A load prediction and intelligent scheduling system of a virtual power plant
Patent Information
- Application Number
- CN202611091283.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-22
- Publication Date
- 2026-09-15
Smart Images
Figure CN122763348A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to, specifically to, a load forecasting and intelligent dispatching system for a virtual power plant. Background Technology
[0002] A virtual power plant is a cloud-based intelligent Internet of Things (IoT) platform that aggregates a large number of dispersed, adjustable electricity loads, distributed power sources, and energy storage facilities through advanced information and communication technologies, forming a unified and coordinated "virtual" power resource. Virtual power plant deployment is mainly divided into two levels: "campus-level / local virtual power plants" and "wide-area / aggregated virtual power plants." The service scope of "campus-level / local virtual power plants" is limited to a clearly defined physical campus. By coordinating photovoltaic, energy storage, gas generators, and controllable loads (such as air conditioning and lighting) within the campus, it achieves overall self-regulation of electricity consumption within the campus.
[0003] However, in the existing technology of "park-level / local virtual power plants", when scheduling multiple controllable loads of equipment in the park, the scheduling order is usually set based on fixed priority or human experience. This results in a lack of objective and quantifiable evaluation basis for scheduling decisions, which leads to insufficient flexibility in load regulation and affects the final scheduling efficiency. Summary of the Invention
[0004] To address the aforementioned shortcomings of existing technologies, this invention provides a load forecasting and intelligent dispatching system for virtual power plants, which can effectively solve the problem of lack of quantitative assessment when scheduling load regulation for multi-point equipment in existing technologies.
[0005] To achieve the above objectives, the present invention provides the following technical solution: This invention provides a load forecasting and intelligent dispatching system for a virtual power plant, comprising at least: The demand forecasting unit divides each calendar day into 24 time periods. Based on historical load data of multiple electrical facilities within the target park, it analyzes the historical load data to calculate the predicted electricity consumption for the current time period. The electricity demand of electrical facilities is divided into fixed electrical facilities and callable electrical facilities. Based on empirical values, the electricity consumption of fixed electrical facilities and callable electrical facilities in a unit time period is predicted, and the predicted electricity consumption is obtained by summing them. The output prediction unit predicts the projected power generation of the park in the current time period based on the parameter analysis of the photovoltaic power generation system in the target park. The integrated dispatch unit generates a load reduction signal when the predicted power generation in the current unit time period is less than or equal to the predicted power consumption. Based on the difference between the predicted power generation and the predicted power consumption, and combined with the equipment operation parameters of each available power facility in the target park, a comprehensive dispatchability score is calculated. Based on the comprehensive dispatchability score, the control priority of the available power facilities in the target park is ranked. When the total photovoltaic power supply of the target park's internal power grid is less than the total power consumption, available power facilities are selected for load reduction based on the control priority.
[0006] Furthermore, the calculation process for predicted electricity consumption is as follows: The historical experience value of electricity consumption for each fixed power facility in each unit time period is recorded as the unit's basic electricity consumption. Based on the analysis of historical power generation data of the available power facilities, a power change prediction map of the available power facilities is obtained. The power change broken line in the power change prediction map of the available power facilities is recorded as the target broken line. The area enclosed by the target broken line and the time axis in the current unit time period is calculated and recorded as the unit variable power consumption. The predicted electricity consumption is obtained by summing the basic electricity consumption per unit and the variable electricity consumption per unit.
[0007] Furthermore, the process for obtaining the power change prediction map of the available electrical facilities is as follows: Each natural day is used as a monitoring cycle, and the week number corresponding to the monitoring cycle is used as its cycle number. The load power of the available power facilities is recorded once every unit time period within the monitoring cycle. A total monitoring duration is set, which is one week. In the same coordinate system, the load power change over time corresponding to each cycle number within the total monitoring period is plotted as a load change line. The similarity between any two load change lines is calculated. Based on the similarity, the load change lines are classified to obtain multiple load similarity sets. Each load similarity set corresponds to multiple cycle numbers to form a cycle set. The fitted line of each load change line in the same load similarity set is plotted as the prediction fitted line. Each predicted fitted line corresponds to a set of periods. The period number corresponding to the next day is obtained, and the predicted fitted line corresponding to the period set to which the period number belongs is used as the power prediction line and a power change prediction graph is drawn.
[0008] Furthermore, the process of partitioning the load similarity set is as follows: Let any two load change line graphs be denoted as the first line graph and the second line graph, respectively. Draw the regions enclosed by the first line graph, the second line graph, and the horizontal axis of the coordinate system, denoted as the first projection region and the second projection region, respectively. Calculate the areas of the first projection region and the second projection region, denoted as S1 and S2, respectively. Calculate the area of the overlapping portion of the first projection region and the second projection region, denoted as S3. Let the similarity be equal to S3 / (S1+S2). A preset similarity threshold is given. Within the same load similarity set, the following conditions must be met: Condition 1: The similarity between any two load change lines is greater than or equal to the similarity threshold; Condition 2: The total number of load-similar sets is the minimum value that satisfies Condition 1.
[0009] Furthermore, the process of plotting the predicted fitted polyline is as follows: The load power values of multiple load change polygons in the load similarity set at the same time are obtained and the average value is calculated and recorded as the load mean. The load mean corresponding to each time is recorded as the fitted power value corresponding to that time. In this way, multiple fitted power values uniformly distributed along the time axis within a single monitoring period are obtained, and the predicted fitted polygon is obtained by connecting them.
[0010] Furthermore, the comprehensive score calculation process for the dispatchability of available power facilities is as follows: The callable electrical facilities are denoted as target equipment. Based on the equipment parameters of the target equipment, the response time, response depth, and response reliability index are calculated. The response time is a negative indicator, while the response depth and response reliability index are positive indicators. The response time is reverse normalized, and the response depth is directly normalized. The system includes preset duration weight, depth weight, and reliability weight, which correspond to response duration, response depth, and response reliability index, respectively. The schedulability comprehensive score is obtained by multiplying the response time, response depth, and response reliability index by their respective weighting coefficients and summing them. Among them, the depth weight is proportional to the difference between the predicted power generation and the predicted power consumption, the sum of the duration weight, the depth weight and the reliability weight is 1, and the ratio of the duration weight to the reliability weight is a preset value.
[0011] Furthermore, the response time of the target device is calculated as follows: Obtain the maximum and minimum adjustable power of the target device. Test the target device to reduce its power from the maximum to the minimum adjustable power and record the time required as the first adjustment time. Divide the difference between the maximum and minimum adjustable power by the adjustment time to calculate the average power adjustment speed. Obtain the rated power of the target device. Test the target device to reduce its power from the rated power to the minimum adjustable power and record the time required as the second adjustment time. Calculate the theoretical time for the target device to reduce its power from the rated power to the minimum adjustable power based on the average power adjustment speed. Calculate the difference between the second adjustment time and the theoretical time to obtain the response time.
[0012] Furthermore, the response depth calculation process for the target device is as follows: Obtain the load change line of the target device during the previous monitoring period, and divide the power value range of the load change line into multiple continuous power range intervals, with a fixed total number of power range intervals; Each load change line is divided into multiple unit lines according to the power range interval. The sum of the projected lengths of the unit lines within each power range interval is recorded as the duration. A duration threshold is preset. Power range intervals with a duration longer than the duration threshold are recorded as valid intervals. The midpoint of the valid interval is recorded as the power duration value. The power duration value corresponding to the maximum duration value is selected and the minimum power duration value is subtracted to obtain the response depth.
[0013] Furthermore, the calculation process for the response reliability index of the target device is as follows: The adjustment response of the target equipment during the previous monitoring period is statistically analyzed. The total number of adjustment commands sent and the number of responses to the adjustment commands are recorded. The ratio of the number of responses to the total number of commands is recorded as the response rate. The difference between the adjustment result after the adjustment command is executed and the target result is recorded as the adjustment deviation. The average adjustment deviation of all adjustment commands is recorded as the response skewness. The response skewness is reverse normalized and summed with the response rate to calculate the response reliability index.
[0014] Furthermore, a power generation prediction model is pre-set, which further divides the unit time period into multiple time periods, predicts the irradiance in each time period, and substitutes the total irradiance of the tilted surface of the photovoltaic panel, the area of the photovoltaic panel, the preset system conversion efficiency, and the duration of each time period into the model to calculate the predicted power generation.
[0015] The technical solution provided by this invention has the following advantages compared with the known prior art: 1. When predicting the electricity consumption of callable power facilities, this invention constructs a load similarity set to dynamically divide and combine them based on the geometric similarity of the projected area of the load curve. It can autonomously discover and aggregate electricity consumption records with truly similar forms of callable power facilities, so that load change lines with similar load performance can be grouped together, thereby providing more typical prediction samples and making the prediction fitting lines of the callable power facilities more accurate in the final fitting.
[0016] 2. This invention calculates a comprehensive dispatchability score, enabling operations to be performed according to a more reasonable priority order during load reduction and adjustment. Furthermore, the depth weight in the comprehensive dispatchability score calculation process is proportional to the size of the power generation gap, achieving dynamic adjustment of the priority of power equipment reduction. This ensures that when the power generation gap is larger, load reduction will prioritize power equipment that supports larger power adjustments, thereby minimizing the number of equipment to be adjusted and improving the efficiency of load reduction and adjustment. Attached Figure Description
[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.
[0018] Figure 1 This is an overall module block diagram of the present invention. Detailed Implementation
[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.
[0020] The present invention will be further described below with reference to embodiments.
[0021] See Figure 1 A load forecasting and intelligent dispatching system for virtual power plants is applicable to the integrated management of power resources within industrial parks, where the parks served by the virtual power plants are designated as target parks. The system includes at least: The demand forecasting unit divides each natural day into 24 time periods (each time period is 1 hour long). Based on the analysis of historical load data of multiple power load units in the target park, it calculates the predicted electricity consumption. By calculating the predicted electricity consumption, it predicts the load demand of the park in the current time period, thereby making targeted adjustments to the energy management strategy.
[0022] Specifically, the calculation process for predicted electricity consumption is as follows: Multiple power-consuming facilities (i.e., power load units) within the target park are acquired. Based on the power demand of these facilities, they are divided into fixed power-consuming facilities and callable power-consuming facilities. The historical experience value of the power consumption of each fixed power-consuming facility in each unit time period is obtained and recorded as the unit basic power consumption (for example, if the power consumption of a certain machine tool in the same unit time period of the previous week was 5 kWh, then its historical experience value of power consumption in that unit time period is 5 kWh). Based on the historical power generation data of callable power-consuming facilities, a power change prediction map of callable power-consuming facilities is obtained. Combined with the unit basic power consumption, the predicted power consumption is calculated.
[0023] Specifically, the calculation process for predicted electricity consumption is as follows: The power change line in the power change prediction graph of the available power facilities is recorded as the target line. The area enclosed by the target line and the time axis in the current unit time period is recorded as the unit variable power consumption. The predicted power consumption is obtained by summing the unit basic power consumption and the unit variable power consumption.
[0024] It should be noted that fixed power facilities refer to power facilities that, during operation within the target park, must strictly meet the requirements of production, safety, or continuous service in terms of power consumption, operating time, or operating status, have regular load operation, and have little or no adjustment space. Demandable power facilities refer to power facilities whose power consumption can be adjusted, shifted, or controlled by external systems on demand, while meeting basic usage needs (for example, central air conditioning can appropriately reduce its operating power and power load, and the cooling temperature can be increased by 1-2 degrees Celsius, provided that basic cooling needs are met).
[0025] More specifically, the process for obtaining the power change prediction map of the available power facilities is as follows: Each natural day is used as a monitoring cycle, and the week number corresponding to the monitoring cycle is used as its cycle number. The load power of the available power facilities is recorded once every hour within the monitoring cycle. A total monitoring period is set to one week. On the same coordinate system, the load power corresponding to each cycle number is plotted as a time-varying line within the total monitoring period, and recorded as the load change line. The similarity between any two load change lines is calculated. Based on the similarity, the load change lines are classified to obtain multiple load similarity sets. Each load similarity set corresponds to multiple cycle numbers to form a cycle set (the cycle set contains multiple cycle numbers, such as Saturday and Sunday). The fitted lines of each load change line in the same load similarity set are plotted as the predicted fitted lines. Each predicted fitted line corresponds to a cycle set. The cycle number corresponding to the next day is obtained. The predicted fitted line corresponding to the cycle set to which the cycle number belongs is used as the power prediction line, and a power change prediction graph is plotted.
[0026] The process of partitioning the load similarity set is as follows: Let any two load change line graphs be denoted as the first line graph and the second line graph, respectively. The regions enclosed by the first line graph, the second line graph, and the horizontal axis (time axis) are denoted as the first projection region and the second projection region, respectively. The areas of the first projection region and the second projection region are calculated and denoted as S1 and S2, respectively. The area of the overlapping portion of the first projection region and the second projection region is calculated and denoted as S3. Let the similarity be equal to S3 / (S1+S2). A preset similarity threshold is given. Within the same load similarity set, the following conditions must be met: Condition 1: The similarity between any two load change lines is greater than or equal to the similarity threshold; (ensuring the similarity between elements within the load similarity set) Condition 2: The total number of load-similar sets is the minimum value that satisfies Condition 1. (Merge load-similar sets with similar elements as much as possible.) It should be noted that by dynamically dividing and combining the load curves by calculating the geometric similarity of the projected area, the system can autonomously discover and aggregate truly similar electricity consumption days (for example, Tuesdays and Thursdays with similar production tasks may be classified into the same category, unlike the conventional division based on weekdays), thereby providing more typical learning samples for the prediction model and improving the accuracy of the prediction.
[0027] The process of plotting the predicted fitted line is as follows: The load power values of multiple load change polygons in the load similarity set at the same time are obtained and the average value is calculated and recorded as the load mean. The load mean corresponding to each time is recorded as the fitted power value corresponding to that time. In this way, multiple fitted power values uniformly distributed along the time axis within a single monitoring period are obtained, and the predicted fitted polygon is obtained by connecting them.
[0028] It should be noted that the predicted fitted line curve can, to some extent, simulate the load changes of available power facilities within a regular cycle, thereby predicting the load of power facilities over a certain period of time and achieving load forecasting based on historical data. It is worth noting that although available power facilities can be adjusted on demand and their load power fluctuates significantly, their actual operation still exhibits patterns related to the weekday sequence. For example, the load variation of the central air conditioning in a production workshop is related to weekdays and weekends, thus allowing for corresponding simulations and forecasts based on whether the next day falls on a weekend or a weekday.
[0029] The output forecasting unit predicts the projected power generation of the photovoltaic power generation system in the target park within the current time period based on parameter analysis, and then combines this prediction with load demand for power resource reserve adjustment.
[0030] It should be noted that the power generation prediction of photovoltaic power generation modules is based on existing technology, including but not limited to "physical model method that calculates theoretical power generation through mathematical formulas based on photoelectric conversion principle and system parameters", "data-driven method that learns mapping relationships from meteorological and temporal characteristics using historical data and machine learning", and hybrid methods that combine the above two, which will not be elaborated on here.
[0031] In one specific embodiment, the formula for predicting power generation is as follows: ; in: This represents the predicted total power generation; n represents the time period number (each time period is 15 minutes), n=1,2,3,4; Let A represent the total irradiance of the tilted surface of the photovoltaic panel during time period n; and let A represent the area of the photovoltaic panel. The preset system conversion efficiency (usually set to 0.75-0.85); This indicates the duration of each time period (specifically 0.25h).
[0032] The integrated dispatch unit generates a load reduction signal if the predicted power generation in the next unit time period is less than or equal to the predicted power consumption. Based on the difference between the predicted power generation and the predicted power consumption, and combined with the equipment operation parameters of each available power facility in the target park, a comprehensive dispatchability score is calculated. Based on the comprehensive dispatchability score, the control priority of available power facilities in the target park is ranked. When the power supply of the microgrid (the power grid system inside the park) is insufficient (when the total photovoltaic power supply of the power grid inside the target park is less than the total power consumption), available power facilities are selected for load reduction according to the control priority to keep the power supply and consumption in the microgrid balanced (when the load reduction space is insufficient to support the power generation gap, power supply from the external power grid or energy storage equipment can be introduced).
[0033] It should be noted that load reduction of callable power facilities refers to reducing the operating power of callable power facilities through a series of operating instructions, thereby reducing the overall load of the power grid within the park and maintaining a balance between supply and demand. The specific operating process is based on existing technology and will not be elaborated on here. For example, central air conditioning can reduce load by providing temperature in cooling mode.
[0034] Specifically, the comprehensive score calculation process for the dispatchability of available power facilities is as follows: Based on the equipment parameter analysis of the callable power facilities, the response duration (seconds), response depth (kW), and response reliability index are calculated. The response duration is a negative indicator (the smaller the value, the higher the score), while the response depth and response reliability index are positive indicators (the larger the value, the higher the score). The response duration is reverse normalized, while the response depth is directly normalized. There are preset duration weights, depth weights, and reliability weights, which correspond to the response duration, response depth, and response reliability index, respectively. The depth weight is proportional to the difference between the predicted power generation and the predicted power consumption. The sum of the duration weight, depth weight, and reliability weight is 1, and the ratio of the duration weight to the reliability weight is a preset value.
[0035] The time length weight, depth weight, and reliability weight satisfy the following formula ; in: These represent the duration weight, depth weight, and reliability weight, respectively. This represents the difference between predicted power generation and predicted power consumption, where k represents the depth weighting. The proportionality coefficient Z represents the ratio of duration weight to reliability weight.
[0036] The schedulability comprehensive score is obtained by multiplying the response time, response depth, and response reliability index by their respective weighting coefficients and summing them.
[0037] It should be noted that the dispatchability comprehensive score reflects the priority of any dispatchable power facility when participating in load regulation as a load reduction target. It is obtained by weighted summation of multiple data points, which are mainly divided into two categories: regulation factors and cost factors. Regulation factors (response depth, response reliability index) are related to the regulation capacity of the dispatchable power facility, while cost factors are related to the time loss during regulation. A higher dispatchability comprehensive score means a higher priority for the dispatchable power facility when participating in load regulation. Reducing the load on this facility can meet load fluctuation demands with minimal impact, ensuring a balance between power supply and demand within the park.
[0038] Furthermore, by setting the depth weight of the dispatchability comprehensive score to be proportional to the size of the power generation gap, the priority of power-consuming equipment can be dynamically adjusted. This ensures that when the power generation gap is larger, load reduction will give priority to power-consuming equipment that supports larger power adjustments, thereby minimizing the number of equipment to be adjusted and improving the efficiency of load reduction adjustments.
[0039] Furthermore, the available electrical facilities are designated as target devices, and the response time of the target devices is calculated as follows: Obtain the maximum and minimum adjustable power of the target device. Test the target device to reduce its power from the maximum to the minimum adjustable power and record the time required, which is recorded as the first adjustment time. Divide the difference between the maximum and minimum adjustable power by the adjustment time to calculate the average power adjustment speed. Obtain the rated power of the target device. Test the target device to reduce its power from the rated power to the minimum adjustable power and record the time required, which is recorded as the second adjustment time. Substitute these values into the formula... The response time is calculated in the process. ,in Indicates the credit limit power. Indicates the minimum adjustable power. Indicates the average speed of power regulation. This indicates the duration of the second adjustment.
[0040] Response time is usually related to the time required for the target device to respond to and actually execute the adjustment command after receiving it. The longer the response time, the longer it takes for the target device to execute the adjustment command, which indirectly affects the command execution efficiency. In other words, the longer the response time, the less suitable the target device is as a load adjustment device in terms of response efficiency.
[0041] The process of calculating the response depth of the target device is as follows: The load change curve of the target device within the previous monitoring period is obtained. The power value range of the load change curve is divided into multiple continuous power range intervals. The total number of power range intervals is fixed (10 in a specific embodiment). Each load change curve is divided into multiple unit curves according to the power range intervals. The sum of the projected lengths of the unit curves in each power range interval is recorded as the duration. A duration threshold is preset. Power range intervals with a duration greater than the duration threshold are recorded as valid intervals. The midpoint of the valid interval is recorded as the power duration value. The power duration value corresponding to the maximum duration value is selected (that is, the duration of the valid interval corresponding to the power duration value is the longest, which usually corresponds to the normal operating power of the target device) and the minimum power duration value is subtracted (which usually corresponds to the minimum acceptable power of the target device) to obtain the response depth.
[0042] It should be noted that the response depth reflects the magnitude of power regulation performed by the target device under normal circumstances, and the greater the response depth, the greater the power reduction that the target device can achieve, and therefore the higher the priority it is selected when participating in load shedding.
[0043] The calculation process for the response reliability index of the target device is as follows: The adjustment response of the target equipment during the previous monitoring period is statistically analyzed. The total number of adjustment commands sent and the number of responses to the adjustment commands are recorded. The ratio of the number of responses to the total number of commands is recorded as the response rate. The difference between the adjustment result after the adjustment command is executed and the target result is recorded as the adjustment deviation. The average adjustment deviation of all adjustment commands is recorded as the response skewness. The response skewness is reverse normalized (the larger the original value, the smaller the normalized value) and summed with the response rate to calculate the response reliability index.
[0044] It should be noted that the adjustment command refers to the instruction action to control the increase or decrease of the operating power of the target equipment. The response reliability index reflects the reliability of the target equipment in responding to the command when the load is adjusted. The higher the response reliability index, the stronger the reliability of the equipment in responding to the adjustment command.
[0045] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the protection scope of the technical solutions of the embodiments of the present invention.
Claims
1. A load prediction and intelligent scheduling system for a virtual power plant, characterized in that, include: The demand forecasting unit divides each calendar day into 24 time periods. Based on historical load data of multiple electrical facilities within the target park, it analyzes the historical load data to calculate the predicted electricity consumption for the current time period. The electricity demand of electrical facilities is divided into fixed electrical facilities and callable electrical facilities. Based on empirical values, the electricity consumption of fixed electrical facilities and callable electrical facilities in a unit time period is predicted, and the predicted electricity consumption is obtained by summing them. The output prediction unit predicts the projected power generation of the park in the current time period based on the parameter analysis of the photovoltaic power generation system in the target park. The integrated dispatch unit generates a load reduction signal when the predicted power generation in the current unit time period is less than or equal to the predicted power consumption. Based on the difference between the predicted power generation and the predicted power consumption, and combined with the equipment operation parameters of each available power facility in the target park, a comprehensive dispatchability score is calculated. Based on the comprehensive dispatchability score, the control priority of the available power facilities in the target park is ranked. When the total photovoltaic power supply of the target park's internal power grid is less than the total power consumption, available power facilities are selected for load reduction based on the control priority. 2.The load forecasting and intelligent scheduling system of a virtual power plant according to claim 1, wherein, The calculation process for predicted electricity consumption is as follows: The historical experience value of electricity consumption for each fixed power facility in each unit time period is recorded as the unit's basic electricity consumption. Based on the analysis of historical power generation data of the available power facilities, a power change prediction map of the available power facilities is obtained. The power change broken line in the power change prediction map of the available power facilities is recorded as the target broken line. The area enclosed by the target broken line and the time axis in the current unit time period is calculated and recorded as the unit variable power consumption. The predicted electricity consumption is obtained by summing the basic electricity consumption per unit and the variable electricity consumption per unit. 3.The load forecasting and intelligent scheduling system of a virtual power plant according to claim 2, characterized in that, The process for obtaining the power change prediction map of available power facilities is as follows: Each natural day is used as a monitoring cycle, and the week number corresponding to the monitoring cycle is used as its cycle number. The load power of the available power facilities is recorded once every unit time period within the monitoring cycle. A total monitoring duration is set, which is one week. In the same coordinate system, the load power change over time corresponding to each cycle number within the total monitoring period is plotted as a load change line. The similarity between any two load change lines is calculated. Based on the similarity, the load change lines are classified to obtain multiple load similarity sets. Each load similarity set corresponds to multiple cycle numbers to form a cycle set. The fitted line of each load change line in the same load similarity set is plotted as the prediction fitted line. Each predicted fitted line corresponds to a set of periods. The period number corresponding to the next day is obtained, and the predicted fitted line corresponding to the period set to which the period number belongs is used as the power prediction line and a power change prediction graph is drawn.
4. The load forecasting and intelligent dispatching system of a virtual power plant according to claim 3, characterized in that, The process of partitioning the load similarity set is as follows: Let any two load change line graphs be denoted as the first line graph and the second line graph, respectively. Draw the regions enclosed by the first line graph, the second line graph, and the horizontal axis of the coordinate system, denoted as the first projection region and the second projection region, respectively. Calculate the areas of the first projection region and the second projection region, denoted as S1 and S2, respectively. Calculate the area of the overlapping portion of the first projection region and the second projection region, denoted as S3. Let the similarity be equal to S3 / (S1+S2). A preset similarity threshold is given. Within the same load similarity set, the following conditions must be met: Condition 1: The similarity between any two load change lines is greater than or equal to the similarity threshold; Condition 2: The total number of load-similar sets is the minimum value that satisfies Condition 1.
5. The load forecasting and intelligent dispatching system of a virtual power plant according to claim 3, characterized in that, The process of plotting the predicted fitted line is as follows: The load power values of multiple load change polygons in the load similarity set at the same time are obtained and the average value is calculated and recorded as the load mean. The load mean corresponding to each time is recorded as the fitted power value corresponding to that time. In this way, multiple fitted power values uniformly distributed along the time axis within a single monitoring period are obtained, and the predicted fitted polygon is obtained by connecting them. 6.The load forecasting and intelligent scheduling system of a virtual power plant according to claim 3, characterized in that, The calculation process for the comprehensive score of dispatchability of available power facilities is as follows: The callable electrical facilities are denoted as target equipment. Based on the equipment parameters of the target equipment, the response time, response depth, and response reliability index are calculated. The response time is a negative indicator, while the response depth and response reliability index are positive indicators. The response time is reverse normalized, and the response depth is directly normalized. The system includes preset duration weight, depth weight, and reliability weight, which correspond to response duration, response depth, and response reliability index, respectively. The schedulability comprehensive score is obtained by multiplying the response time, response depth, and response reliability index by their respective weighting coefficients and summing them. Among them, the depth weight is proportional to the difference between the predicted power generation and the predicted power consumption, the sum of the duration weight, the depth weight and the reliability weight is 1, and the ratio of the duration weight to the reliability weight is a preset value.
7. The load forecasting and intelligent dispatching system of a virtual power plant according to claim 6, characterized in that, The response time of the target device is calculated as follows: Obtain the maximum and minimum adjustable power of the target device. Test the target device to reduce its power from the maximum to the minimum adjustable power and record the time required as the first adjustment time. Divide the difference between the maximum and minimum adjustable power by the adjustment time to calculate the average power adjustment speed. Obtain the rated power of the target device. Test the target device to reduce its power from the rated power to the minimum adjustable power and record the time required as the second adjustment time. Calculate the theoretical time for the target device to reduce its power from the rated power to the minimum adjustable power based on the average power adjustment speed. Calculate the difference between the second adjustment time and the theoretical time to obtain the response time. 8.The load forecasting and intelligent scheduling system of a virtual power plant of claim 6, wherein, The process of calculating the response depth of the target device is as follows: Obtain the load change line of the target device during the previous monitoring period, and divide the power value range of the load change line into multiple continuous power range intervals, with a fixed total number of power range intervals; Each load change line is divided into multiple unit lines according to the power range interval. The sum of the projected lengths of the unit lines within each power range interval is recorded as the duration. A duration threshold is preset. Power range intervals with a duration longer than the duration threshold are recorded as valid intervals. The midpoint of the valid interval is recorded as the power duration value. The power duration value corresponding to the maximum duration value is selected and the minimum power duration value is subtracted to obtain the response depth. 9.The load forecasting and intelligent scheduling system of a virtual power plant of claim 6, wherein, The calculation process for the response reliability index of the target device is as follows: The adjustment response of the target equipment during the previous monitoring period is statistically analyzed. The total number of adjustment commands sent and the number of responses to the adjustment commands are recorded. The ratio of the number of responses to the total number of commands is recorded as the response rate. The difference between the adjustment result after the adjustment command is executed and the target result is recorded as the adjustment deviation. The average adjustment deviation of all adjustment commands is recorded as the response skewness. The response skewness is reverse normalized and summed with the response rate to calculate the response reliability index. 10.The load forecasting and intelligent scheduling system of a virtual power plant of claim 1, wherein, A pre-set power generation prediction model is used to further divide the unit time period into multiple time periods. The irradiance within each time period is predicted. The total irradiance of the tilted surface of the photovoltaic panel, the area of the photovoltaic panel, the preset system conversion efficiency, and the duration of each time period are substituted into the model to calculate the predicted power generation.