Power distribution network power demand prediction method and device based on big data analysis, and medium

Abstract

The invention discloses a power distribution network power demand prediction method and device based on big data analysis and a medium, and belongs to the technical field of power demand prediction, and the method comprises the steps: dividing a power distribution network into a plurality of power nodes, collecting the power data of the power nodes, and carrying out the screening to obtain power abnormal nodes; demand information of the power abnormal node is collected, and a power consumption demand event is extracted according to the demand information; acquiring event information of the power consumption demand event, and evaluating according to the event information to obtain a propagation effect condition of the power consumption demand event; acquiring a real-time prediction demand value of a power abnormal node, and determining power change values of other power nodes in combination with a propagation effect condition; and collecting real-time environment data of the power node, predicting according to the real-time environment data to obtain a basic power demand value, and superposing the power change value to obtain a real-time power demand value. The accuracy of power distribution network power demand prediction based on big data analysis is improved.

Description

Technical Field

[0001] This invention relates to the field of power demand forecasting technology, specifically to a method, equipment, and medium for forecasting power demand in distribution networks based on big data analysis. Background Technology

[0002] In the power system sector, the stable operation of the distribution network is crucial for ensuring social production and people's lives, and accurate electricity demand forecasting is the foundation for optimizing distribution network scheduling and rationally allocating resources. Traditional electricity demand forecasting methods typically rely on static data such as historical load data, meteorological information, and economic indicators to build models. While these methods can reflect basic electricity demand patterns to some extent, their models are rigid and lack real-time response mechanisms for dynamic emergencies. This results in poor performance when dealing with extreme weather, equipment failures, sudden events, or rapid changes in energy supply and demand, and an inability to dynamically adjust forecasting strategies to adapt to unforeseen circumstances. Consequently, the electricity demand forecasting of the distribution network is inaccurate, further impacting the stable operation of the distribution network. Summary of the Invention

[0003] In view of the above-mentioned problems, the present invention is proposed.

[0004] Therefore, this invention aims to solve the problem of inaccurate power demand forecasting in the power grid, which further affects the stable operation of the distribution network.

[0005] To address the aforementioned technical problems, this invention provides the following technical solution: a method for predicting power demand in a distribution network based on big data analysis, comprising, The power distribution network is divided into multiple power nodes. Power data of the power nodes is collected, and abnormal power nodes are identified based on the power data. Demand information of abnormal power nodes is collected, and electricity demand events are extracted based on the demand information. Event information of electricity demand events is obtained, and the propagation effect of electricity demand events is evaluated based on the event information. Real-time predicted demand values ​​of abnormal power nodes are collected, and the power change values ​​of the remaining power nodes are determined by combining the propagation effect. Real-time environmental data of power nodes is collected, and the basic power demand value is predicted based on the real-time environmental data. The real-time power demand value is obtained by superimposing the power change value.

[0006] As a preferred embodiment of the power demand forecasting method for distribution networks based on big data analysis described in this invention, the power anomaly nodes obtained through screening include: Collect power data from power nodes, and determine whether there is insufficient power supply at the power grid nodes based on the power data. If there is insufficient power supply, the power node is determined to be a power abnormal node. If there is no power shortage, the real-time load of the power node is extracted based on the power data; Collect historical load data for power nodes, compare the real-time load with the historical load, and determine whether the real-time load is abnormal. If the real-time load is abnormal, the power node is determined to be an abnormal power node.

[0007] As a preferred embodiment of the power demand forecasting method for power distribution networks based on big data analysis described in this invention, the method of obtaining power demand events includes: collecting demand information of abnormal power nodes, collecting the usage time points corresponding to power equipment, and determining whether the power consumption time points meet the usage time points. If the usage time point is consistent, the usage quantity of electrical equipment corresponding to the historical electricity usage time point is counted and recorded as the historical usage quantity. The real-time usage quantity of electrical equipment is collected, and the difference between the historical usage quantity and the real-time usage quantity is calculated to obtain the equipment usage difference. If the usage time does not meet the requirements, the real-time number of electrical devices will be used as the device usage difference. Devices whose device usage difference reaches the preset usage difference threshold will be recorded as abnormal devices. The power demand event will be obtained by combining the abnormal devices and the node location.

[0008] As a preferred embodiment of the power demand forecasting method for power distribution networks based on big data analysis described in this invention, the step of obtaining power demand events by combining abnormal equipment and node locations includes collecting the equipment efficacy of abnormal equipment and determining whether the use of abnormal equipment is related to geographical location based on the equipment efficacy. If the abnormal equipment usage is related to geographical location, all events affecting power demand are collected as basic events; events related to geographical location are selected from the basic events and recorded as geographical events; geographical features of geographical events are extracted; location features of node locations are obtained; and geographical events with consistent geographical features and location features are used as power demand events. If the use of equipment is not related to geographical location, then events related to abnormal equipment are selected from the basic events as abnormal events; the equipment usage characteristics corresponding to the abnormal events are collected, and the power demand events are confirmed based on the equipment usage characteristics.

[0009] As a preferred embodiment of the power demand forecasting method for power distribution networks based on big data analysis described in this invention, the step of confirming the power demand event based on equipment usage characteristics includes obtaining the feature similarity of equipment usage characteristics corresponding to all abnormal events and determining whether the feature similarity is within a preset similarity range. If the similarity is within a preset range, collect the power impact values ​​corresponding to the abnormal events when they occurred in the past; The system acquires real-time power change values ​​of abnormal power nodes, calculates the difference between the power impact value and the real-time change value, collects the probability of abnormal events, and comprehensively determines the power demand events. If it is not within the preset similarity range, the device operation characteristics of the abnormal device are extracted, and the similarity between the device operation characteristics and the device usage characteristics is compared and recorded as the operation similarity. Select the abnormal event with the highest similarity to the actual event as the electricity demand event.

[0010] The beneficial effects of the preferred technical solution in the embodiments of the present invention are as follows: by comparing the real-time load with the historical load, nodes with insufficient power supply or abnormal load can be accurately identified, thereby improving the timeliness and accuracy of fault detection.

[0011] As a preferred embodiment of the power demand forecasting method for power distribution networks based on big data analysis described in this invention, the step of obtaining the propagation effect of power demand events includes determining whether the power demand event is a geographical event. If it is a geographical event, then determine whether the electricity demand event has spread; If electricity demand events spread, determine whether the occurrence of electricity demand events depends on geographical features; If relying on geographical features, the location features of the remaining power nodes are obtained, and the similarity between the geographical features and location features is compared and recorded as geographical similarity. The range of power nodes whose geographical similarity reaches a preset geographical similarity threshold is taken as the node propagation range. The node propagation speed and node propagation strength of the power nodes are obtained based on the geographical similarity. If geographical features are not relied upon, historical diffusion data of electricity demand events are collected, and the propagation effect is obtained based on the historical diffusion data; If it is not a geographical event, the propagation effect of the electricity demand event can be assessed based on the event information.

[0012] The beneficial effects of the preferred technical solution in the embodiments of the present invention are as follows: by analyzing the difference between the time points and the quantity of equipment used, combined with geographical location or equipment characteristics, abnormal events affecting power demand can be accurately identified, thereby improving the comprehensiveness and reliability of event identification.

[0013] As a preferred embodiment of the power demand forecasting method for power distribution networks based on big data analysis described in this invention, the step of obtaining the propagation effect based on historical diffusion data includes collecting historical diffusion data of power demand events, wherein the historical diffusion data includes historical diffusion range, historical diffusion speed and historical diffusion intensity. Determine whether the message spread of an electricity demand event affects electricity demand. If it does, collect message spread data of the electricity demand event. The message spread data includes message spread range, message spread speed, and message spread intensity. The propagation effect was obtained by combining historical diffusion data and message diffusion data. The assessment of the propagation effect of electricity demand events based on event information includes, If it is not a geographical event, then the event information of the electricity demand event is collected, including the duration of the event and the people who are interested in the event; Determine whether the electricity demand event is a sudden event. If it is not a sudden event, collect the historical propagation time of the electricity demand event and determine whether the duration of the event has reached the historical propagation time. If the historical propagation duration is not reached, the difference between the historical propagation duration and the duration of the event is calculated. Establish a correlation table between the difference in propagation duration and the propagation effect, and find the propagation effect based on the correlation table; If the electricity demand event is a sudden event, the information dissemination channels of the electricity demand event are collected, and the dissemination effect is combined with the information of the people concerned about the event. The aforementioned situation regarding the dissemination effect obtained by combining the event's target audience includes using the average speed at which the event's target audience receives information in the statistical information dissemination channels as the node dissemination speed. Based on the characteristics of the people who are interested in the event, we can determine their distribution range and use this range as the propagation range of the nodes. The percentage of people who pay attention to events at different power nodes is used as a measure of the node's dissemination strength.

[0014] The beneficial effects of the preferred technical solution in the embodiments of the present invention are as follows: by evaluating the geographical characteristics, historical diffusion data and information dissemination channels of an event, it is possible to dynamically predict the spread and impact of the event in the power distribution network, thereby enhancing the foresight of demand forecasting.

[0015] As a preferred embodiment of the power demand forecasting method for distribution networks based on big data analysis described in this invention, the step of determining the power change value of the remaining power nodes includes: recording the power nodes within the node propagation range as power forecast nodes, and calculating the distance between the power anomaly nodes and the power forecast nodes. The time points of power change are calculated based on the node propagation speed and distance, and the power anomaly values ​​of nodes with power anomalies are collected. Obtain real-time time points, calculate the time difference between the power change time points and the real-time time points, and obtain the real-time predicted demand value based on the time difference; When the real-time predicted demand value reaches the preset demand value threshold, the power change value of the power prediction node at the power change time point is calculated based on the power anomaly value and the node propagation strength.

[0016] The preferred technical solution in the embodiments of the present invention has the following beneficial effects: by combining the propagation effect and real-time predicted demand value, it is possible to calculate the power changes of other nodes at a specific point in time, thereby achieving more accurate power demand allocation and adjustment.

[0017] The present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the steps of the aforementioned method for predicting power demand in a power distribution network based on big data analysis.

[0018] The present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements the steps of the power demand forecasting method for power distribution networks based on big data analysis.

[0019] The beneficial effects of this invention are as follows: This invention divides the power distribution network into multiple power nodes, collects the voltage, equipment load, and frequency of the power nodes, and identifies nodes with insufficient power supply as power anomaly nodes. For nodes without insufficient power supply, the existence of anomalies is determined based on their real-time load fluctuations, thereby identifying power anomaly nodes. Power demand events are extracted from the demand information of power anomaly nodes, and the propagation effect of these events is evaluated based on their event information. Power nodes within the propagation range of a node are recorded as power prediction nodes, and the power change time point is calculated based on the distance between the power anomaly node and the power prediction node, as well as the node propagation speed. The real-time time point is obtained, and the time difference between the power change time point and the real-time time point is calculated. The real-time predicted demand value is obtained based on this time difference. When the real-time predicted demand value reaches a preset demand value threshold, the power change value of the power prediction node at the power change time point is calculated based on the power anomaly value of the power anomaly node and the node propagation strength. Real-time environmental data of the power nodes is collected, and the basic power demand value is predicted based on this real-time environmental data. The real-time power demand value is then obtained by superimposing the power change value. By dividing the power grid into multiple nodes and predicting the power demand of other nodes based on known abnormal fluctuations in power demand, this method is more accurate than conventional static forecasting. Because it dynamically predicts sudden events that have already occurred in power demand, it improves the accuracy of power demand forecasting for distribution networks based on big data analysis.

[0020] The system collects the usage time points of electrical equipment corresponding to power anomaly nodes and determines whether these usage time points match the established usage time points. If they match, the difference between the number of devices used at historical usage time points and the current usage number is calculated as the device usage difference. If they do not match, the current usage number of devices is used as the device usage difference. Devices whose usage difference reaches a preset threshold are marked as abnormal devices. The system then determines whether the usage of abnormal devices is related to their geographical location based on their device effectiveness. If related, events whose geographical features match the node's location are selected from all events affecting power demand as power demand events. If unrelated, events related to abnormal devices are selected from basic events as abnormal events. If all abnormal events are similar, power demand events are selected based on their historical power impact value and frequency. If they are not similar, the abnormal event that best matches the actual situation is selected as the power demand event. By identifying nodes with abnormal electricity demand, the events causing these abnormalities can be inferred. This not only facilitates power dispatching based on the development of these events but also allows for the prediction of changes in electricity demand at other nodes, thus improving the convenience of power demand forecasting for distribution networks based on big data analysis.

[0021] If the electricity demand event is a geographical event and its spread depends on geographical features, then the range of power nodes with location features similar to geographical features is taken as the node propagation range. Simultaneously, the node propagation speed and strength are obtained based on geographical similarity. If it does not depend on geographical features, historical diffusion data and message diffusion data of the electricity demand event are collected to comprehensively determine the propagation effect. If it is neither a geographical event nor a sudden event, the historical propagation duration of the electricity demand event is collected to determine if the event occurrence duration has reached the historical propagation duration. If not, the difference between the historical propagation duration and the event occurrence duration is calculated. A correlation table is established between the propagation duration difference and the propagation effect, and the propagation effect is obtained from the correlation table. If it is a sudden event, the average speed at which the event's target audience receives information through the information diffusion channels is used as the node propagation speed. The distribution range of the event's target audience is taken as the node propagation range, and the node propagation strength is obtained based on the proportion of the event's target audience at different power nodes. Because the spread of an event takes time, its impact on other power nodes is delayed. By assessing the propagation speed, scope, and intensity of events that cause power changes, the impact on other nodes can be identified, improving the accuracy of forecasts for other power nodes. This also makes forecasting for other nodes simpler and faster, thus increasing the speed of power demand forecasting for distribution networks based on big data analysis. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 The above is a flowchart of a power demand forecasting method for power distribution networks based on big data analysis, provided as an embodiment of the present invention. Detailed Implementation

[0024] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0025] Example 1, referring to Figure 1 This is one embodiment of the present invention, which provides a method for predicting power demand in a distribution network based on big data analysis, including: S1. Divide the power distribution network into multiple power nodes, collect power data from the power nodes, and filter out abnormal power nodes based on the power data; S2. Collect demand information from power anomaly nodes and extract power demand events based on the demand information; S3. Obtain event information of electricity demand events, and evaluate the propagation effect of electricity demand events based on the event information; S4. Collect real-time predicted demand values ​​of power anomaly nodes and determine the power change values ​​of the remaining power nodes in combination with the propagation effect. S5. Collect real-time environmental data of power nodes, predict the basic power demand value based on the real-time environmental data, and superimpose the power change value to obtain the real-time power demand value. It should be noted that in practical applications, real-time environmental data of power nodes is collected. This real-time environmental data includes meteorological conditions, calendar information, etc. Based on existing power demand forecasting technologies for distribution networks, a comprehensive analysis is performed using historical load data and real-time environmental data, among other factors. A basic power demand value is obtained using existing power demand analysis techniques, and then the real-time power demand value is obtained by overlaying power change values ​​at different time points. Existing power demand forecasting data for distribution networks is typically analyzed based on static, conventional factors; however, it is difficult to accurately predict some sporadic or sudden power consumption events.

[0026] Therefore, to address the aforementioned problems, steps S1-S5 are used to select abnormal power nodes to identify occasional power consumption events. Based on these events' propagation effects, the power demand of other nodes can be predicted in a timely manner, reducing the inaccuracy of power forecasting caused by sudden power outages. Because sudden and occasional events are difficult to predict, identifying events requiring power consumption based on abnormal power demand allows for comprehensive forecasting of other nodes in the distribution network, significantly reducing the impact of dynamic events on power forecasting.

[0027] Example 2, refer to Figure 1 This is one embodiment of the present invention, which provides a method for predicting power demand in a distribution network based on big data analysis, including: In this embodiment of the invention, step S1 involves dividing the power distribution network into multiple power nodes, collecting power data from these nodes, and filtering out abnormal power nodes based on the power data. This includes the following steps S11-S12: Step S11: Collect power data from the power nodes. The power data includes voltage, equipment load, and power node frequency.

[0028] In an embodiment of the present invention, step S12, determining whether there is insufficient power supply at a power grid node based on power data, and if there is insufficient power supply, determining the power node as a power anomaly node, includes the following steps A1-A2: A1. A persistently low voltage is one of the most obvious signs of insufficient power supply. If the data shows that the current and power (apparent power) of the transformer or line are consistently close to or exceed the rated value (e.g., the load rate is consistently >80%), it indicates that the equipment is overloaded or overloaded, which also reflects insufficient power supply. The stability of the power grid frequency is the result of the real-time balance between power generation and power consumption. Insufficient power supply will cause the system frequency to be consistently lower than the rated value.

[0029] A2. By using some power data, we can determine whether there is a power shortage at the current power node. The power grid supplies power based on historical power demand or current forecasts. When a power shortage occurs, it means that an event requiring electricity has occurred that was not anticipated.

[0030] In an optional implementation, whether there is insufficient power supply in S12 can be determined based on voltage deviation and equipment temperature. Voltage deviation data of power nodes and temperature data of key equipment are collected. Voltage deviation refers to the continuous deviation between the actual voltage and the rated voltage. Equipment temperature is monitored by sensors to reflect the operating status of the equipment. If the voltage deviation continuously exceeds ±5% of the rated voltage and the equipment temperature continuously exceeds the safety threshold, it is determined that there is insufficient power supply at the power node.

[0031] In another optional implementation, whether there is a power shortage in S12 can also be determined based on load fluctuation and power interruption frequency. Load fluctuation data and power interruption frequency data of the power node are collected. Load fluctuation refers to the load changing drastically in a short period of time. Power interruption frequency refers to the number of power interruptions that occur per unit time. If the load fluctuation rate exceeds 15% of the historical average level and the power interruption frequency has increased significantly in the recent period, it is determined that there is a power shortage in the power node.

[0032] Step S13: If there is no power shortage, the real-time load of the power node is extracted based on the power data.

[0033] In an embodiment of the present invention, step S14, collecting historical load data from power nodes, comparing real-time load with historical load, and determining whether the real-time load is abnormal, includes the following steps B1-B2: B1. Calculate the difference between real-time load and historical load, and determine whether the difference is within the preset standard range; B2. If it is not within the preset standard range, then the real-time load is judged to be abnormal.

[0034] In an optional implementation, in S14, the determination of whether the real-time load is abnormal can be based on the abnormality judgment of the sliding window average value. Historical load data of the power node is collected, and a sliding window is set. The average value of the historical load within the window is calculated as a reference load value. The difference between the real-time load and the above reference load value is calculated, and it is determined whether the difference is within the preset tolerance range. If the difference is not within the tolerance range, the real-time load is determined to be abnormal.

[0035] In another optional implementation, in S14, the determination of whether the real-time load is abnormal can also be based on the abnormality judgment of the historical load distribution range. Historical load data of the power node is collected, and a specific quantile of the historical load is calculated to form a normal load range. The real-time load is directly compared with the normal load range. If the real-time load is lower than the lower limit or higher than the upper limit, the real-time load is determined to be abnormal.

[0036] Step S15: If the real-time load is abnormal, then the power node is determined to be a power abnormal node.

[0037] Most existing technologies rely on historical load data to predict electricity demand because users' electricity consumption behavior usually does not change much. When the load is significantly different from the historical load, it is judged that an event that would not normally occur has occurred, which leads to a significant change in electricity demand.

[0038] Therefore, identifying power anomaly nodes is beneficial for reducing subsequent power demand forecasting errors based on power anomalies.

[0039] For example, a major accident occurs at node A, causing a sudden increase in electricity consumption. Since this major accident was not predicted beforehand, the electricity demand forecast for that node is inaccurate. Now that the accident has occurred, if it spreads, the electricity demand of other nodes will also change. Therefore, determining the specific circumstances and making timely electricity demand forecasts for other nodes is crucial to reduce the likelihood of similar events occurring at other nodes and improve the accuracy of power demand forecasts for the distribution network.

[0040] In this embodiment of the invention, step S2 involves collecting demand information from abnormal power nodes and extracting power demand events based on the demand information, including the following steps S21-S23: Step S21: Collect demand information of power outage nodes. The demand information includes the electrical equipment, the time of power consumption, and the location of the node.

[0041] Electrical equipment refers to the equipment that is currently using electricity at the abnormal node, and the electricity usage time point refers to the time when electricity is being used, i.e., the real-time time point.

[0042] Modern smart meters collect detailed data such as voltage, current, power, and power factor at high frequencies.

[0043] By analyzing the current waveform, it can be determined whether devices with unique power consumption characteristics are working, and the devices currently using electricity can be confirmed through smart meters.

[0044] Step S22: Collect the usage time points corresponding to the electrical equipment and determine whether the power consumption time points match the usage time points.

[0045] By collecting historical usage time points of electrical equipment, the corresponding usage time points are determined based on the frequency of use at different time points. If the frequency of use is roughly the same at each point in time, then the point in time for use is considered to be any time.

[0046] Step S23: If the usage time point is met, the usage quantity of electrical equipment corresponding to the historical electricity usage time point is counted and recorded as the historical usage quantity.

[0047] Step S24: Collect the real-time usage of electrical equipment, calculate the difference between the historical usage and the real-time usage, and obtain the equipment usage difference.

[0048] Step S25: If the usage time does not match, the real-time quantity of electrical equipment is used as the equipment usage difference.

[0049] Step S26: Record devices whose usage difference reaches the preset usage difference threshold as abnormal devices, and combine the abnormal devices and node locations to obtain the power demand event.

[0050] In practical applications, if a device's usage time aligns with a given timeframe—meaning its use at that time is normal—then the change in its usage is statistically analyzed. For example, using a television at 9 PM is normal, but a surge in television usage indicates a correlation between electricity demand and television usage, thus the corresponding electricity demand event is also related to televisions. Conversely, if usage doesn't align with a given timeframe, it means that using the device at that time is abnormal, so it's assumed that no one should be using the device at that time. Therefore, the real-time number of devices in use is used as the device usage difference. Device usage differences are only considered abnormal when they reach a threshold. This threshold can be set based on user experience due to individual differences. If only a very small number of users use a particular device, it indicates that no specific event occurred, and therefore, the change in electricity demand is not considered to be caused by that device.

[0051] The steps for obtaining power demand events by combining abnormal equipment and node locations are as follows: Step S261: Collect the device function of the abnormal device, and determine whether the use of the abnormal device is related to the geographical location based on the device function.

[0052] The use of some devices is related to geographical location. For example, the use of air conditioners is affected by temperature and humidity, which are caused by geographical environment. Therefore, it is believed that the use of air conditioners is related to geographical location.

[0053] Step S262: If the abnormal equipment usage is related to geographical location, collect all events affecting power demand as base events.

[0054] Obtain the incidental events corresponding to historical electricity demand fluctuations, and statistically analyze all incidental events to obtain all events affecting electricity demand.

[0055] Step S263: Filter out events related to geographical location from the basic events and record them as geographical events, and extract the geographical features of the geographical events.

[0056] Step S264: Obtain the location features of the node location, and take the geographical events that match the location features as electricity demand events.

[0057] If the device is geographically related, the corresponding geographical features of the event are statistically analyzed. For example, users typically use heaters for warmth only when the temperature is below 10 degrees Celsius, so the geographical feature corresponding to a heater use event is a temperature below 10 degrees Celsius. Heaters are sometimes also used to dry clothes, so the geographical feature corresponding to drying clothes is high humidity or continuous rain. By matching geographical features, the corresponding electricity demand events are obtained.

[0058] Step S265: If the use of the device is not related to the geographical location, then select events related to the abnormal device from the basic events as abnormal events.

[0059] Step S266: Collect the equipment usage characteristics corresponding to the abnormal event, and confirm the power demand event based on the equipment usage characteristics.

[0060] In practical applications, if the use of equipment is unrelated to geographical location, such as the use of televisions and washing machines, which can be used in any geographical environment, then it is considered that the use of equipment is unrelated to geographical location. In this case, it is impossible to determine the electricity demand event through geographical features.

[0061] Further analysis is needed to determine electricity demand events based on equipment usage characteristics. Confirmation of electricity demand events will help to make more accurate predictions of electricity demand at other nodes based on the event circumstances.

[0062] Different abnormal devices use different methods to confirm electricity demand events, obtain more accurate results, and identify the reasons for changes in user electricity demand.

[0063] The steps for collecting equipment usage characteristics corresponding to abnormal events and confirming electricity demand events based on these characteristics are as follows: Step S2661: Obtain the feature similarity of the device usage features corresponding to all abnormal events, and determine whether the feature similarity is within the preset similarity range.

[0064] Feature similarity is obtained by comparing the cosine similarity model. Feature similarity refers to the usage characteristics of all devices corresponding to all abnormal events, that is, whether the event characteristics that confirm the use of the device are difficult to distinguish.

[0065] For example, when a user uses a television, the corresponding events may include watching a ball game, watching a sports meet, playing a game, etc. However, the usage characteristics corresponding to these events are very similar, such as the duration of use and the time of use.

[0066] By comparing the usage characteristics of these devices, it can be determined whether the events corresponding to the use of the abnormal device can be easily identified.

[0067] Step S2662: If the power impact value corresponding to the abnormal event when it occurred in the past is within the preset similarity range.

[0068] If the similarity is within the preset range, it means that the event corresponding to the use of the device is difficult to distinguish. Then, the power load corresponding to the abnormal event in the historical process is statistically analyzed, and the power impact value is obtained by calculating the difference between the power load and the power load when there is no such event and all other conditions are the same.

[0069] Step S2663: Obtain the real-time change value of the power at the abnormal power node, calculate the difference between the power impact value and the real-time change value, collect the probability of the occurrence of the abnormal event, and comprehensively determine the power demand event.

[0070] The difference between the real-time power load and the historical average power load is used as the real-time change value to calculate the power impact value and the difference between the real-time change value. Set the proportional coefficients for the difference and the probability of occurrence, respectively. Then, sum the products of the difference and the probability of occurrence with their respective proportional coefficients to obtain the abnormal value of the abnormal event.

[0071] For example, if the difference and the probability of occurrence are 200 and 80% respectively, and the corresponding scaling factors are set to 0.4 and 0.6 respectively, then the corresponding outlier is 200×0.4+80%×0.6=80.48.

[0072] The event with the largest outlier is selected as the electricity demand event. The smaller the difference, the larger the outlier, indicating that the electricity demand of the event is more similar to and closer to the existing electricity demand. The higher the probability of the event occurring, the more likely the change in existing electricity demand is to be caused by the current event, and therefore the larger the corresponding outlier.

[0073] Step S2664: If the similarity is not within the preset range, extract the device operation features of the abnormal device, compare the similarity between the device operation features and the device usage features, and record it as the operation similarity.

[0074] Step S2665: Select the abnormal event with the highest similarity as the electricity demand event.

[0075] If it is not within the preset similarity range, it means that the event characteristics of using the device are different. In this case, it is only necessary to collect the operating characteristics of the abnormal device, compare them through the cosine similarity model, and select the event with the highest similarity. Because its operating characteristics match the characteristics corresponding to the event, it can be confirmed that the power demand was caused by this abnormal event.

[0076] In this embodiment of the invention, step S3 involves acquiring event information of an electricity demand event and evaluating the propagation effect of the electricity demand event based on the event information, including the following steps S31-S2: Step S31: Determine whether the electricity demand event is a geographical event. If it is a geographical event, determine whether the electricity demand event has spread.

[0077] Step S32: If the electricity demand event spreads, determine whether the occurrence of the electricity demand event depends on geographical features.

[0078] Different electricity demand events have different characteristics, so whether they will spread also varies.

[0079] For example, if the electricity demand event is due to excessively high temperature, the temperature in that area will not spread. If the electricity demand event does not spread, it is considered that there is no propagation effect, and other nodes will not be affected by the event of that node.

[0080] If a large number of range hoods are used in area A, it is because a fire has occurred, and the fire will spread, causing the incident to escalate.

[0081] While events of excessively high regional temperatures depend on geographical features for them to occur, fires do not depend on geographical features. They are only related to location because they occur in that region, and are therefore classified as geographical events.

[0082] Step S33: If relying on geographical features, obtain the location features of other power nodes, compare the similarity between geographical features and location features, and record it as geographical similarity.

[0083] Step S34: The range of power nodes whose geographical similarity reaches a preset geographical similarity threshold is taken as the node propagation range.

[0084] If it relies on geographical features, the same event may occur in places with similar geographical features. Therefore, the propagation range of nodes can be determined based on the similarity of geographical features. The geographical similarity threshold can be set by the user based on experience.

[0085] Step S35: Obtain the node propagation speed and node propagation strength of the power nodes based on geographical similarity.

[0086] Establish a correlation curve between geographical similarity and node propagation speed, and establish a correlation curve between geographical similarity and node propagation strength. Find the corresponding node propagation speed and node propagation strength through the corresponding correlation curves. The greater the geographical similarity, the greater the probability that the same event will occur at that node, so the node will spread faster and the corresponding node will spread more powerfully.

[0087] For example, floods are more common in low-lying areas, floodplains, and urban areas with weak drainage systems. Their occurrence depends on the amount of heavy rainfall in the area, and due to the fluidity of water flow, other low-lying areas, floodplains, and areas with weak drainage systems are also prone to this event.

[0088] The more similar the geographical features, the greater the probability of flooding, and the greater the severity and speed of the flooding.

[0089] For example, if flooding occurs in area A, and area B has very similar geographical features to area A, then the degree of flooding in area B will be similar to that in area A. However, area C has fewer low-lying areas, so its similarity is not high, and it is clear that even if flooding occurs in area C, the degree of flooding in area B will be lower.

[0090] Step S36: If not dependent on geographical features, collect historical diffusion data of electricity demand events and obtain the propagation effect based on the historical diffusion data.

[0091] Step S37: If it is not a geographical event, the propagation effect of the electricity demand event is assessed based on the event information.

[0092] When the occurrence of electricity demand events depends on geographical features and the events are prone to spread, areas with similar geographical features are more likely to be affected by the events, so their spread will be faster. For regions with different geographical characteristics, since the occurrence of electricity demand events depends on geographical characteristics, these regions do not have the corresponding geographical characteristics, so the events cannot spread to these regions. Therefore, the electricity demand events do not affect the electricity demand in these regions, and the electricity demand forecast for these regions does not need to take the events into account.

[0093] If electricity demand events do not depend on geographical features, and spread to areas with different geographical features, further analysis and confirmation based on historical diffusion data are needed.

[0094] The steps for collecting historical diffusion data of electricity demand events and obtaining the propagation effect based on the historical diffusion data are as follows: Step S361: Collect historical diffusion data of electricity demand events. The historical diffusion data includes historical diffusion range, historical diffusion speed, and historical diffusion intensity.

[0095] If it does not rely on geographical features, it can be directly spread. For example, in the spread of fire, based on historical fire situations, the average spread range, spread speed and spread intensity of the fire can be statistically analyzed to obtain historical spread data.

[0096] The diffusion intensity refers to the extent to which the event spreads to other areas. Because different events spread in different ways, the diffusion of some events exhibits a decaying trend. For example, the impact of some faults spreads less as the distance from the fault increases, so the diffusion intensity is smaller.

[0097] Step S362: Determine whether the message diffusion of the electricity demand event affects the electricity demand. If it does, collect the message diffusion data of the electricity demand event. The message diffusion data includes the message diffusion range, message diffusion speed, and message diffusion strength.

[0098] Step S363: Combine historical diffusion data and message diffusion data to obtain the propagation effect.

[0099] The electricity demand event itself does not spread, but the spread of its information can still affect changes in electricity demand. For example, in the case of a natural disaster, although it only occurs in location A, when users in other areas become aware of it, they will make preparations in advance out of concern for the spread of the natural disaster, such as leaving lights on for extended periods of time.

[0100] The spread of information can also affect electricity demand. The scope, speed, and intensity of information spread refer to the extent, speed, and intensity of the electricity impact caused by the spread of information.

[0101] For example, after a message spreads naturally, users directly affected by the event will turn on devices such as air conditioners, while those affected by the message will leave lights on for extended periods. The electricity consumed by turning on the lights represents the extent of the message's spread.

[0102] By superimposing the diffusion range, diffusion speed, and diffusion strength from historical diffusion data and message diffusion data one by one, we can obtain the node propagation range, node propagation speed, and node propagation strength.

[0103] If it is not a geographical event, the steps for assessing the propagation effect of the electricity demand event based on event information are as follows: Step S371: If it is not a geographical event, collect the event information of the electricity demand event. The event information includes the duration of the event and the people who are interested in the event.

[0104] Step S372: Determine whether the electricity demand event is a sudden event. If it is not a sudden event, collect the historical propagation time of the electricity demand event and determine whether the event duration has reached the historical propagation time.

[0105] Some electricity demand events are sudden events, such as natural disasters, malfunctions, sudden weather changes, etc., while other electricity demand events are not sudden events, such as watching a ball game, the time of the game is announced in advance, so it is not a sudden event.

[0106] The historical dissemination time of ball games is calculated, that is, the time from the announcement of the ball game to the awareness of the target audience. The time spent watching the ball game is calculated from the time the ball game was announced.

[0107] Step S373: If the historical propagation duration has not been reached, calculate the propagation duration difference between the historical propagation duration and the duration of the event.

[0108] Step S374: Establish a correlation table between the difference in propagation duration and the propagation effect, and find the propagation effect based on the correlation table.

[0109] Step S375: If the electricity demand event is a sudden event, collect the information dissemination channels of the electricity demand event and combine them with the dissemination effect of the people concerned about the event.

[0110] If the event is not a sudden event, it is an occasional event. Although such events are not routine events, they have already undergone a publicity period before they occur. If the event has reached the historical dissemination time, it means that the event has been disseminated to a considerable extent, and it is believed that it will not spread further.

[0111] For example, the publicity for a sports match is almost complete within 5 days, and there won't be any additional viewers afterward. Therefore, the actual number of viewers is basically determined before the match even starts, so the event is considered to have stopped spreading due to the dissemination of the news, and the electricity demand is no longer affected.

[0112] If the historical dissemination duration has not been reached, it means that the event is still spreading. Taking a ball game as an example, as shown in Table 1, a relevant table can be created to find the corresponding dissemination effect. In practice, the table will contain specific data based on the actual situation.

[0113] Table 1. Transmission Effects

[0114] The steps for collecting information on electricity demand events and understanding their dissemination channels, combined with the dissemination effect among the target audience, are as follows: Step S3751: The average speed at which the public concerned about the event receives information in the statistical information dissemination channel is taken as the node propagation speed.

[0115] Step S3752: Extract the characteristics of the people who are interested in the event based on the characteristics of the people who are interested in the event, and determine the distribution range of the people who are interested in the event based on the characteristics of the people, which is used as the node propagation range.

[0116] Step S3753: Calculate the percentage of people who are interested in the event at different power nodes, as a measure of the node's dissemination strength.

[0117] In practical applications, when an electricity demand event is sudden, its propagation can occur immediately, without a pre-propagation period. Furthermore, because different events attract different audiences, the electricity demand of these audiences only changes during the propagation process due to the event's occurrence. Therefore, the propagation effect can be determined based on the characteristics of the audience interested in the event.

[0118] In this embodiment of the invention, step S4 involves collecting real-time predicted demand values ​​of power anomaly nodes and determining the power change values ​​of the remaining power nodes based on propagation effects. This includes the following steps S41-S43: Step S41: Record the power nodes within the node propagation range as power prediction nodes, and calculate the distance between the power anomaly node and the power prediction node.

[0119] Step S42: Calculate the time point of power change based on the node propagation speed and distance, and collect the power anomaly value of the power anomaly node.

[0120] Step S43: Obtain the real-time time point, calculate the time difference between the power change time point and the real-time time point, and obtain the real-time predicted demand value based on the time difference.

[0121] In an embodiment of the present invention, step S44, when the real-time predicted demand value reaches a preset demand value threshold, calculates the power change value of the power prediction node at the power change time point based on the power anomaly value and the node propagation strength, including the following steps C1-C4: C1. The propagation of electricity demand events takes time. Therefore, when an anomaly occurs at a power anomaly node, the changes in electricity demand at other nodes do not occur simultaneously, but are delayed. Whether it is necessary to predict the changes in electricity demand at other nodes caused by the electricity demand event depends on whether the time is close.

[0122] C2. As the timing of an event approaches and the resulting changes in electricity demand become more imminent, forecasting electricity demand becomes even more crucial. Premature forecasting can actually negatively impact the accuracy of electricity demand predictions.

[0123] C3. Multiply the node propagation strength and the power anomaly value to obtain the power change value of the power prediction node. For example, if the node propagation strength and the power anomaly value are 50% and 90 respectively, then the power change value is 90 × 50% = 45. Because some changes in power demand decay as the event propagates, the strength naturally decreases, and the power change value is obviously different from the power anomaly value.

[0124] C4. Based on the propagation speed and the actual distance between nodes, the actual time when the event propagation affects the power prediction point is determined. This allows for a more accurate prediction of the power demand changes at the corresponding time points. The demand threshold can be set by the user based on the actual situation.

[0125] In an optional implementation, the power change value in S44 can be adjusted based on a weighted adjustment of geographical similarity. The geographical similarity between the power prediction node and the power anomaly node is calculated, and the geographical similarity is used as a weighting coefficient to adjust the node propagation strength. The weighted node propagation strength is then multiplied by the power anomaly value to obtain the power change value. However, the weighting of geographical similarity in this implementation depends on human experience, which can easily introduce subjective errors in complex environments.

[0126] In another optional implementation, the power change value in S44 can also be estimated based on empirical estimation by analogy with historical events. Historical events similar to the current power demand event are retrieved from the historical database, and the power change value recorded at a similar power prediction node for the historical event is obtained as a reference value. The reference value is then fine-tuned based on the slight differences between the current event and the historical event to obtain the power change value. However, this implementation is less effective for estimating rare or new events.

[0127] In an embodiment of the present invention, S5 collects real-time environmental data of the power node, predicts the basic power demand value based on the real-time environmental data, and superimposes the power change value to obtain the real-time power demand value.

[0128] In an optional implementation, the real-time power demand value obtained in S5 can be based on fixed rules and environmental thresholds to predict the basic power demand value. Real-time environmental data of power nodes is collected, but only key environmental factors are considered, and environmental thresholds are preset. Based on the comparison between environmental data and thresholds, the basic power demand value is directly calculated using a predefined rule table or linear interpolation method. The basic power demand value is then superimposed with the power change value obtained from the propagation effect to obtain the real-time power demand value. However, this implementation simplifies the influence of environmental factors and cannot handle complex nonlinear relationships, resulting in a large prediction deviation in variable environments.

[0129] Example 3 is an embodiment of the present invention. The above is an illustrative scheme of a power demand forecasting method for distribution networks based on big data analysis. It should be noted that the technical solution of a power demand forecasting device and medium for distribution networks based on big data analysis belongs to the same concept as the technical solution of the power demand forecasting method for distribution networks based on big data analysis described above. For details not described in detail in the technical solution of the power demand forecasting device and medium for distribution networks based on big data analysis in this embodiment, please refer to the description of the technical solution of the power demand forecasting method for distribution networks based on big data analysis described above.

[0130] This embodiment also provides an electronic device applicable to a power demand forecasting method for power distribution networks based on big data analysis, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the power demand forecasting method for power distribution networks based on big data analysis as proposed in the above embodiment.

[0131] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements a power demand forecasting method for power distribution networks based on big data analysis as proposed in the above embodiments.

[0132] The storage medium proposed in this embodiment belongs to the same inventive concept as the method for predicting power demand in a distribution network based on big data analysis proposed in the above embodiments. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0133] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0134] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for forecasting power demand in a distribution network based on big data analysis, characterized in that: include, The power distribution network is divided into multiple power nodes, power data of the power nodes is collected, and power anomaly nodes are identified based on the power data. Collect demand information from nodes with abnormal power supply, and extract electricity demand events based on the demand information; Obtain event information on electricity demand events, and assess the propagation effect of electricity demand events based on the event information; Collect real-time predicted demand values ​​of power anomaly nodes and determine the power change values ​​of the remaining power nodes based on the propagation effect. Real-time environmental data of power nodes is collected, and the basic power demand value is predicted based on the real-time environmental data. The real-time power demand value is obtained by superimposing the power change value.

2. The power demand forecasting method for distribution networks based on big data analysis as described in claim 1, characterized in that: The filtered nodes with abnormal power include, Collect power data from power nodes, and determine whether there is insufficient power supply at the power grid nodes based on the power data. If there is insufficient power supply, the power node is determined to be a power abnormal node. If there is no power shortage, the real-time load of the power node is extracted based on the power data; Collect historical load data for power nodes, compare the real-time load with the historical load, and determine whether the real-time load is abnormal. If the real-time load is abnormal, the power node is determined to be an abnormal power node.

3. The power demand forecasting method for distribution networks based on big data analysis as described in claim 2, characterized in that: The process of obtaining electricity demand events includes collecting demand information from abnormal power nodes, collecting the usage time points corresponding to electrical equipment, and determining whether the electricity demand time points meet the usage time points. If the usage time point is consistent, the usage quantity of electrical equipment corresponding to the historical electricity usage time point is counted and recorded as the historical usage quantity. The real-time usage quantity of electrical equipment is collected, and the difference between the historical usage quantity and the real-time usage quantity is calculated to obtain the equipment usage difference. If the usage time does not meet the requirements, the real-time number of electrical devices will be used as the device usage difference. Devices whose device usage difference reaches the preset usage difference threshold will be recorded as abnormal devices. The power demand event will be obtained by combining the abnormal devices and the node location.

4. The power demand forecasting method for distribution networks based on big data analysis as described in claim 3, characterized in that: The process of combining abnormal equipment and node location to obtain electricity demand events includes collecting the equipment efficacy of abnormal equipment and determining whether the use of abnormal equipment is related to geographical location based on the equipment efficacy. If the abnormal equipment usage is related to geographical location, then all events affecting power demand are collected as base events; Select geographically related events from the basic events and record them as geographical events. Extract the geographical features of the geographical events and obtain the location features of the nodes. Geographical events with consistent geographical features and location features are used as electricity demand events. If the use of equipment is not related to geographical location, then events related to abnormal equipment are selected from the basic events as abnormal events; the equipment usage characteristics corresponding to the abnormal events are collected, and the power demand events are confirmed based on the equipment usage characteristics.

5. The power demand forecasting method for distribution networks based on big data analysis as described in claim 4, characterized in that: The step of confirming the electricity demand event based on the equipment usage characteristics includes obtaining the feature similarity of the equipment usage characteristics corresponding to all abnormal events, and determining whether the feature similarity is within a preset similarity range; If the similarity is within a preset range, collect the power impact values ​​corresponding to the abnormal events when they occurred in the past; The system acquires real-time power change values ​​of abnormal power nodes, calculates the difference between the power impact value and the real-time change value, collects the probability of abnormal events, and comprehensively determines the power demand events. If it is not within the preset similarity range, the device operation characteristics of the abnormal device are extracted, and the similarity between the device operation characteristics and the device usage characteristics is compared and recorded as the operation similarity. Select the abnormal event with the highest similarity to the actual event as the electricity demand event.

6. The power demand forecasting method for distribution networks based on big data analysis as described in claim 5, characterized in that: The process of obtaining the propagation effect of electricity demand events includes determining whether the electricity demand event is a geographical event. If it is a geographical event, then determine whether the electricity demand event has spread; If electricity demand events spread, determine whether the occurrence of electricity demand events depends on geographical features; If relying on geographical features, the location features of the remaining power nodes are obtained, and the similarity between the geographical features and location features is compared and recorded as geographical similarity. The range of power nodes whose geographical similarity reaches a preset geographical similarity threshold is taken as the node propagation range. The node propagation speed and node propagation strength of the power nodes are obtained based on the geographical similarity. If geographical features are not relied upon, historical diffusion data of electricity demand events are collected, and the propagation effect is obtained based on the historical diffusion data; If it is not a geographical event, the propagation effect of the electricity demand event can be assessed based on the event information.

7. The power demand forecasting method for distribution networks based on big data analysis as described in claim 6, characterized in that: The process of obtaining the propagation effect based on historical diffusion data includes collecting historical diffusion data of electricity demand events, including historical diffusion range, historical diffusion speed, and historical diffusion intensity. Determine whether the message spread of an electricity demand event affects electricity demand. If it does, collect message spread data of the electricity demand event. The message spread data includes message spread range, message spread speed, and message spread intensity. The propagation effect was obtained by combining historical diffusion data and message diffusion data. The assessment of the propagation effect of electricity demand events based on event information includes, If it is not a geographical event, then the event information of the electricity demand event is collected, including the duration of the event and the people who are interested in the event; Determine whether the electricity demand event is a sudden event. If it is not a sudden event, collect the historical propagation time of the electricity demand event and determine whether the duration of the event has reached the historical propagation time. If the historical propagation duration is not reached, the difference between the historical propagation duration and the duration of the event is calculated. Establish a correlation table between the difference in propagation duration and the propagation effect, and find the propagation effect based on the correlation table; If the electricity demand event is a sudden event, the information dissemination channels of the electricity demand event are collected, and the dissemination effect is combined with the information of the people concerned about the event. The aforementioned situation regarding the dissemination effect obtained by combining the event's target audience includes using the average speed at which the event's target audience receives information in the statistical information dissemination channels as the node dissemination speed. Based on the characteristics of the people who are interested in the event, we can determine their distribution range and use this range as the propagation range of the nodes. The percentage of people who pay attention to events at different power nodes is used as a measure of the node's dissemination strength.

8. The power demand forecasting method for distribution networks based on big data analysis as described in claim 7, characterized in that: The process of determining the power change value of the remaining power nodes includes recording power nodes within the node propagation range as power prediction nodes and calculating the distance between power anomaly nodes and power prediction nodes. The time points of power change are calculated based on the node propagation speed and distance, and the power anomaly values ​​of nodes with power anomalies are collected. Obtain real-time time points, calculate the time difference between the power change time points and the real-time time points, and obtain the real-time predicted demand value based on the time difference; When the real-time predicted demand value reaches the preset demand value threshold, the power change value of the power prediction node at the power change time point is calculated based on the power anomaly value and the node propagation strength.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the power demand forecasting method for distribution networks based on big data analysis as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the power demand forecasting method for distribution networks based on big data analysis as described in any one of claims 1 to 8.

Images