index

[0076]

[0077] As can be seen from Table 1, when the multi-source feature data are all smaller than the corresponding reference consistency data, the multi-source data consistency evaluation index increases with the decrease of data merging error, communication interruption rate, and number of equipment warnings. In this case, the data merging error, communication interruption rate, and number of equipment warnings are negatively correlated with the multi-source data consistency evaluation index. For example, when the data merging error decreases from 0.8 in the first row to 0.2 in the fifth row, the communication interruption rate decreases from 0.9 in the first row to 0.4 in the fifth row, and the number of equipment warnings decreases from 7 in the first row to 2 in the fifth row, the multi-source data consistency evaluation index increases from 0.67 in the first row to 1.25 in the fifth row, further confirming the negative correlation between the multi-source feature data and the multi-source data consistency evaluation index. When the communication interruption rate is higher, it may lead to an increase in the data merging error and trigger more equipment warnings, which also increases the number of equipment warnings. The multi-source data consistency evaluation index is only needed when the data feature deviation index is less than the deviation threshold. However, a larger multi-source data consistency evaluation index indicates less conflict between data collected from different data sources, which is more conducive to accurate power load forecasting.

[0078] Specifically, the reference consistency data is obtained from a preset database. In one specific embodiment, the reference consistency data is set by professionals related to the power load system based on specific industrial conditions. For example, in an industrial automation environment, it can be assumed that the maximum combined error is 0.05, the maximum communication interruption rate is 0.001, and the maximum number of device warnings is 5.

[0079] Furthermore, the specific process for determining whether to perform equipment operation and maintenance is as follows: a multi-source threshold is obtained from a preset database, and the multi-source threshold is used to determine whether to perform equipment operation and maintenance; the multi-source data consistency evaluation index is compared with the multi-source threshold: if the multi-source data consistency evaluation index is less than the multi-source threshold, equipment operation and maintenance is performed, and equipment operation and maintenance means maintaining communication links and equipment to ensure network coverage and transmission quality, and at the same time inspecting and collecting the operating status of the equipment and performing maintenance; if the multi-source data consistency evaluation index is not less than the multi-source threshold, equipment operation and maintenance is not performed.

[0080] In this embodiment, during equipment operation and maintenance, the communication link uses Ping and Traceroute to locate the faulty link, check whether there is signal attenuation at the fiber optic node, and check the physical link to check whether the fiber optic connector is loose and replace it to ensure network coverage and transmission quality. Through accurate data consistency evaluation and intelligent operation and maintenance decision-making, data reliability is improved, operation and maintenance decisions are optimized, network stability is enhanced, operation and maintenance efficiency is improved, and system security is guaranteed. In addition, through inspection and maintenance of collection equipment, potential faults can be discovered and eliminated in a timely manner, reducing system downtime and maintenance costs.

[0081] Specifically, the multi-source threshold is obtained from a preset database. In one specific embodiment, the multi-source feature data corresponding to the data source problem in the historical power load data collection is substituted into the specific restriction expression of the multi-source data consistency evaluation index to obtain the corresponding data set, and the result of the mean operation on the data set is recorded as the multi-source threshold.

[0082] To sum up, the embodiment of the present application obtains the time series anomaly index through the acquired time series data and determines whether to take time adjustment measures, then obtains the numerical characteristic anomaly index through the acquired numerical characteristic data to determine whether to take data adjustment measures, and then obtains the data characteristic deviation index through the acquired data feature data to determine whether to regulate the intensity of the data adjustment measures, and finally obtains the multi-source data consistency evaluation index through the acquired multi-source feature data to determine whether to perform equipment operation and maintenance, thereby making the power load data more timely, and thus achieving the acquisition of more accurate data for power load forecasting, effectively solving the problem of unstable data source when performing power load forecasting in the prior art.

[0083] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0084] The present invention is described with reference to flowcharts and / or block diagrams of systems, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0085] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1The function specified in one or more boxes.

[0086] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps for the function specified in one or more boxes.

[0087] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0088] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A virtual power plant load forecasting system based on big data, characterized in that: Including time series evaluation module, numerical characteristic evaluation module, data feature evaluation module and multi-source data evaluation module: The time series evaluation module is used to obtain a time series anomaly index from the acquired time series data, and determine whether to take time adjustment measures based on the time series anomaly index. The time series anomaly index is used to quantify the probability of anomalies in the time aspect of the power load data during the acquisition process. The numerical characteristic evaluation module is used to obtain a numerical characteristic anomaly index from the acquired numerical characteristic data, and determine whether to take data adjustment measures based on the numerical characteristic anomaly index. The numerical characteristic anomaly index is used to quantify the probability of abnormality in the numerical value of the power load data during the collection process; The data feature evaluation module is used to obtain a data feature deviation index from the acquired data feature data, and determine whether to adjust the intensity of data adjustment measures based on the data feature deviation index. The data feature deviation index is used to quantify the probability of deviation of feature data during the collection process of power load data; The multi-source data evaluation module is used to obtain a multi-source data consistency evaluation index by acquiring multi-source feature data within a preset time period, and to determine whether to perform equipment operation and maintenance based on the multi-source data consistency evaluation index. The multi-source data consistency evaluation index is used to quantify the probability of conflicts between multiple data sources during the collection of power load data; The time series data includes collection frequency, collection data volume, data missing rate, data delay duration and timestamp deviation; The numerical characteristic data include short-term fluctuation amplitude, saturation value duration, out-of-range abnormal point ratio and invalid value ratio; The data characteristic data include load stability index, fluctuation frequency, fluctuation amplitude, peak-to-valley deviation and mean deviation; The multi-source feature data includes data merging error, communication interruption rate and number of equipment warnings.

2. The virtual power plant load forecasting system based on big data according to claim 1, characterized in that: The specific steps for obtaining the time series anomaly index are as follows: Numbering the preset time periods and obtaining reference time series data from a preset database, wherein the reference time series data includes a maximum data delay duration and a maximum timestamp deviation; Obtain the corresponding preset duration according to the preset time period; If the time series data meets the first judgment condition, the time series data and the corresponding reference time series data are processed to obtain a time series anomaly index. The first judgment condition indicates that the data delay duration and the timestamp deviation are both less than the corresponding maximum data delay duration and the maximum timestamp deviation.

3. The virtual power plant load forecasting system based on big data according to claim 2, characterized in that: The specific process for determining whether to take time adjustment measures is as follows: If the time series data does not meet the first judgment condition, then directly take time adjustment measures; otherwise, obtain a time threshold from a preset database, and the time threshold is used to determine whether to take time adjustment measures; Compare the time series anomaly index to a time threshold: If the time series anomaly index is not less than the time threshold, time adjustment measures are taken in the next preset time period, and the time adjustment measures include increasing bandwidth and using a time synchronization protocol; If the time series anomaly index is less than the time threshold, the function of the numerical characteristic evaluation module is executed.

4. The big data-based virtual power plant load forecasting system according to claim 1, characterized in that: The specific steps for obtaining the numerical characteristic abnormality index are: Acquire reference numerical characteristic data from a preset database, wherein the reference numerical characteristic data includes a maximum short-time fluctuation amplitude and a maximum saturation value duration; Compare the numerical characteristic data with the reference numerical characteristic data: If the numerical characteristic data does not meet the second judgment condition, data adjustment measures are directly taken, wherein the second judgment condition indicates that the short-term fluctuation amplitude and the saturation value duration are both less than the corresponding short-term fluctuation amplitude maximum value and the saturation value duration maximum value; If the numerical characteristic data meets the second judgment condition, the ratio of the short-term fluctuation amplitude to the maximum short-term fluctuation amplitude and the ratio of the saturation value duration to the maximum saturation value duration are subjected to hyperbolic cotangent processing to obtain the first characteristic index, and the sum of the proportion of out-of-range abnormal points and the proportion of invalid values is subjected to inverse cotangent processing to obtain the second characteristic index; A numerical characteristic abnormality index is obtained according to the first characteristic index and the second characteristic index.

5. The virtual power plant load forecasting system based on big data according to claim 4, characterized in that: The specific process of determining whether to take data adjustment measures is as follows: If the numerical characteristic data does not meet the second judgment condition, the data adjustment measures are directly taken; otherwise, a numerical threshold is obtained from a preset database, and the numerical threshold is used to determine whether to take the data adjustment measures; Compare the numerical property anomaly index to the numerical threshold: If the numerical characteristic abnormality index is not less than the numerical threshold, data adjustment measures are taken in the next preset time period. The data adjustment measures include adjusting the data collection frequency and abnormality marking. The abnormality marking indicates marking the saturated value data and ignoring the marked saturated value in subsequent analysis. If the numerical characteristic anomaly index is less than the numerical threshold, the function of the data characteristic evaluation module is executed.

6. The virtual power plant load forecasting system based on big data according to claim 1, characterized in that: The specific process of obtaining the data feature deviation index is as follows: Obtain reference characteristic data and deviation assessment weights from a preset database, wherein the reference characteristic data includes a maximum stability index, a maximum fluctuation amplitude, a maximum fluctuation frequency, a maximum peak-to-valley deviation, and a maximum mean deviation, and the deviation assessment weights include a load weight, a fluctuation weight, and a deviation weight; Obtaining deviation indicators based on the data feature data and the corresponding reference feature data, wherein the deviation indicators include a first deviation indicator, a second deviation indicator, a third deviation indicator, a fourth deviation indicator, and a fifth deviation indicator; The deviation index and deviation assessment weight are processed to obtain the data feature deviation index.

7. The virtual power plant load forecasting system based on big data according to claim 6, characterized in that: The specific process for determining whether to regulate the intensity of data adjustment measures is as follows; Obtaining a deviation threshold from a preset database, wherein the deviation threshold is used to determine whether to regulate the intensity of data adjustment measures; Compare the data feature deviation index with the deviation threshold: If the data feature deviation index is not less than the deviation threshold, the data collection frequency is increased; If the data feature deviation index is less than the deviation threshold, the function of the multi-source data evaluation module is executed.

8. The virtual power plant load forecasting system based on big data according to claim 1, characterized in that: The specific process of obtaining the multi-source data consistency evaluation index is as follows: Acquire reference consistency data from a preset database, wherein the reference consistency data includes a maximum merge error, a maximum communication interruption rate, and a maximum number of device warnings; A multi-source data consistency evaluation index is obtained based on the multi-source feature data and the reference consistency data. The specific restricted expression of the multi-source data consistency evaluation index is as follows: Where n represents the number of the preset time period, , Indicates the number of the preset time period, Indicates the data merging error of the nth preset time period, Indicates the communication interruption rate of the nth preset time period, Indicates the number of device warnings in the nth preset time period, represents the maximum value of the combined error, Indicates the maximum value of communication interruption rate, Indicates the maximum number of device warnings. Indicates the multi-source data consistency evaluation index for the nth preset time period.

9. The virtual power plant load forecasting system based on big data according to claim 8, characterized in that: The specific process of determining whether to perform equipment operation and maintenance is as follows: Obtaining a multi-source threshold from a preset database, wherein the multi-source threshold is used to determine whether to perform equipment operation and maintenance; Compare the multi-source data consistency assessment index with the multi-source threshold: If the multi-source data consistency evaluation index is less than the multi-source threshold, equipment operation and maintenance is performed. The equipment operation and maintenance means maintaining the communication links and equipment to ensure network coverage and transmission quality, and inspecting the operating status of the collection equipment and performing maintenance. If the multi-source data consistency evaluation index is not less than the multi-source threshold, equipment operation and maintenance will not be performed.

Images