Power grid engineering planning scheme optimization management system based on digital twinning

Abstract

The invention discloses a digital twinborn-based power grid engineering planning scheme optimization management system, and particularly relates to the technical field of power system planning, which comprises a data perception and fusion module, a dynamic interactive digital twinborn body module and an uncertainty quantification and scene generation module, a multi-target adaptive robust optimization module; and a cross-stage collaborative management module. According to the method, the digital twin is driven by real-time data to perform multi-time-scale dynamic deduction, an uncertainty scene set is constructed by using a depth generation model, and then a robust optimal planning scheme with high adaptability is generated by using a distributed robust optimization model with the worst case expectation cost minimization as a target, so that the robustness of the target is improved. And finally, lossless transmission and closed-loop feedback of the planning result to the design and construction stages are realized through a digital passport mechanism, and the robustness, economy and full-life-cycle management efficiency of the planning scheme are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of power system planning technology, and more specifically, to a power grid engineering planning scheme optimization management system based on digital twins. Background Technology

[0002] Power grid engineering planning is the foundation for the safe and stable operation of the power system and economical and efficient investment. With the high proportion of renewable energy access and the development of the electricity market, planning work faces severe challenges brought about by strong uncertainties on both the source and load sides.

[0003] Currently, power grid planning methods based on digital twins have been applied, but there are still obvious limitations: First, existing systems mostly focus on three-dimensional visualization and static data integration, and the digital model lacks dynamic data interaction capabilities with the real environment, making it difficult to achieve forward-looking extrapolation and analysis. Second, the uncertainties in the planning process have not been systematically quantified, and optimization algorithms mostly operate under deterministic or simple probabilistic scenarios, resulting in insufficient robustness of the planning scheme. Furthermore, the data is fragmented between planning and subsequent stages such as design and construction, forming information silos, which not only distorts the transmission of planning intent but also prevents valuable process feedback data from being used for model iterative optimization.

[0004] Therefore, in view of the above situation, the present invention provides a power grid engineering planning scheme optimization management system based on digital twin. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention provides a power grid engineering planning scheme optimization management system based on digital twins to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a power grid engineering planning scheme optimization management system based on digital twins, comprising: The data sensing and fusion module is used to collect dynamic data streams from the external environment in real time or near real time. The dynamic interactive digital twin module communicates with the data sensing and fusion module and is used to perform multi-time-scale simulation of power grid engineering planning schemes based on dynamic data flow. The uncertainty quantification and scenario generation module communicates with the dynamic interactive digital twin module and is used to generate a large number of simulation scenarios representing future uncertainties based on historical data and dynamic data streams through deep learning models. The multi-objective adaptive robust optimization module is connected to the dynamic interactive digital twin module and the uncertainty quantification and scenario generation module, respectively. It is used to solve the planning scheme based on the simulation results and simulation scenario using a distributed robust optimization model to generate the planning scheme with optimal robustness. The cross-stage collaborative management module communicates with the multi-objective adaptive robust optimization module to create a digital passport containing key information throughout the entire lifecycle for the generated robust and optimal planning scheme, and to achieve data collaboration and closed-loop feedback between the planning stage and the subsequent design and construction stages.

[0007] Preferably, the dynamic data stream collected by the data sensing and fusion module includes at least one of the following: Predicted weather data down to the grid level; Dynamic geographic information data monitored by satellite remote sensing; Macroeconomic policy and electricity market trend data obtained through natural language processing technology; Hotspot data on user-side load growth predicted by big data.

[0008] Preferably, the dynamic interactive digital twin module includes a multi-timescale inference engine, which is configured as follows: Rapid simulations are performed at a speed of seconds or minutes to assess the transient stability of the planning scheme under extreme conditions; A medium-speed simulation was conducted at a second speed of hour / day to assess the intraday renewable energy consumption and economic operation level of the planning scheme; A slow-motion simulation was conducted at a third speed, on a monthly / grade-by-grade basis, to simulate the impact of equipment aging and technological iteration on the long-term economic viability of the planning scheme.

[0009] Preferably, the deep learning model used in the uncertainty quantification and scene generation module is a conditional variational autoencoder or a generative adversarial network, which is used to learn the joint probability distribution of historical data and generate a set of physically reasonable, extreme but possible future scenarios.

[0010] Preferably, the multi-objective adaptive robust optimization module employs a distributed robust optimization model, the objective function of which is: in, Indicates the total investment cost; A fuzzy set representing the probability distribution of uncertainty constructed based on a simulation scenario; Representing fuzzy sets A certain probability distribution in, and Indicates the probability distribution The expected value operator; Indicates the probability distribution The following operating and maintenance costs; Indicates the probability distribution The associated risk costs; Indicates in fuzzy set All probability distributions In, operation and maintenance costs The worst-case scenario for the expected value of risk cost.

[0011] Preferably, the multi-objective adaptive robust optimization module further includes an online adjustment unit, which is constructed based on a reinforcement learning algorithm and configured as follows: When a deviation of a preset threshold is detected between real-time environmental data and predicted data, the system automatically calls or fine-tunes optimization strategies from a pre-stored optimization strategy library to achieve online adaptive adjustment of the planning scheme.

[0012] Preferably, the cross-stage collaborative management module creates a digital passport for the most robust planning scheme, embedding at least the following information: Key performance indicators and risk assessment reports derived from simulations in a digital twin; In response to equipment selection recommendations and special technical requirements raised in subsequent design phases; Early warnings and resource allocation suggestions were made for the construction difficulties in the subsequent construction phase.

[0013] Preferably, the cross-stage collaborative management module is further configured as follows: Receive feedback data on digital passports from the design or construction phase; Feedback data is fed into the dynamic interactive digital twin module to correct and update the digital twin model, forming a closed-loop optimization across stages.

[0014] Preferably, it also includes a method for optimizing and managing power grid engineering planning schemes, specifically comprising the following steps: S1. Real-time or near-real-time acquisition of dynamic data streams from the external environment; S2. Based on the acquired dynamic data stream, drive the digital twin to perform multi-time-scale simulation of power grid engineering planning schemes; S3. Based on historical data and dynamic data streams, a massive number of simulation scenarios representing future uncertainties are generated through deep learning models; S4. Based on the simulation results and simulation scenarios, a distributed robust optimization model is used to solve the planning scheme and generate the planning scheme with optimal robustness. S5. Create a digital passport containing key information throughout the entire lifecycle for the generated robust and optimal planning scheme, and implement collaborative management with subsequent stages based on the digital passport.

[0015] Preferably, in step S3, the joint probability distribution of historical data is learned using a conditional variational autoencoder or a generative adversarial network to generate a simulation scenario.

[0016] The technical effects and advantages of this invention are as follows: 1. This invention utilizes a deep generative model in the uncertainty quantification and scenario generation module to actively learn the underlying distribution and generate a massive number of extreme but possible future scenarios, thereby constructing an uncertainty fuzzy set. Subsequently, the distributed robust optimization model adopted by the multi-objective adaptive robust optimization module does not seek the average optimum, but focuses on optimizing the expected total cost under the worst probability distribution for the fuzzy set. This enables the final generated planning scheme to be immune to various unexpected fluctuations in the future, fundamentally improving the investment security and robustness of power grid projects when facing uncertainties such as new energy fluctuations and market changes. 2. This invention breaks through the limitations of traditional static planning and establishes a continuously running dynamic optimization loop. The data perception and fusion module injects external environmental data in real time, driving the dynamic interactive digital twin module to perform multi-timescale extrapolation, realizing continuous verification of the solution from transient stability to the whole life cycle. Crucially, the online adjustment unit in the multi-objective adaptive robust optimization module is based on the principle of reinforcement learning. When a significant deviation between reality and prediction is detected, it can automatically fine-tune the strategy and verify it through simulation, realizing the online adaptation of the solution, which greatly enhances the planning system's response and adaptability to changes in the real world. 3. This invention introduces a digital passport as a data carrier throughout the entire project lifecycle. During the planning phase, the cross-phase collaborative management module structurally embeds all the derivation results, performance indicators, and derived suggestions of the optimization scheme into the digital passport. This passport flows with the project to the design and construction phases, ensuring that the upstream optimization intentions are accurately implemented. At the same time, the downstream feedback data is transmitted back to the digital twin in real time for model correction, completely changing the previous situation of data fragmentation between stages and forming a cross-phase collaborative value-added chain. This effectively reduces design changes and construction rework, and improves the overall project efficiency and quality. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall system modules of the present invention.

[0018] Figure 2 This is a schematic diagram of the multi-timescale inference engine of the present invention.

[0019] Figure 3 This is a schematic diagram illustrating the working principle of the uncertainty quantification and scene generation module of this invention.

[0020] Figure 4 This is the overall flowchart of the present invention.

[0021] The attached diagram is labeled as follows: 1. Data perception and fusion module; 2. Dynamic interactive digital twin module; 3. Uncertainty quantification and scene generation module; 4. Multi-objective adaptive robust optimization module; 5. Cross-stage collaborative management module; 20. Multi-timescale inference engine; 40. Online adjustment unit. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Example 1 As attached Figure 1-3 As shown, this embodiment of the invention provides a power grid engineering planning scheme optimization management system based on digital twins, specifically including a data perception and fusion module 1, a dynamic interactive digital twin module 2, an uncertainty quantification and scenario generation module 3, a multi-objective adaptive robust optimization module 4, and a cross-stage collaborative management module 5. Each module communicates and interacts with data through a system bus or network. Specifically: Data perception and fusion module 1: This module is responsible for providing real-time or near real-time data support for the entire system. In this embodiment, the module collects and integrates the following multi-source dynamic data streams by calling various public API interfaces, deploying IoT sensing devices, and utilizing web crawler technology. Meteorological data is obtained from the National Meteorological Administration or commercial meteorological service providers (such as Windy and AccuWeather), with forecasts of wind speed, light intensity, temperature, and precipitation probability accurate to a 1km×1km grid for the next 7-15 days. The data is updated hourly. Geographic information data: By accessing remote sensing data sources such as high-resolution satellites and sentinel satellites, information such as topography, vegetation cover, and land subsidence risk of the planning area is acquired and updated regularly (e.g., quarterly), and sensitive areas that may affect power grid construction, such as ecological protection areas or geological disaster sites, are identified. Policy and Market Data: Using Python's Scrapy or BeautifulSoup libraries, we can scrape energy policies, regional development planning documents, and clearing price information from power trading centers. Through natural language processing technology based on the BERT model, we can perform sentiment analysis and keyword extraction on the text to quantify policy orientation and market trends. Load hotspot data: Integrating big data such as city information model (CIM), population heat map, and electric vehicle charging station construction plan, the K-means clustering algorithm is used to predict the hotspot areas and load density of regional load growth in the next 3-5 years; All collected heterogeneous data, after being cleaned, standardized in format, and aligned in time and space, are stored in time-series databases (such as InfluxDB) and relational databases (such as MySQL) for use by other modules.

[0024] Dynamic Interactive Digital Twin Module 2: This module is the core simulation environment of the system. Based on the dynamic data provided by the data perception and fusion module 1, it constructs a three-dimensional visualization model of power grid facilities (such as substations, transmission lines, and distributed power sources) in the Unity3D or UE4 engine, and integrates the computing kernel of power system analysis software (such as DigSILENT PowerFactory or OpenDSS) to form a multi-timescale inference engine 20. The specific working method of the multi-timescale inference engine 20 is as follows: Rapid simulation at the second / minute level: When receiving warnings of extreme weather such as typhoons and ice storms, the engine starts rapid simulation. Based on real-time meteorological data, it simulates the impact of extreme operating conditions on the power grid, calculates transient stability and voltage stability, assesses whether the planned scheme will cause cascading failures, and visualizes the failure propagation path. Hourly / Daily Medium-Speed ​​Simulation: Used to evaluate the daily operational performance of planning schemes. The engine simulates a typical day at 15-minute intervals, and performs power flow calculations based on wind and solar power output forecasts and load curves. It outputs indicators such as network loss, voltage deviation, and renewable energy curtailment rate for the entire day, and calculates the corresponding operating costs. Slow-motion simulation for months / grades: Used to evaluate the long-term benefits of planning schemes. The engine simulates the changes in reliability indicators (such as SAIDI, SAIFI) of the planning scheme over the next 20 years by introducing equipment aging models (such as Weibull distribution) and technology iteration curves (such as the annual efficiency improvement rate of photovoltaic modules), and calculates its total life cycle cost.

[0025] Uncertainty Quantification and Scene Generation Module 3: This module is responsible for addressing various challenges that may arise in the future. In this embodiment, the module uses the PyTorch or TensorFlow framework to build a conditional variational autoencoder as a deep learning model. The specific implementation steps are as follows: Model training: The training set consists of historical meteorological data, load data, and economic data from the past 10 years. The conditional variational autoencoder learns to map the high-dimensional historical data to a probability distribution in a latent space, while the decoder learns to sample and reconstruct new data consistent with the statistical characteristics of the training data from this latent space under certain conditions (such as "summer" and "high load growth"). Scenario Generation: After training, by inputting different conditional vectors and adding random noise into the decoder, thousands of sets of "physically reasonable" but historically unprecedented extreme future scenarios can be generated in batches. For example, scenarios such as "a week without wind and load reaching 120% of the historical peak" or "sharp fluctuations in photovoltaic output and a surge in electricity prices" can be generated. These scenario sets constitute the uncertainty fuzzy set required for subsequent optimization. The sample basis.

[0026] Multi-objective adaptive robust optimization module 4: This module is the decision-making brain of the system. It calls the deduction results of the dynamic interactive digital twin module 2 and the set of scenarios generated by the uncertainty quantification and scenario generation module 3 to optimize the planning scheme (such as route selection, substation capacity, and location and capacity of new energy installations). Its core is a distributed robust optimization model, whose objective function is as follows: in: This represents the total investment cost, including equipment purchase costs and civil construction costs, etc. A fuzzy set of uncertainty probability distributions constructed from the simulation scene generated by the uncertainty quantification and scene generation module 3; Representing fuzzy sets A certain probability distribution in, and Indicates the probability distribution The expected value operator; Indicates the probability distribution The operating and maintenance costs, including network loss electricity costs and equipment maintenance costs, are calculated using the medium-speed simulation results of the dynamic interactive digital twin module 2. Indicates the probability distribution The risk cost is used to quantify the failure risk of the planning scheme in different scenarios. For example, if load loss occurs in extreme scenarios, then... The value of this lost load is determined if the equipment is overloaded. For the corresponding penalty costs; Indicates in fuzzy set All probability distributions In, operation and maintenance costs The worst-case scenario for the expected value of risk cost; This optimization problem is a complex bi-level optimization problem. In this embodiment, a column and constraint generation algorithm is used to solve it, and the final output is a robust optimal planning scheme with the lowest total cost when facing the worst future situation. In addition, the module also includes an online adjustment unit 40, which is built based on reinforcement learning algorithms (such as Deep Q-Network, DQN). It continuously monitors the deviation between real environmental data and predicted data. When the deviation exceeds a preset threshold (such as wind speed prediction error exceeding 30%), the online adjustment unit 40 will automatically select a fine-tuning strategy (such as "emergency start of a backup gas turbine") from the pre-stored strategy library and drive the digital twin to perform rapid verification and execution, so as to realize the online adaptive adjustment of the planning scheme.

[0027] Cross-stage collaboration management module 5: This module is responsible for transmitting the results and value of the optimal planning scheme to the downstream stage. It creates a unique digital passport for each planning scheme generated by the multi-objective adaptive robust optimization module 4. The digital passport is a structured XML or JSON file that embeds at least the following information: Key performance indicators and risk assessment report: directly derived from the derivation and optimization results of the dynamic interactive digital twin module 2 and the multi-objective adaptive robust optimization module 4; Recommendations during the design phase include: recommended equipment models, a list of technical parameters, and requirements for special technologies (such as high seismic resistance and corrosion resistance); Recommendations for the construction phase include early warning of construction difficulties (such as crossing rivers or traversing forests) identified based on geographic information data, as well as suggested construction techniques and key resource allocation schemes. This module also provides a web service interface through which design or construction teams can access and read the contents of the digital passport. At the same time, they can upload feedback data such as construction drawing changes, equipment selection confirmation, on-site construction progress and problems through this interface. The cross-stage collaborative management module 5 then transmits this valuable on-site data to the dynamic interactive digital twin module 2 for correction and updating of the digital twin model to make it more realistic, thereby forming a closed-loop optimization process across planning, design and construction.

[0028] Example 2 As attached Figure 4As shown, based on Embodiment 1, the present invention also provides a method for optimizing and managing power grid engineering planning schemes, including the following steps: S1. Collect dynamic data streams from the external environment in real time or near real time, including future weather forecast data accurate to the grid, dynamic geographic information data monitored by satellite remote sensing, macroeconomic policy and power market trend data obtained through natural language processing technology, and user-side load growth hotspot data based on big data prediction, and clean, merge and store these multi-source heterogeneous data. S2. Based on the dynamic data stream collected in step S1, drive the dynamic interactive digital twin to perform multi-time-scale simulation of power grid engineering planning schemes. Specifically, this includes: rapid simulation at the second / minute level to evaluate transient stability; medium-speed simulation at the hour / day level to evaluate the level of new energy consumption and economic operation; and slow simulation at the month / year level to simulate the impact of equipment aging and technology iteration on long-term economic efficiency. S3. Based on historical data and the dynamic data flow in step S1, use deep learning models such as conditional variational autoencoders or generative adversarial networks to learn the joint probability distribution of historical data and generate a large number of physically reasonable, extreme but possible future simulation scenarios to construct an uncertain fuzzy set. S4. Based on the simulation results from step S2 and the simulation scenario generated in step S3, a distributed robust optimization model is used to solve the planning scheme. The objective function is: Ultimately, the most robust planning scheme is generated.

[0029] S5. Create a digital passport containing key information throughout the entire lifecycle for the robust optimal planning scheme generated in step S4. Based on this digital passport, achieve data collaboration with subsequent design and construction phases. At the same time, receive feedback data from downstream phases to correct and update the digital twin model, forming a closed-loop optimization.

[0030] In summary: This invention drives a digital twin to dynamically extrapolate power grid planning schemes across multiple time scales by collecting meteorological, geographical, policy, and load data in real time. At the same time, it uses deep learning to generate massive amounts of extreme future scenarios to quantify uncertainty. Based on this, a worst-case-oriented distributed robust optimization model is adopted to solve for the robust optimal planning scheme that minimizes the sum of total investment cost and worst-case expected operating risk. Ultimately, a digital passport carrying full lifecycle information is created for the solution, enabling seamless transfer of planning results to downstream stages and closed-loop optimization based on on-site feedback, thereby forming a dynamic adaptive management system from perception, prediction, optimization to collaborative execution.

[0031] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A digital-twin-based power grid engineering planning scheme optimization management system, characterized in that: Comprise: a data perception and fusion module (1) for collecting dynamic data streams from external environment in real-time or quasi-real-time; a dynamic interactive digital twin module (2) in communication connection with the data perception and fusion module (1) for performing multi-time scale deduction simulation on the power grid engineering planning scheme based on the dynamic data streams; an uncertainty quantification and scenario generation module (3) in communication connection with the dynamic interactive digital twin module (2) for generating a large number of simulation scenarios representing future uncertainty factors based on historical data and dynamic data streams through a deep learning model; a multi-objective adaptive robust optimization module (4) in communication connection with the dynamic interactive digital twin module (2) and the uncertainty quantification and scenario generation module (3) respectively for solving the planning scheme based on the deduction simulation results and the simulation scenarios by using a distributed robust optimization model to generate a robustly optimal planning scheme; a cross-stage collaborative management module (5) in communication connection with the multi-objective adaptive robust optimization module (4) for creating a digital passport containing key information throughout the life cycle for the generated robustly optimal planning scheme and realizing data collaboration and closed-loop feedback between the planning stage and the subsequent design and construction stages.

2. The digital-twin-based power grid engineering planning scheme optimization management system according to claim 1, characterized in that: The dynamic data streams collected by the data perception and fusion module (1) include at least one of: future weather forecast data accurate to the grid; geographic information dynamic data monitored by satellite remote sensing; macro policy and power market trend data obtained through natural language processing technology; user-side load growth hotspot data predicted based on big data.

3. The digital-twin-based power grid engineering planning scheme optimization management system according to claim 1, characterized in that: The dynamic interactive digital twin module (2) includes a multi-time scale deduction engine (20) configured to: perform rapid deduction at a first speed of seconds / minutes to evaluate the transient stability of the planning scheme under extreme working conditions; perform medium-speed deduction at a second speed of hours / days to evaluate the intra-day new energy consumption and economic operation level of the planning scheme; perform slow deduction at a third speed of months / years to simulate the impact of device aging and technology iteration on the long-term economy of the planning scheme.

4. The digital-twin-based power grid engineering planning scheme optimization management system according to claim 1, characterized in that: The deep learning model used in the uncertainty quantification and scenario generation module (3) is a conditional variational autoencoder or a generative adversarial network for learning the joint probability distribution of historical data and generating a set of physically reasonable and extreme but possible future scenarios.

5. The digital-twin-based power grid engineering planning scheme optimization management system according to claim 1, characterized in that: The distributed robust optimization model used by the multi-objective adaptive robust optimization module (4) has an optimization objective function as follows: wherein, represents the total investment cost; a fuzzy set representing an uncertainty probability distribution based on the simulation scenario construction; representing a certain probability distribution in a fuzzy set and denotes the expected value operator under the probability distribution ​ representing the operating and maintenance costs under the probability distribution ​ represents the risk cost under the probability distribution ​ represents the worst value of the expected value of the risk cost among all probability distributions in the fuzzy set of the running maintenance cost .

6. The digital-twin-based power grid engineering planning scheme optimization management system according to claim 1, characterized in that: The multi-objective adaptive robust optimization module (4) further includes an online adjustment unit (40) constructed based on a reinforcement learning algorithm and configured to: when a preset threshold deviation occurs between real-time environmental data and predicted data, automatically call or fine-tune an optimization strategy from a pre-stored optimization strategy library to realize online adaptive adjustment of the planning scheme.

7. The digital-twin-based power grid engineering planning scheme optimization management system according to claim 1, characterized in that: The digital passport created by the cross-stage collaborative management module (5) for the robustly optimal planning scheme at least embeds the following information: key performance indicators and risk assessment reports obtained through deduction in the digital twin; The equipment selection suggestion and special technical requirement proposed for the subsequent design stage; The construction difficulty early warning and resource allocation suggestion proposed for the subsequent construction stage.

8. The digital-twin-based power grid engineering planning scheme optimization management system according to claim 1, characterized in that: The cross-stage collaborative management module (5) is further configured to: Receive feedback data from the design stage or the construction stage for the digital passport; Deliver the feedback data to the dynamic interactive digital twin module (2) for correcting and updating the digital twin model, forming a cross-stage closed-loop optimization.

9. The digital-twin-based power grid engineering planning scheme optimization management system according to claim 1, characterized in that: Further comprising a power grid engineering planning scheme optimization management method, specifically comprising the following steps: S1, real-time or quasi-real-time acquisition of dynamic data flow from the external environment; S2, based on the collected dynamic data flow, driving the digital twin to perform multi-time scale deduction simulation on the power grid engineering planning scheme; S3, based on historical data and dynamic data flow, generate a large number of simulation scenarios representing future uncertainty factors through a deep learning model; S4, based on the deduction simulation results and the simulation scenarios, use a distributed robust optimization model to solve the planning scheme to generate a robustly optimal planning scheme; S5, create a digital passport containing key information throughout the life cycle for the generated robustly optimal planning scheme, and realize collaborative management with the subsequent stage based on the digital passport.

10. The digital-twin-based power grid engineering planning scheme optimization management system according to claim 9, characterized in that: In the S3 step, the conditional variational autoencoder or the generative adversarial network is used to learn the joint probability distribution of the historical data to generate the simulation scenario.

Images