Power grid intelligent bearing capacity collaborative optimization method and system, product and medium
By combining deep reinforcement learning and a digital twin environment, real-time collaborative optimization between the power grid and the market was achieved, solving the problem of lagging power grid capacity assessment, improving the power grid's ability to accept and maintain green electricity, and reducing congestion costs.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-12
- Publication Date
- 2026-04-10
AI Technical Summary
Existing power grid capacity assessment methods are outdated and cannot respond in real time to changes in renewable energy output and market transactions, resulting in low power grid operating efficiency, insufficient stability, and a disconnect between the market and the power grid, leading to high congestion costs.
A dynamic carrying capacity assessment method based on deep reinforcement learning is adopted, combined with digital twin environment and multi-dimensional information data, and through iterative calculation and two-layer model optimization, the market and power grid are coordinated and optimized, mandatory safety boundary conditions are set, and real-time scheduling and market clearing are carried out.
It improves the grid's real-time acceptance of green electricity, reduces congestion costs, enhances the grid's flexibility and stability, achieves coordinated optimization between the market and the grid, and ensures the economy and security of power dispatch.
Smart Images
Figure CN121835984A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power grid assessment technology, and in particular to a method, system, product, and medium for collaborative optimization of power grid intelligent carrying capacity. Background Technology
[0002] As the global energy structure transitions towards a low-carbon model, the large-scale integration of green electricity (such as wind and solar power) places higher demands on the grid's capacity. Traditional grids, relying on fixed dispatch strategies and a single energy supply model, struggle to adapt to the highly volatile and unevenly distributed nature of renewable energy, leading to low grid efficiency and insufficient stability. Simultaneously, the accelerated pace of electricity market reforms, with the nationwide rollout of the electricity spot market and the gradual full participation of new energy sources in market transactions, promotes the consumption of green electricity while simultaneously placing higher demands on the grid's capacity, flexibility, and stability.
[0003] Grid carrying capacity refers to the maximum capacity of green electricity that a grid can accept while ensuring its safe and stable operation. Due to the strong randomness and volatility of green electricity output, problems such as grid power imbalance and frequency fluctuations are easily caused. Increased volatility in the electricity spot market necessitates a balance between economic efficiency and security for the grid in a market-driven environment. The reverse distribution of load centers and energy bases increases the pressure on inter-regional transmission, and the lagging construction of ultra-high-voltage (UHV) power transmission infrastructure may restrict the absorption of green electricity. Furthermore, existing grid carrying capacity assessments are mostly based on static or semi-static models, making it difficult to respond in real-time to changes in renewable energy output, market trading behavior, and load fluctuations. Therefore, the grid still faces multiple challenges in accommodating a high proportion of green electricity.
[0004] In existing technologies, grid capacity assessments mostly employ static, offline calculations based on fixed rules. These methods are time-consuming, cannot account for continuous temporal fluctuations in renewable energy output and load, and cannot quantify the expansion of capacity by flexible adjustment resources such as energy storage and demand response. After market clearing, safety checks are performed by dispatching agencies, requiring manual adjustments upon detection of exceeding limits. This leads to frequent modifications of market results, incurring high congestion costs, and a complete disconnect between market signals and grid physical security. Furthermore, most systems rely on historical data for training, making it difficult to respond in real-time to sudden factors such as extreme weather and market policy changes, resulting in lagging model updates. Therefore, the shortcomings of existing technologies lie in the disconnect between grid capacity calculations and the market, exhibiting a lag. Summary of the Invention
[0005] This invention provides a method, system, product, and medium for the coordinated optimization of smart grid carrying capacity, which can achieve coordinated optimization of green power consumption and safe grid operation.
[0006] This invention provides a method for collaborative optimization of smart power grid carrying capacity, comprising: Obtain source data from the main source system and extract the state vector of the source data; Based on the state vector, the dynamic bearing capacity of the current iteration node is calculated iteratively until the dynamic bearing capacity of all nodes is calculated, and the dynamic bearing capacity data sequence of each node at different time periods is output; wherein, the current iteration node is updated in each iteration. The clearing result is initialized based on the dynamic carrying capacity data sequence and the upper-level model. The clearing result is then input into the lower-level model, and the current market clearing result is iteratively calculated until the current iteration result meets the first preset condition and converges. Finally, the final market clearing result and the corresponding scheduling plan are output. If the obtained clearing result does not converge, the price signal is updated and the iteration continues. Execute the scheduling plan corresponding to the final market clearing result, and return the execution result of the scheduling plan corresponding to the final market clearing result to the data fusion processing center.
[0007] This application acquires data from multiple major source systems, obtaining multi-dimensional information data including electrical data and market factors, providing comprehensive foundational data for subsequent capacity analysis. Dynamic capacity data sequences are obtained through iterative calculations. The acquired multi-dimensional source data is transformed into dynamic capacity, serving as a core constraint on subsequent market clearing results. This achieves integrated market and grid collaborative optimization, ensuring that the collaborative optimization results naturally meet the real-time operational safety requirements of the grid during the planning stage, fundamentally eliminating the conflict between market and dispatch. By acquiring market clearing results, judgment data is provided for subsequent instruction generation. This judgment data is converted into actual operational instructions, resulting in a power dispatch plan corresponding to the clearing results. This systematically improves the resilience and flexibility of the grid, maximizing green power consumption, and is a key step in achieving subsequent power dispatch. By returning the instruction execution results to the data fusion processing center, a complete data feedback and model self-evolution closed loop is formed. Incremental training and online parameter calibration are performed on the prediction model and market behavior model, improving the system's real-time adaptive capability and maintaining the advancement and robustness of the decision-making strategy. Compared to existing technologies, this application combines a deep reinforcement learning-based dynamic power grid carrying capacity assessment method with a two-layer collaborative optimization method for power market and power grid operation safety to conduct real-time regulation of the green power grid under a multi-dimensional value system, achieving significant technical effects in reducing congestion costs and enabling ex-post adjustments.
[0008] Furthermore, the step of iteratively calculating the dynamic bearing capacity of the current iteration node based on the state vector specifically involves: Based on the state vector and the deep reinforcement learning network, the reward result and dynamic carrying capacity of the current iteration node are iteratively calculated until the reward result meets the second preset condition, at which point the dynamic carrying capacity of the current iteration node is output; wherein, during each iteration, the reward result is stored in the experience pool, and samples are taken from the experience pool to train and update the deep reinforcement learning network.
[0009] By introducing a dynamic carrying capacity assessment model based on deep reinforcement learning and continuous interaction with the digital twin environment, the accuracy of the deep reinforcement learning network in calculating the real-time acceptance capacity of the power grid for green electricity is improved through multiple iterations. By treating the power grid as a temporal game environment, a dynamic carrying capacity data sequence with high spatiotemporal resolution (e.g., 15 minutes) considering both safety and economy is obtained, providing a fundamental basis and hard boundaries for subsequent optimization decisions.
[0010] Further, the step of iteratively calculating the reward result of the current iteration node based on the state vector and the deep reinforcement learning network until the reward result satisfies the second preset condition, and then outputting the dynamic carrying capacity of the current iteration node, includes: The state vector is input into a pre-trained deep reinforcement learning network, and a set of action commands is output through the deep reinforcement learning network; wherein, the action commands coordinate and control at least two types of resources among thermal power units, energy storage devices and adjustable loads; A digital twin simulation environment is constructed based on a power grid model. In the digital twin simulation environment, the action commands are executed to perform line power flow calculation and security verification, and the security verification results are obtained. Based on the security verification results, the maximum power value of renewable energy that each regional power grid can additionally accept is calculated, and the dynamic carrying capacity of the current iteration node is generated.
[0011] This approach, employing deep reinforcement learning networks to acquire action commands, enables the coordinated scheduling of flexible resources such as thermal power, energy storage, and loads. Simulation experiments model the evolution of the power grid's state, conducting safety checks to improve the reliability of subsequent capacity results. This allows for autonomous learning, enabling the grid to proactively expand its real-time capacity to accept green electricity through coordinated scheduling of flexible resources while meeting all safety constraints. By generating a dynamic capacity data sequence indexed by time, the grid's real-time capacity to accept green electricity is expressed quantifiably, providing a fundamental basis and hard boundaries for subsequent clearing result calculations.
[0012] Furthermore, the initialization of the clearing result based on the dynamic bearing capacity data sequence and the upper-level model, the input of the clearing result into the lower-level model, and the iterative calculation of the current market clearing result are specifically as follows: Based on the declaration information of different market entities, the upper-level model runs a clearing algorithm under the constraints of mandatory safety boundary conditions to initialize the clearing results; Based on the clearing results, the efficiency of different market participants is solved through the lower-level model to obtain the bidding curves and planned quantities of different market participants. Based on the bidding curves and planned quantities of different market participants, a new clearing result is obtained.
[0013] By constructing a two-layer model and iterating multiple times under the constraint of mandatory safety boundary conditions, the upper-layer model initializes the clearing results and calculates preliminary clearing prices and quantities. The lower-layer model optimizes the calculations for different market participants, achieving synchronous and joint optimization of market transactions and power grid physical operation. This ensures that the market clearing results naturally meet the real-time operational safety requirements of the power grid during the planning stage, fundamentally eliminating the conflict between the market and dispatch. By embedding the constraints of mandatory safety boundary conditions into the market clearing model, guidance signals based on these conditions are released before market participants submit their information. During the market clearing process, the mandatory safety boundary conditions, pricing, and supply-demand relationships are simultaneously optimized, achieving the integration of "transaction" and "dispatch" during the planning stage and eliminating deviations between planning and execution.
[0014] Furthermore, the upper-level model runs a clearing algorithm under the constraints of mandatory safety boundary conditions, specifically as follows: Mandatory safety boundary conditions are set based on the dynamic bearing capacity data sequence; The mandatory safety boundary conditions include power balance constraints, line power flow constraints, dynamic carrying capacity constraints, and unit operation constraints. Among them, based on the dynamic carrying capacity data sequence, power generation equals load as the power balance constraint; Among them, based on the dynamic carrying capacity data sequence, the limitation on line transmission power is used as the line power flow constraint; Among them, based on the dynamic carrying capacity data sequence, the limitation that the total output of new energy sources does not exceed the carrying capacity is used as the dynamic carrying capacity constraint; Among them, based on the dynamic bearing capacity data sequence, the upper and lower limits of output and the ramp rate limit are used as the unit operation constraints.
[0015] By setting multiple constraints, the calculation process of the clearing results is constrained from four aspects: power balance, line flow, dynamic carrying capacity, and unit operation, ensuring that the final market clearing results are referable.
[0016] Further, the step of obtaining source data from the main source system and extracting the state vector of the source data includes: Source data is obtained from the main source system, and the source data is cleaned and spatiotemporally aligned to form an input data packet with a unified format; The power grid topology features, electrical operation features, and new energy output timing features are extracted from the input data packets to construct a state vector that can be recognized by a deep reinforcement learning network.
[0017] This process of cleaning the acquired source data removes outliers, improving the accuracy of the final results. Spatiotemporal alignment creates a standardized input package, facilitating subsequent integration and computation. By extracting topological features, the acquired basic data is transformed into feature vectors that can be input into the artificial intelligence network, providing input data that supports subsequent deep learning computations.
[0018] Furthermore, the acquisition of source data from the main source system specifically includes: The main source systems include: power grid dispatch automation system, new energy power forecasting system, power market trading platform, meteorological monitoring system, and load management system; Among them, the power grid topology and real-time power flow are obtained through the power grid dispatch automation system; Among them, wind and solar power generation forecasts are provided through the new energy power forecasting system; Among these, market transaction information is obtained through the electricity market trading platform; Meteorological data is obtained through a meteorological monitoring system. Load information and adjustability potential are obtained through the load management system.
[0019] By acquiring source data from five main source systems, and using electrical data, market data, load data, and green grid data as data support from multiple perspectives, subsequent collaborative optimization calculations of the grid's intelligent carrying capacity are carried out. This ensures that the final results include multi-dimensional influencing factors, thereby improving the reliability and practical significance of the calculation results.
[0020] This invention provides a smart power grid carrying capacity collaborative optimization system, comprising: a data input module, a carrying capacity analysis module, an instruction generation module, and an instruction execution module; The data input module is used to obtain source data from the main source system and extract the state vector of the source data; The load-bearing capacity analysis module iteratively calculates the dynamic load-bearing capacity of the current iteration node based on the state vector until the dynamic load-bearing capacity calculation of all nodes is completed, and outputs a dynamic load-bearing capacity data sequence formed by the load-bearing capacity data of each node at different time periods; wherein, the current iteration node is updated in each iteration. The instruction generation module is used to initialize the clearing result based on the dynamic bearing capacity data sequence and the upper-level model, input the clearing result into the lower-level model, iteratively calculate the current market clearing result, and output the final market clearing result and the corresponding scheduling plan when the current iteration result meets several preset conditions and converges. The instruction execution module is used to execute the scheduling plan corresponding to the final market clearing result and return the execution result of the scheduling plan corresponding to the final market clearing result to the data fusion processing center.
[0021] Another embodiment of this application provides a computer-readable storage medium storing a computer program product thereon, which, when executed by a processor, implements the steps of the near-grid intelligent carrying capacity collaborative optimization method of this application.
[0022] Another embodiment of this application provides a computer program product, including a computer program or instructions, which, when executed by a communication device, implement the steps of the smart grid carrying capacity collaborative optimization method of this application. Attached Figure Description
[0023] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0024] Figure 1 This is a flowchart illustrating one embodiment of the collaborative optimization method for intelligent power grid carrying capacity provided in this application.
[0025] Figure 2 This is a schematic diagram of one embodiment of a smart power grid capacity collaborative optimization system provided in this application. Detailed Implementation
[0026] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0027] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.
[0028] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.
[0029] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.
[0030] Example 1 See Figure 1 To address the problem of lagging power grid carrying capacity assessment in existing technologies, this application provides an embodiment of a collaborative optimization method for intelligent power grid carrying capacity, comprising steps 11 to 14, the specific steps of which are as follows: Step 11: Obtain source data from the main source system and extract the state vector of the source data.
[0031] Furthermore, source data is obtained from the main source system, and the source data is cleaned and spatiotemporally aligned to form a unified format input data packet; power grid topology features, electrical operation features, and new energy output timing features are extracted from the input data packet to construct a state vector that can be recognized by a deep reinforcement learning network.
[0032] Furthermore, the main source systems include: a power grid dispatch automation system, a new energy power prediction system, a power market trading platform, a meteorological monitoring system, and a load management system; wherein, the power grid dispatch automation system obtains power grid topology and real-time power flow data; wherein, the new energy power prediction system provides predicted output for wind and solar power generation; wherein, the power market trading platform obtains market transaction information, including but not limited to price information and transaction results; wherein, the meteorological monitoring system obtains meteorological data, including but not limited to temperature, wind speed, and irradiance; and wherein, the load management system obtains load data and adjustable potential.
[0033] In one embodiment, source data is obtained from a main source system, and the state vector of the source data is extracted, including steps 1101 to 1103, each of which is as follows: Step 1101: Obtain source data from the main source system.
[0034] Data was obtained from five main source systems: grid topology and real-time power flow from the power grid dispatch automation system; wind and solar power generation forecasts from the renewable energy power forecasting system; market transaction information from the electricity market trading platform; meteorological data from the meteorological monitoring system; and load information and adjustability potential from the load management system. By acquiring data from these five main systems, and using electrical data, market data, load data, and green grid data as supporting data from multiple perspectives, the final results incorporate multi-dimensional influencing factors, improving the reliability and scientific rigor of the obtained results.
[0035] Step 1102: Preprocess the collected source data to obtain a data packet in a uniform format.
[0036] In this process, the source data obtained in step 1101 is cleaned, aligned, and normalized in the data fusion processing center to form a data packet with a unified format. Abnormal data is removed, data quality is improved, and the accuracy of subsequent calculation results is enhanced. By aligning data from different sources or formats, the consistency of data in time, space, or logic is ensured. By unifying the data format, the standardization and interoperability of the data are improved.
[0037] Step 1103: Extract data features from the unified format data packet and construct a state vector.
[0038] Specifically, real-time power grid status data and external prediction data are obtained from the input data packet, and topology feature encoding is performed. Specifically, the node-branch correlation matrix of the power grid in the real-time state data is converted into an input format for graph neural network recognition; electrical measurements are normalized, and voltage and power values are converted to per-unit values; an LSTM network is used to extract the time-series variation patterns of renewable energy output and load, performing time-series feature extraction. The processed features are combined to form a state vector S_t. This standardized feature vector set lays a solid data foundation for subsequent artificial intelligence-based analysis.
[0039] The real-time status data of the power grid includes data such as node voltage, line power flow, generator output, and load distribution.
[0040] The external forecast data includes forecasts of new energy output, load, and weather trends for the next four hours.
[0041] Step 12: Based on the state vector, iteratively calculate the dynamic bearing capacity of the current iteration node until the dynamic bearing capacity of all nodes is calculated, and output the dynamic bearing capacity data sequence formed by the bearing capacity of each node at different time periods; wherein, the current iteration node is updated in each iteration.
[0042] Furthermore, based on the state vector and the deep reinforcement learning network, the reward result and dynamic carrying capacity of the current iteration node are iteratively calculated until the reward result meets the second preset condition, at which point the dynamic carrying capacity of the current iteration node is output; wherein, during each iteration, the reward result is stored in the experience pool, and samples are taken from the experience pool to train and update the deep reinforcement learning network.
[0043] Furthermore, the state vector is input into a pre-trained deep reinforcement learning network, and a set of action commands is output through the deep reinforcement learning network; wherein, the action commands coordinate and control at least two types of resources among thermal power units, energy storage devices, and adjustable loads; a digital twin simulation environment is constructed based on the power grid model, and the action commands are executed in the digital twin simulation environment to perform line power flow calculation and safety verification, and obtain safety verification results; based on the safety verification results, the maximum power value of renewable energy that the power grid in each region can additionally accept is calculated, and the dynamic carrying capacity of the current iteration node is generated.
[0044] In one embodiment, the dynamic bearing capacity of the current iteration node is iteratively calculated based on the state vector until the dynamic bearing capacity of all nodes is calculated, and the dynamic bearing capacity data sequence of each node at different time periods is output; wherein, during each iteration, the current iteration node is updated, including steps 1201 to 1207, and the specific steps are as follows: Step 1201: Input the state vector into the AI prediction module to perform high-precision new energy load prediction.
[0045] Step 1202: Input the state vector into the deep reinforcement learning network for forward computation to generate action vectors.
[0046] The deep reinforcement learning network is a DRL actor network, which includes an input layer, a hidden layer, and an output layer.
[0047] The input layer receives a feature matrix of dimension n×m. Feature transformation is performed in the hidden layer, and spatial features are extracted through a convolutional layer. The specific calculation formula for the feature transformation in the hidden layer is: h1=ReLU(W1·S_t+b1). Based on the spatial features extracted by the convolutional layer, action vectors are generated through the output layer. The specific calculation formula is: A_t=tanh(W2·h1+b2). The output value of the action vector ranges from [-1,1]. Through this DRL actor network, the feature matrix incorporating spatiotemporal topological information is converted into action vectors, providing crucial foundational data for subsequent generation of control commands.
[0048] Step 1203: The motion vector is decoded and constrained to convert it into actual control commands.
[0049] The action vector is decoded into an actual control quantity, specifically calculated as: ΔP_i = A_t[i] × ΔP_max_i. The ramp rate constraint is checked; if the adjustment exceeds the unit's ramp rate capability, adjustment is made according to the ramp rate limit. Finally, a complete action command is generated, encompassing unit power adjustment, energy storage charging / discharging, and load regulation. This mapping of the output action vector to actual control commands, along with constraint checks, ensures the physical feasibility and safety of the control commands, giving subsequent calculation results practical significance.
[0050] Step 1204: Input the action command into the digital twin simulation environment based on the power grid model to simulate the evolution of the power grid state and calculate the line power flow and node voltage for safety verification.
[0051] In this digital twin simulation environment, power flow calculations are performed on the power grid state at each moment. The line power flow is calculated by solving the equation system Bθ=P: P_l=B_l(θ_i-θ_j). The obtained line power flow is then subjected to safety verification to check whether the line power flow exceeds limits and whether the node voltage is within the allowable range. Finally, the new power grid state S_{t+1} is calculated. This high-fidelity digital twin environment provides accurate and reliable feedback data through closed-loop, dynamic safety simulation verification of control commands.
[0052] Step 1205: Calculate the reward function based on the security verification result.
[0053] The reward function comprises four parts: absorption reward, cost penalty, safety penalty, and smoothing penalty. The sum of the reward functions of each part is taken as the total reward. By performing a weighted combination of multiple reward functions, the reward function maximizes the absorption of new energy sources while taking into account the economy, safety, and smoothness of the system operation, which has guiding significance.
[0054] The specific formula for calculating the consumption reward is: R_green=α·ΣΔP_renewable, which encourages the consumption of more new energy sources.
[0055] The specific formula for calculating the cost penalty is: R_cost=-β·Σ(a·ΔP²+b·ΔP+c), which penalizes the power generation cost.
[0056] The specific calculation formula for the safety penalty is: R_safety=-γ·Σmax(0,|P_l|-P_l_max)²-δ·Σmax(0,|V|-V_limit)², which penalizes exceeding the safety limit.
[0057] The specific calculation formula for the smoothing penalty is: R_smooth=-ε·∥A_t-A_{t-1}∥², to avoid drastic changes in control instructions.
[0058] Step 1206: Based on the obtained reward function, perform experience storage and learning.
[0059] Among them, storage experience tuples<S_t,A_t,R_total,S_{t+1}> The network is fed into the experience pool, from which N=128 samples are randomly sampled. The critic network and the actor network are updated. By utilizing the experience pool and the dual-network soft update mechanism, stable and efficient learning is achieved, ensuring that the actor network can be optimized in real time and reliably in complex electrical environments, and guaranteeing the accuracy of the load-bearing capacity output results.
[0060] Specifically, the commentator network is updated by calculating the target Q value: y_i=r_i+γ·Q'(s_{i+1},μ'(s_{i+1})), minimizing the loss function: L=1 / NΣ(y_i-Q(s_i,a_i))².
[0061] Specifically, the policy gradient is calculated as follows: The target network parameters are updated using the policy gradient soft update, specifically calculated as: θ^Q'←0.001θ^Q+0.999θ^Q', θ^μ'←0.001θ^μ+0.999θ^μ', thereby updating the actor network.
[0062] Step 1207: Repeat steps 1202 to 1206 to iteratively calculate the dynamic carrying capacity of each node until convergence, that is, when the second preset condition is met, output the dynamic carrying capacity curve to express the real-time acceptance capacity of the power grid for green electricity in a data-driven manner, providing a fundamental basis and hard boundary for subsequent clearing result calculation.
[0063] The second preset condition is specifically: the average reward result change in the most recent 100 iterations is less than 0.1%.
[0064] The dynamic bearing capacity curve contains bearing capacity data for each node at different time periods, with a time resolution of 15 minutes.
[0065] Step 13: Initialize the clearing result based on the dynamic carrying capacity data sequence and the upper-level model, input the clearing result into the lower-level model, iteratively calculate the current market clearing result until the current iteration result meets the first preset condition and converges, and output the final market clearing result and the corresponding scheduling plan; wherein, if the obtained clearing result does not converge, update the price signal and continue iterating.
[0066] Furthermore, based on the declaration information of different market entities, the upper-level model runs a clearing algorithm under the constraint of mandatory safety boundary conditions to initialize the clearing results; based on the clearing results, the lower-level model solves for the benefits of different market entities to obtain the bidding curves and planned quantities of different market entities, and obtains new clearing results through the bidding curves and planned quantities of different market entities.
[0067] Furthermore, mandatory safety boundary conditions are set based on the dynamic carrying capacity data sequence; wherein, the mandatory safety boundary conditions include power balance constraints, line power flow constraints, dynamic carrying capacity constraints, and unit operation constraints; wherein, based on the dynamic carrying capacity data sequence, power generation equals load as the power balance constraint; wherein, based on the dynamic carrying capacity data sequence, the limitation on line transmission power is used as the line power flow constraint; wherein, based on the dynamic carrying capacity data sequence, the limitation on the total output of new energy sources not exceeding the carrying capacity is used as the dynamic carrying capacity constraint; wherein, based on the dynamic carrying capacity data sequence, the upper and lower limits of output and the ramp rate limit are used as the unit operation constraints.
[0068] In one embodiment, source data is obtained from a main source system, and the state vector of the source data is extracted, including steps 1301 to 1305, each of which is as follows: Step 1301: Initialize the upper-level model, set mandatory safety boundary conditions based on the dynamic bearing capacity curve obtained in step 12, and calculate the initial clearing results.
[0069] The upper-level model is the objective function set by the market operator. The specific initialization calculation formula is: minΣ_t[Σ_gC_g(P_g,t)+Σ_lC_congestion,l(P_l,t)], which minimizes the total power generation cost and congestion cost.
[0070] Specifically, the mandatory safety boundary conditions are: power balance constraints ensure that power generation equals load; line power flow constraints limit the power transmitted through the line; dynamic carrying capacity constraints limit the total output of new energy sources to not exceed the carrying capacity; and unit operation constraints include upper and lower limits of output and ramp rate limits.
[0071] Specifically, based on historical data, the node marginal electricity price LMP_i,t^0 is initialized to obtain the initial clearing result.
[0072] Step 1302: Based on the initial clearing results, solve the benefits of different market participants through the lower-level model.
[0073] The initial clearing result is used as a market signal LMP^k and published to all registered market participants through the trading platform.
[0074] The specific calculation formula for photovoltaic power plants among the various market entities is: maxΣ_t[LMP_i,t·P_pv,t], with the output value not exceeding the predicted value as a constraint. The calculation method for wind farm power plants is similar to that for photovoltaic power plants. The specific calculation formula for energy storage power plants is: maxΣ_t[LMP_i,t·P_discharge,t-LMP_i,t·P_charge,t], and the calculation process must meet the SOC constraint and charging / discharging power limit. The specific calculation formula for thermal power plants is: maxΣ_t[LMP_i,t·P_thermal,t-(aP²+bP+c)], and the calculation process considers power generation cost and operating constraints. The specific calculation formula for adjustable load is: minΣ_t[LMP_i,t·P_load,t+C_discomfort(ΔP_load)], which needs to balance electricity cost and dissipation cost. By solving the initial clearing results obtained from the upper-level model, a clear and efficient two-layer interactive market mechanism was constructed. This mechanism deeply integrates the electricity market trading mechanism and combines the impact of physical operation and electricity spot market price signals on power generation behavior and load response. It drives various market participants at the bottom level to optimize in parallel based on individual objectives, thus achieving spontaneous coordination between the evaluation results and the actual market operation.
[0075] Step 1303: Calculate the nodal marginal electricity price based on the optimized reporting results of each market participant.
[0076] Specifically, based on the optimized price curves and planned quantities obtained by each market participant, the results are reported to the market operation agency. The market operation agency aggregates the market responses and calculates the total supply curve. The specific calculation formula is: P_total^k=Σ_mP_m^k.
[0077] The optimized price curves and planned quantities obtained by each market participant include: photovoltaic price curve, wind power price curve, energy storage charging and discharging plan, thermal power price curve, and load demand response curve.
[0078] The process of implementing safety-constrained economic dispatch includes: constructing an optimization problem that minimizes total cost; linearizing the nonlinear cost function piecewise under carrying capacity and technical constraints; and simplifying AC power flow into DC power flow. By solving the problem under safety constraints, this step establishes an efficient and standardized market clearing and pricing process, achieving economic dispatch of resources and accurate price discovery while ensuring the safe operation of the power grid. Simplifying AC power flow into DC power flow reduces computational complexity and facilitates subsequent solutions.
[0079] This involves calling mathematical programming solvers such as CPLEX or Gurobi to solve mixed-integer linear programming or quadratic programming problems.
[0080] The specific formula for calculating the marginal electricity price at each node is as follows: LMP_i= Total cost / P_load_i = system λ + blocking component μ.
[0081] Step 1304: Repeat steps 1301 to 1303 to iteratively calculate the clearing results and perform convergence checks. Output the converged results as the final market clearing results.
[0082] The convergence check criteria are used as the first preset condition, including: the price change in the infinite norm is less than 0.01 yuan / MWh; the change in cleared electricity is less than 1MW; the number of iterations does not exceed 50; if the result does not converge, the price signal LMP^{k+1} is updated, and the iteration continues. By setting multiple clear and quantifiable convergence criteria, the stability and reliability of the iterative solution process are ensured.
[0083] The final market clearing result includes cleared electricity tables for each time period, a node marginal price matrix, a congestion management scheme, and ancillary service arrangements. The cleared electricity tables for each time period include time and electricity consumption of different generating units. The node marginal price matrix is formatted as time × node. The congestion management scheme includes a list of congested lines and congestion cost allocation. The ancillary service arrangements include frequency regulation reserve capacity and dispatch price.
[0084] Step 1305: Generate a scheduling plan corresponding to the final market clearing result based on the final market clearing result.
[0085] The dispatch plan includes: generation control commands to adjust the output of traditional generating units, including unit start-up, shutdown, and power regulation; energy storage dispatch commands to arrange charging and discharging schedules; demand response commands to output load regulation signals to guide flexible load regulation; and market guidance signals to issue price signals and congestion warnings to market participants. By generating a complete and executable market clearing scheme and dispatch command set, the optimization theory results can be directly applied to the safe and economical operation of actual power systems.
[0086] Step 14: Execute the scheduling plan corresponding to the final market clearing result and return the execution result of the scheduling plan corresponding to the final market clearing result to the data fusion processing center.
[0087] In one embodiment, the scheduling plan corresponding to the final market clearing result is executed, and the execution result of the scheduling plan corresponding to the final market clearing result is returned to the data fusion processing center, including step 1401, the specific steps of which are as follows: Step 1401 involves feeding back the actual execution effect of the scheduling plan to the monitoring system for data collection and returning it to the data fusion processing center. By forming a closed-loop optimization and learning mechanism of "source-grid-load-storage-market", rapid response and adaptive optimization of carrying capacity assessment can be achieved under scenarios such as extreme weather and sudden market changes.
[0088] Among them, based on the visualization platform, it provides decision support information such as carrying capacity heat map, operation risk warning and optimization benefit analysis.
[0089] See Figure 2 Another embodiment of this application also provides a smart grid carrying capacity collaborative optimization system, including: a data input module 201, a carrying capacity analysis module 202, an instruction generation module 203, and an instruction execution module 204; Data input module 201 is used to obtain source data from the main source system and extract the state vector of the source data; The bearing capacity analysis module 202 is used to iteratively calculate the dynamic bearing capacity of the current iteration node based on the state vector until the dynamic bearing capacity calculation of all nodes is completed, and output the dynamic bearing capacity data sequence formed by the bearing capacity data of each node at different time periods; wherein, the current iteration node is updated in each iteration. The instruction generation module 203 is used to initialize the clearing result based on the dynamic carrying capacity data sequence and the upper-level model, input the clearing result into the lower-level model, iteratively calculate the current market clearing result, and output the final market clearing result and the corresponding scheduling plan when the current iteration result meets several preset conditions and converges; wherein, if the obtained clearing result does not converge, the price signal is updated and the iteration continues. The instruction execution module 204 is used to execute the scheduling plan corresponding to the final market clearing result and return the execution result of the scheduling plan corresponding to the final market clearing result to the data fusion processing center.
[0090] It is understood that the above system embodiments correspond to the method embodiments of this application, and can implement the power grid intelligent carrying capacity collaborative optimization method provided by any of the above method embodiments of this application.
[0091] It should be noted that the system embodiments described above are merely illustrative, and some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided in this application, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.
[0092] Based on the above embodiments of the near-grid intelligent carrying capacity collaborative optimization method, another embodiment of the present invention provides a terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the grid intelligent carrying capacity collaborative optimization method of any embodiment of the present invention.
[0093] For example, in this embodiment, the computer program can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the terminal device.
[0094] The terminal device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.
[0095] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting all parts of the terminal device via various interfaces and lines.
[0096] Based on the above-described method embodiments, another embodiment of the present invention provides a computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to execute the power grid intelligent carrying capacity collaborative optimization method described in any of the above-described method embodiments of the present invention.
[0097] The modules / units integrated in the device / terminal equipment, if implemented as software functional units and sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.
[0098] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.
Claims
1. A method for collaborative optimization of smart power grid carrying capacity, characterized in that, include: Obtain source data from the main source system and extract the state vector of the source data; Based on the state vector, the dynamic bearing capacity of the current iteration node is calculated iteratively until the dynamic bearing capacity of all nodes is calculated, and the dynamic bearing capacity data sequence of each node at different time periods is output; wherein, the current iteration node is updated in each iteration. The clearing result is initialized based on the dynamic carrying capacity data sequence and the upper-level model. The clearing result is then input into the lower-level model, and the current market clearing result is iteratively calculated until the current iteration result meets the first preset condition and converges. Finally, the final market clearing result and the corresponding scheduling plan are output. If the obtained clearing result does not converge, the price signal is updated and the iteration continues. Execute the scheduling plan corresponding to the final market clearing result, and return the execution result of the scheduling plan corresponding to the final market clearing result to the data fusion processing center.
2. The method for collaborative optimization of smart grid carrying capacity according to claim 1, characterized in that, The step of iteratively calculating the dynamic bearing capacity of the current iteration node based on the state vector is as follows: Based on the state vector and the deep reinforcement learning network, the reward result and dynamic carrying capacity of the current iteration node are iteratively calculated until the reward result meets the second preset condition, at which point the dynamic carrying capacity of the current iteration node is output; wherein, during each iteration, the reward result is stored in the experience pool, and samples are taken from the experience pool to train and update the deep reinforcement learning network.
3. The method for collaborative optimization of smart grid carrying capacity according to claim 2, characterized in that, The step of iteratively calculating the reward result of the current iteration node based on the state vector and the deep reinforcement learning network until the reward result meets the second preset condition, and then outputting the dynamic carrying capacity of the current iteration node, includes: The state vector is input into a pre-trained deep reinforcement learning network, and a set of action commands is output through the deep reinforcement learning network; wherein, the action commands coordinate and control at least two types of resources among thermal power units, energy storage devices and adjustable loads; A digital twin simulation environment is constructed based on a power grid model. In the digital twin simulation environment, the action commands are executed to perform line power flow calculation and security verification, and the security verification results are obtained. Based on the security verification results, the maximum power value of renewable energy that each regional power grid can additionally accept is calculated, and the dynamic carrying capacity of the current iteration node is generated.
4. The method for collaborative optimization of smart grid carrying capacity according to claim 1, characterized in that, The process of initializing the clearing results based on the dynamic bearing capacity data sequence and the upper-level model, inputting the clearing results into the lower-level model, and iteratively calculating the current market clearing results involves the following steps: Based on the declaration information of different market entities, the upper-level model runs a clearing algorithm under the constraints of mandatory safety boundary conditions to initialize the clearing results; Based on the clearing results, the efficiency of different market participants is solved through the lower-level model to obtain the bidding curves and planned quantities of different market participants. Based on the bidding curves and planned quantities of different market participants, a new clearing result is obtained.
5. The method for collaborative optimization of smart grid carrying capacity according to claim 4, characterized in that, The upper-level model runs a clearing algorithm under the constraints of mandatory safety boundary conditions, specifically as follows: Mandatory safety boundary conditions are set based on the dynamic bearing capacity data sequence; The mandatory safety boundary conditions include power balance constraints, line power flow constraints, dynamic carrying capacity constraints, and unit operation constraints. Among them, based on the dynamic carrying capacity data sequence, power generation equals load as the power balance constraint; Among them, based on the dynamic carrying capacity data sequence, the limitation on line transmission power is used as the line power flow constraint; Among them, based on the dynamic carrying capacity data sequence, the limitation that the total output of new energy sources does not exceed the carrying capacity is used as the dynamic carrying capacity constraint; Among them, based on the dynamic bearing capacity data sequence, the upper and lower limits of output and the ramp rate limit are used as the unit operation constraints.
6. The method for collaborative optimization of smart grid carrying capacity according to claim 1, characterized in that, The step of obtaining source data from the main source system and extracting the state vector of the source data includes: Source data is obtained from the main source system, and the source data is cleaned and spatiotemporally aligned to form an input data packet with a unified format. The power grid topology features, electrical operation features, and new energy output timing features are extracted from the input data packets to construct a state vector that can be recognized by a deep reinforcement learning network.
7. The method for collaborative optimization of smart grid carrying capacity according to claim 6, characterized in that, The process of obtaining source data from the main source system specifically includes: The main source systems include: power grid dispatch automation system, new energy power forecasting system, power market trading platform, meteorological monitoring system, and load management system; Among them, the power grid topology and real-time power flow are obtained through the power grid dispatch automation system; Among them, wind and solar power generation forecasts are provided through the new energy power forecasting system; Among these, market transaction information is obtained through the electricity market trading platform; Meteorological data is obtained through a meteorological monitoring system. Load information and adjustability potential are obtained through the load management system.
8. A smart power grid carrying capacity collaborative optimization system, characterized in that, include: Data input module, load-bearing capacity analysis module, instruction generation module, and instruction execution module; The data input module is used to obtain source data from the main source system and extract the state vector of the source data; The bearing capacity analysis module is used to iteratively calculate the dynamic bearing capacity of the current iteration node based on the state vector until the dynamic bearing capacity calculation of all nodes is completed, and output the dynamic bearing capacity data sequence formed by the bearing capacity data of each node at different time periods; wherein, the current iteration node is updated in each iteration; The instruction generation module is used to initialize the clearing result based on the dynamic carrying capacity data sequence and the upper-level model, input the clearing result into the lower-level model, iteratively calculate the current market clearing result, and output the final market clearing result and the corresponding scheduling plan when the current iteration result meets several preset conditions and converges; wherein, if the obtained clearing result does not converge, the price signal is updated and the iteration continues. The instruction execution module is used to execute the scheduling plan corresponding to the final market clearing result and return the execution result of the scheduling plan corresponding to the final market clearing result to the data fusion processing center.
9. A computer-readable storage medium, characterized in that, The storage medium stores a computer program or instructions, which, when executed by a communication device, implement the power grid intelligent carrying capacity collaborative optimization method as described in any one of claims 1 to 7.
10. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by the communication device, the collaborative optimization method for intelligent power grid carrying capacity as described in any one of claims 1 to 7 is implemented.