Interactive power grid dispatching feedback method and system
By establishing a user behavior prediction model and a dynamic feedback adjustment database, and combining machine learning algorithms to optimize power scheduling, the problem of traditional power grid scheduling methods insufficient response to real-time data and user behavior changes is solved, and the stable operation and efficient scheduling of the power grid are achieved.
Patent Information
- Application Number
- CN202411475753.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-22
- Publication Date
- 2024-12-27
AI Technical Summary
Traditional grid scheduling methods lack agile response to real-time data and changes in user behavior, leading to problems such as supply and demand imbalance and energy waste.
By establishing a user behavior prediction model and a dynamic feedback adjustment database, combining machine learning algorithms for online learning and reinforcement learning, capturing user power consumption patterns and environment changes, optimizing power scheduling, and setting up a distributed dynamic feedback adjustment database to deal with emergencies.
It realizes more accurate power consumption mode capture and scheduling decisions, quickly responds to abnormal loads, improves the risk resistance and operational efficiency of the power grid, and reduces operating costs.
Smart Images
Figure CN119203778A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of electric power dispatching, and in particular, relates to an interactive power grid dispatching feedback method and system. Background Art
[0002] With the widespread access to renewable energy and the diversified development of user-side resources, the supply and demand relationship of modern power systems has become increasingly complex. The traditional power grid dispatching method mainly relies on pre-set plans, lacks full utilization of real-time data and agile response to changes in user behavior. For example, the invention patent with announcement number CN103683498B discloses a power grid dispatching system, including a dispatching center, several control stations, and multiple measurement and control terminals; the dispatching center includes a fault detection unit, a real-time load calculation unit, a short-term load forecasting unit, a backup power supply control unit, a reactive power compensation unit, and a message unit, and each unit is connected through a communication bus; the dispatching center, the control station, and the measurement and control terminal jointly complete the dispatching of the power grid, and the dispatching center is the core. The units of the dispatching center communicate with each other and work together to complete the dispatching task of the entire power grid.
[0003] This static scheduling method is unable to cope with the rapidly changing and diverse user demands, and can easily lead to problems such as supply and demand imbalance and energy waste.
[0004] In view of this, this application is hereby filed. Summary of the invention
[0005] The technical problem to be solved by the present invention is to overcome the deficiencies of the prior art and provide an interactive power grid dispatching feedback method and system. The present invention is achieved through the following technical solutions:
[0006] An interactive power grid dispatch feedback method, the method comprising:
[0007] Establish a user behavior prediction model to capture the user's electricity usage pattern and the changing trend of the user's electricity usage pattern over time;
[0008] Setting up a dynamic feedback regulation database based on the user behavior prediction model;
[0009] When a corresponding abnormal load is detected, the dynamic feedback regulation database is called accordingly.
[0010] Preferably, the method further comprises:
[0011] Set up machine learning algorithms to continuously optimize user behavior prediction models and dynamic feedback regulation databases.
[0012] Preferably, the machine learning algorithm comprises:
[0013] Set up an online learning framework to accumulate learning data and strengthen the training model through incremental learning and / or transfer learning and / or federated learning;
[0014] Reinforcement learning applications are used to conduct environmental modeling, build a simulation environment for power grid operation, simulate different operation scenarios and results, introduce strategy exploration algorithms, and allow the model to synchronously trial and error in the simulation environment during power grid operation to generate different operation strategies and sort them according to their priorities.
[0015] Preferably, establishing a user behavior prediction model includes:
[0016] Multi-source data fusion analysis, based on information feedback from multiple sources, evaluates the impact of information from each source on electricity demand, optimizes user behavior prediction models, and adjusts power dispatch according to the impact on electricity demand.
[0017] Preferably, the multi-source data includes:
[0018] User feedback sentiment analysis: Through the natural language processing model, we analyze the user's information feedback on various power consumption platforms, extract the user's demand for their own power consumption, and judge the impact on power dispatch based on the demand;
[0019] Life data association: combining city event announcements and traffic flow data to predict the impact of large-scale events, emergencies and traffic conditions on power grid load;
[0020] Environmental monitoring association: Introduce weather conditions, air quality index and noise pollution data into the user behavior prediction model, and judge the impact of environmental information on residents' electricity consumption behavior based on changes in various environmental information.
[0021] Preferably, establishing a user behavior prediction model includes:
[0022] Set up privacy protection to erase personal identification information after data collection and only retain information related to electricity usage behavior;
[0023] The collected data is homomorphically encrypted and calculations are performed directly on the encrypted data.
[0024] Preferably, the dynamic feedback regulation database is provided with a mesh division, and corresponding dynamic feedback regulation databases are provided at different power consumption nodes.
[0025] The present invention also provides an interactive power grid dispatching feedback system, and the interactive power grid dispatching feedback system is applied with the interactive power grid dispatching feedback method as described above.
[0026] Compared with the prior art, the present invention has the following beneficial effects:
[0027] 1. By establishing a user behavior prediction model, it is possible to more accurately capture the user's electricity consumption pattern and its changing trend, provide a scientific basis for subsequent scheduling decisions, and quickly call on preset response measures when abnormal loads are detected, effectively preventing the power grid from being overloaded or underloaded, and ensuring the stable operation of the power grid.
[0028] 2. Through the continuous optimization of machine learning algorithms, the model can continuously adapt to new power consumption patterns and external environmental changes, maintain the accuracy of predictions, and the continuous optimization of the dynamic feedback adjustment database enables the system to respond to various emergencies more flexibly, improve the risk resistance of the power grid, and reduce operating costs.
[0029] 3. The online learning framework allows the system to update the model in real time without stopping, ensuring that the model is always in the best state. Through transfer learning and federated learning, the model can borrow knowledge from other tasks or user groups to improve the generalization and robustness of the model. The application of reinforcement learning enables the system to continuously trial and error in a simulated environment, generate and optimize operating strategies, and improve the quality and efficiency of scheduling decisions. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION
[0031] The present invention will be further described below in conjunction with the accompanying drawings.
[0032] Embodiment 1
[0033] like Figure 1 As shown, this embodiment provides an interactive power grid dispatch feedback method, the method comprising:
[0034] Establish a user behavior prediction model to capture the user's electricity usage pattern and the changing trend of the user's electricity usage pattern over time;
[0035] Setting up a dynamic feedback regulation database based on the user behavior prediction model;
[0036] When a corresponding abnormal load is detected, the dynamic feedback regulation database is called accordingly.
[0037] By establishing a user behavior prediction model, we can more accurately capture the user's electricity consumption pattern and its changing trends, provide a scientific basis for subsequent scheduling decisions, and quickly call on preset response measures when abnormal loads are detected, effectively preventing the power grid from being overloaded or underloaded, and ensuring the stable operation of the power grid.
[0038] Preferably, the method further comprises:
[0039] Set up machine learning algorithms to continuously optimize the user behavior prediction model and dynamic feedback adjustment database. Through the continuous optimization of the machine learning algorithm, the model can continuously adapt to new power consumption patterns and external environmental changes, maintain the accuracy of the prediction, and the continuous optimization of the dynamic feedback adjustment database enables the system to respond to various emergencies more flexibly, improve the risk resistance of the power grid, and reduce operating costs.
[0040] Preferably, the machine learning algorithm comprises:
[0041] Set up an online learning framework to accumulate learning data and strengthen the training model through incremental learning and / or transfer learning and / or federated learning;
[0042] Reinforcement learning applications are used to conduct environmental modeling, build a simulation environment for power grid operation, simulate different operation scenarios and results, introduce strategy exploration algorithms, and allow the model to synchronously trial and error in the simulation environment during power grid operation to generate different operation strategies and sort them according to their priorities.
[0043] Specifically, incremental learning only updates part of the model's weights each time new data is received, avoiding retraining the entire model. Transfer learning can use the knowledge of existing models to migrate to new tasks, accelerating the convergence of new models. Federated learning allows multiple participants to collaboratively train models without sharing original data, protecting user privacy. A further specific setting of federated learning can be that the power system terminal of grassroots power workers provides data and improvement references for federated learning. If it is a power system terminal with higher authority, it can also directly modify the preliminary parameters in the dynamic feedback adjustment database to be executed.
[0044] The online learning framework allows the system to update the model in real time without stopping, ensuring that the model is always in the best state. Through transfer learning and federated learning, the model can borrow knowledge from other tasks or user groups to improve the generalization and robustness of the model. The application of reinforcement learning enables the system to continuously trial and error in a simulated environment, generate and optimize operating strategies, and improve the quality and efficiency of scheduling decisions.
[0045] Further, model training can follow the following three steps:
[0046] Baseline model: First, use simple models such as linear regression and decision tree as a baseline to evaluate basic prediction effects.
[0047] Advanced models: Introduce complex models such as random forest, support vector machine, gradient boosted tree (GBDT), long short-term memory network (LSTM), and use cross-validation to determine the best hyperparameter combination.
[0048] Model fusion: Use ensemble learning methods, such as bagging and boosting, to combine the advantages of multiple models to improve the robustness and accuracy of predictions.
[0049] Then the model can be evaluated and optimized, for example:
[0050] Error analysis: The prediction performance of the model is evaluated through indicators such as MAE and MSE, with a focus on prediction errors during special periods such as holidays and extreme weather.
[0051] Sensitivity analysis: Test the model’s sensitivity to different input features to identify the key factors that affect the prediction results.
[0052] Real-time correction: Set up an online learning module to dynamically adjust model parameters based on the latest data to maintain the timeliness of the model.
[0053] Preferably, establishing a user behavior prediction model includes:
[0054] Multi-source data fusion analysis, based on information feedback from multiple sources, evaluates the impact of information from each source on electricity demand, optimizes user behavior prediction models, and adjusts power dispatch according to the impact on electricity demand.
[0055] Preferably, the multi-source data includes:
[0056] User feedback sentiment analysis: Through the natural language processing model, we analyze the user's information feedback on various power consumption platforms, extract the user's demand for their own power consumption, and judge the impact on power dispatch based on the demand;
[0057] Life data association: combining city event announcements and traffic flow data to predict the impact of large-scale events, emergencies and traffic conditions on power grid load;
[0058] Environmental monitoring association: Introduce weather conditions, air quality index and noise pollution data into the user behavior prediction model, and judge the impact of environmental information on residents' electricity consumption behavior based on changes in various environmental information.
[0059] Furthermore, data collection includes data collection and preprocessing and feature engineering analysis, where data collection and preprocessing include:
[0060] Historical electricity consumption data: Obtain detailed electricity consumption records of users from smart meters, power grid management systems and other channels, including daily and hourly electricity consumption, electricity consumption type (such as industrial, commercial, residential), etc.
[0061] Socio-economic indicators: Collect macroeconomic data closely related to electricity consumption, such as GDP growth rate, population density, per capita income level, etc.
[0062] Weather forecast: Integrate meteorological API to obtain past and future weather data, including temperature, humidity, wind speed, etc.
[0063] Holiday arrangement: Organize information such as national statutory holidays and local special holidays so that the model can distinguish the difference in electricity consumption between weekdays and weekends.
[0064] Feature engineering includes:
[0065] Time series features: Extract features of time dimensions such as date, day of the week, and season.
[0066] Weather characteristics: Convert weather variables such as temperature and humidity into features that have a direct impact on electricity consumption.
[0067] Economic characteristics: Calculate the changing trends of economic indicators such as economic growth rate and inflation rate.
[0068] Social behavior characteristics: Analyze the discussion heat about electricity consumption on social media and understand the public's attention to changes in electricity prices.
[0069] Preferably, establishing a user behavior prediction model includes:
[0070] Set up privacy protection to erase personal identification information after data collection and only retain information related to electricity usage behavior;
[0071] The collected data is homomorphically encrypted and calculations are performed directly on the encrypted data.
[0072] Preferably, the dynamic feedback regulation database is provided with a mesh division, and corresponding dynamic feedback regulation databases are provided at different power consumption nodes.
[0073] The mesh-distributed dynamic feedback regulation database enables the system to manage each power consumption node more finely, improving the flexibility and response speed of scheduling. Through distributed management, the system can allocate computing resources more reasonably, avoid single point overload, and improve the overall performance of the system. The distributed architecture enhances the fault tolerance of the system. Even if a node fails, it will not affect the normal operation of the entire system.
[0074] Furthermore, a dynamic feedback regulation database is directly set up at the corresponding grassroots power nodes, which can be directly managed and executed by local administrators to avoid network congestion or the impact on power dispatching due to untimely signal feedback.
[0075] Specifically, simulation or actual data analysis examples can be introduced:
[0076] 1. Historical electricity consumption data:
[0077] The electricity consumption type is residential electricity consumption: the hourly electricity consumption of residents in a certain place from October 1 to October 7 of a certain year is shown in the following table (partial data):
[0078] date time Electricity consumption (kWh) 10-01 00:00 0.5 10-01 01:00 0.6 ... ... ... 10-07 23:00 1.2
[0079] The types of electricity consumption are industrial, commercial, and residential: The distribution of electricity consumption types in a certain area from October 1 to October 7 of a certain year is shown in the following table:
[0080] date Industrial electricity consumption (kWh) Commercial electricity (kWh) Residential electricity consumption (kWh) 10-01 1200 800 500 10-02 1100 850 550 ... ... ... ... 10-07 1300 900 600
[0081] 2. Socio-economic indicators:
[0082] The data can be obtained directly from official websites such as the National Bureau of Statistics and local statistical bureaus.
[0083] GDP growth rate: quarterly growth rate (unit: %)
[0084] For example, the GDP growth rate of a region from the first quarter to the third quarter of a certain year is shown in the following table:
[0085]
[0086]
[0087] Population density: the number of people per square kilometer (unit: person / km 2 )
[0088] For example, the monthly population density of a city in a certain year is shown in the following table:
[0089] month <![CDATA[Population density (persons / km 2 )]]> 01 1200 02 1210 ... ... 12 1250
[0090] Per capita income level: Annual per capita income (unit: yuan)
[0091] For example, the monthly per capita income of a region in a certain year is shown in the following table:
[0092] month Per capita income (yuan) 01 5000 02 5100 ... ... 12 5500
[0093] 3. Weather forecast information:
[0094] Data source: Meteorological Bureau API
[0095] Specific parameters:
[0096] Temperature: Daily maximum and minimum temperatures (unit: °C)
[0097] For example, the temperature in a certain area from October 1 to October 7 of a certain year is shown in the following table:
[0098]
[0099]
[0100] Humidity: Daily average humidity (unit: %)
[0101] For example, the humidity in a certain area from October 1 to October 7 of a certain year is shown in the following table:
[0102] date Average humidity (%) 10-01 70 10-02 72 ... ... 10-07 75
[0103] Wind speed: daily average wind speed (unit: m / s)
[0104] For example, the wind speed in a certain area from October 1 to October 7 of a certain year is shown in the following table:
[0105] date Average wind speed (m / s) 10-01 2.5 10-02 2.7 ... ... 10-07 2.8
[0106] 4. Holiday arrangements
[0107] Data source: Government announcement
[0108] Specific parameters:
[0109] Statutory Holidays: Date
[0110] For example, the statutory holidays for a certain year are shown in the following table:
[0111] Holidays date National Day 10-01 Mid-Autumn Festival 09-29
[0112] Furthermore, the characteristic festivals of a certain region are shown in the following table:
[0113]
[0114] The above data can also contain more information and more information types, which will not be elaborated here. For the above information, the following conclusions can be drawn based on the data source processing:
[0115] 1. Analysis of historical electricity consumption data
[0116] Electricity consumption trends:
[0117] From October 1 to October 7 of a certain year, the residential user's electricity consumption was lower at night (00:00-06:00) and higher during the day (07:00-22:00), especially during the evening peak period (18:00-22:00), when electricity consumption increased significantly.
[0118] Industrial electricity consumption in the region is higher on weekdays (Monday to Friday) and lower on weekends (Saturday and Sunday); residential electricity consumption is higher on weekends.
[0119] 2. Analysis of social and economic indicators
[0120] GDP Growth Rate:
[0121] From the first quarter to the third quarter of a certain year, the region's GDP growth rate increased quarter by quarter, indicating that economic activities were gradually becoming more active, which might drive the growth of electricity demand.
[0122] Population Density:
[0123] In a certain year, the city's population density increased month by month, indicating that the urbanization process was accelerating, which might lead to an increase in electricity demand.
[0124] Per capita income level:
[0125] In a certain year, the per capita income of the region increased month by month, indicating that residents' consumption capacity increased, which may increase the frequency of use of household appliances and thus increase electricity consumption.
[0126] 3. Weather forecast analysis
[0127] Relationship between temperature and power consumption:
[0128] From October 1 to October 7 of a certain year, when the temperature was high (such as October 1, the highest temperature was 25°C), the air conditioner was used more frequently and the electricity consumption was high; when the temperature was low (such as October 7, the highest temperature was 23°C), the electricity consumption was relatively low.
[0129] Relationship between humidity and electricity consumption:
[0130] When humidity is high (such as on October 7, when the average humidity was 75%), residents may turn on dehumidification equipment, resulting in increased electricity consumption.
[0131] Relationship between wind speed and power consumption:
[0132] When the wind speed is low (such as on October 1, the average wind speed was 2.5m / s), residents may rely more on air conditioning to adjust the indoor temperature, resulting in higher electricity consumption.
[0133] 4. Holiday arrangement analysis
[0134] Changes in electricity consumption during holidays:
[0135] During the National Day (October 1st) and Mid-Autumn Festival (September 29th) of a certain year, residents' electricity consumption increased significantly, especially residential electricity consumption, because many people chose to rest or party at home.
[0136] Electricity consumption in certain specific areas may also increase during local festivals (such as the Dragon Boat Festival and Lantern Festival), as these festivals are usually accompanied by large-scale public events and light shows.
[0137] We can further draw the following conclusions from the above user behavior prediction model (as well as the influencing factors of related parameters):
[0138] Basic power consumption mode:
[0139] Electricity consumption by residential users is lower at night and higher during the day, especially during the evening peak period.
[0140] Industrial electricity consumption is higher on weekdays and lower on weekends; residential electricity consumption is higher on weekends.
[0141] Socioeconomic factors:
[0142] Economic growth, urbanization process and improvement of residents' income level may increase electricity demand and become a significant influencing factor.
[0143] Weather factors:
[0144] Changes in temperature, humidity and wind speed directly affect residents' electricity consumption. Electricity consumption increases when temperatures are high and humidity is high. The influencing factor may be that high-power electrical appliances such as air conditioners and dehumidifiers may need to work for a long time to eliminate the effects of high temperature and humidity.
[0145] Holiday factors:
[0146] Electricity consumption increases significantly during statutory holidays and local festivals, especially in residential areas.
[0147] These specific data and conclusions provide a solid foundation for establishing user behavior prediction models, help to more accurately capture users' electricity consumption patterns and their changing trends, and provide a scientific basis for power grid dispatching decisions.
[0148] Furthermore, the interactive power grid dispatch feedback method can also include real-time monitoring and early warning, response strategy design and user participation incentives.
[0149] Real-time monitoring and early warning include:
[0150] Data collection: Deploy smart sensors with high-frequency sampling to monitor key parameters such as current, voltage, power factor, etc. at each node of the power grid in real time.
[0151] Anomaly detection: Use outlier detection algorithms (such as Isolation Forest and LOF) to identify abnormal conditions in power grid operation and issue alerts in advance.
[0152] Response strategy design includes:
[0153] Graded response: According to the different degrees of abnormality, three-level response strategies are designed: mild, moderate and severe, corresponding to different regulatory measures.
[0154] Energy storage scheduling: During high-load periods, energy storage facilities are prioritized to release stored energy and smooth the load curve; during low-load periods, excess electricity is stored.
[0155] Load shifting: Through price signals or direct control measures, users are guided to shift non-critical loads to time periods when the grid is less stressed, such as at night or on weekends.
[0156] Emergency power generation: In extreme cases, start backup generators or call on surplus production capacity in neighboring areas to ensure that important loads are not affected.
[0157] User participation incentives include:
[0158] Points reward: A points system is established, and users can earn points by participating in demand response activities. Points can be used to redeem incentives such as electricity discounts and gift cards.
[0159] Information transparency: Real-time grid status and response suggestions are pushed to users through mobile phone apps, text messages, etc. to enhance user participation awareness.
[0160] Embodiment 2
[0161] This embodiment provides an interactive power grid dispatching feedback system, and the interactive power grid dispatching feedback system applies the interactive power grid dispatching feedback method as described in the first embodiment.
[0162] The interactive power grid dispatching feedback system at least includes a storage device, a processor, and a computer program stored in the storage device and executed by the processor.
Claims
1. An interactive power grid dispatching feedback method, characterized in that: The method comprises: Establish a user behavior prediction model to capture the user's electricity usage pattern and the changing trend of the user's electricity usage pattern over time; Setting up a dynamic feedback regulation database based on the user behavior prediction model; When a corresponding abnormal load is detected, the dynamic feedback regulation database is called accordingly.
2. The interactive power grid dispatching feedback method according to claim 1, characterized in that: The method further comprises: Set up machine learning algorithms to continuously optimize user behavior prediction models and dynamic feedback regulation databases.
3. The interactive power grid dispatching feedback method according to claim 2, characterized in that: The machine learning algorithm includes: Set up an online learning framework to accumulate learning data and strengthen the training model through incremental learning and / or transfer learning and / or federated learning; Reinforcement learning applications are used to conduct environmental modeling, build a simulation environment for power grid operation, simulate different operation scenarios and results, introduce strategy exploration algorithms, and allow the model to synchronously trial and error in the simulation environment during power grid operation to generate different operation strategies and sort them according to their priorities.
4. The interactive power grid dispatching feedback method according to claim 1, characterized in that: Building a user behavior prediction model includes: Multi-source data fusion analysis, based on information feedback from multiple sources, evaluates the impact of information from each source on electricity demand, optimizes user behavior prediction models, and adjusts power dispatch according to the impact on electricity demand.
5. The interactive power grid dispatching feedback method according to claim 4, characterized in that: The multi-source data includes: User feedback sentiment analysis: Through the natural language processing model, we analyze the user's information feedback on various power consumption platforms, extract the user's demand for their own power consumption, and judge the impact on power dispatch based on the demand; Life data association: combining city event announcements and traffic flow data to predict the impact of large-scale events, emergencies and traffic conditions on power grid load; Environmental monitoring association: Introduce weather conditions, air quality index and noise pollution data into the user behavior prediction model, and judge the impact of environmental information on residents' electricity consumption behavior based on changes in various environmental information.
6. The interactive power grid dispatching feedback method according to claim 1, characterized in that: Building a user behavior prediction model includes: Set up privacy protection to erase personal identification information after data collection and only retain information related to electricity usage behavior; The collected data is homomorphically encrypted and calculations are performed directly on the encrypted data.
7. The interactive power grid dispatching feedback method according to claim 1, characterized in that: The dynamic feedback regulation database is provided with a mesh division, and corresponding dynamic feedback regulation databases are provided at different power consumption nodes.
8. Interactive power grid dispatch feedback system, characterized in that: The interactive power grid dispatching feedback system is applied with the interactive power grid dispatching feedback method as described in any one of claims 1 to 7.