Method and system for optimizing low-carbon operation of power system

Abstract

The invention discloses a power system low-carbon operation optimization method and system. The method comprises the steps of S1, collecting multi-source operation data in a target power grid dispatching period; s2, an optimal scheduling model is established based on the multi-source operation data, a target function is the weighted sum of economic cost and carbon emission cost, and limiting conditions include unit operation limitation, system power balance limitation and safe operation limitation; s3, introducing a demand side response variable and constraint into the optimal scheduling model, establishing a mathematical model of an interruptible load and a peak-shifting load, and bringing the power adjustment amounts of the two types of loads into system global power balance calculation; and S4, solving the optimization scheduling model after introduction of the demand side response variables and constraints by using a hybrid intelligent optimization algorithm, and outputting a scheduling scheme including a unit output plan and an adjustable load response plan. Through the improved optimization scheduling model and algorithm, a scheduling scheme for reducing power grid carbon emission, peak load shifting and guaranteeing operation safety can be output.

Description

Technical Field

[0001] This invention relates to the field of power system dispatching and low-carbon technology, and in particular to a method and system for optimizing the low-carbon operation of power systems. Background Technology

[0002] Amid the severe global climate change situation, the power industry, as a major source of carbon emissions, faces urgent requirements for carbon reduction. The goals of carbon peaking and carbon neutrality are driving the power system's transformation from a traditional high-carbon model to a clean and low-carbon one. However, existing power dispatching systems and technologies suffer from numerous inadequacies in adaptability. Traditional power dispatching, centered on economic efficiency and power supply reliability, fails to incorporate carbon emission costs into its optimization framework. This leads to the extensive deployment of high-emission coal-fired power units during peak load periods, limiting greenhouse gas emission reduction effects. Simultaneously, the large-scale grid connection of renewable energy sources such as wind and solar power, with their inherent intermittency and volatility, significantly increases the complexity of power system dispatching. This not only necessitates additional reserve capacity to ensure safe operation but also increases system operating costs and the risk of wind and solar curtailment.

[0003] Demand-side response, as a key means to improve grid flexibility and promote the consumption of clean energy, has been implemented in some regions. Through economic incentives to guide industrial users to use electricity in an orderly manner and to encourage residential users to respond to time-of-use pricing, it has achieved certain results in peak shaving, valley filling, energy conservation, and emission reduction. However, the existing grid dispatching system does not fully tap into and utilize demand-side resources, and has not yet formed a complete mechanism for coordinated optimization with the generation side, thus failing to fully unleash the potential for coordinated emission reduction. Meanwhile, with the expansion of the power system and the increase in constraints, traditional dispatching optimization methods such as single mathematical programming are gradually proving insufficient in terms of solution speed and uncertainty handling. Although artificial intelligence technologies (such as reinforcement learning) can autonomously learn control strategies through interaction with the environment, providing new ideas for power dispatching, these intelligent technologies have not yet been organically integrated with low-carbon dispatching.

[0004] In summary, existing technologies for low-carbon dispatching in power systems still suffer from three major shortcomings: 1) The lack of quantitative consideration of carbon emission costs in dispatch optimization makes it difficult to effectively reduce system carbon emissions; 2) The potential for demand-side response is not fully utilized, and the emission reduction effect of source-load coordination is not realized; 3) Faced with the uncertainty of renewable energy output, existing optimization algorithms struggle to solve large-scale complex models in a timely and efficient manner, lacking adaptive real-time adjustment capabilities. These technical deficiencies severely limit the emission reduction efficiency and operational efficiency of power systems.

[0005] The statements herein provide only background information in relation to this invention and do not necessarily constitute prior art. Summary of the Invention

[0006] The purpose of this invention is to address the lack of coordination and intelligence in existing low-carbon dispatching technologies by comprehensively considering both generation-side carbon emissions and load-side demand response. Through an improved optimization dispatching model and algorithm, a unified dispatching scheme is output to reduce grid carbon emissions, smooth peak and valley loads, and ensure operational safety. The provided system is then used to implement this dispatching scheme in actual grid dispatching, thereby reducing fossil fuel consumption and promoting the use of clean energy.

[0007] To achieve the above objectives, the present invention provides a method for optimizing the low-carbon operation of a power system, comprising the following steps: S1. Collect multi-source operational data within the target power grid dispatch cycle. The multi-source operational data includes power source side parameters, load side parameters, renewable energy data, and power grid physical constraint parameters. S2. Establish an optimized dispatch model based on the multi-source operational data. The objective function of the optimized dispatch model is a weighted sum of economic costs and carbon emission costs. The constraints of the optimized dispatch model include unit operation constraints, system power balance constraints, and safe operation constraints. S3. Introduce demand-side response variables and constraints into the optimized dispatch model to establish mathematical models for interruptible loads and peak-shaving loads. Incorporate the power adjustment amounts of the interruptible loads and peak-shaving loads into the system's global power balance calculation. S4. Solve the optimized dispatch model with the introduced demand-side response variables and constraints using a hybrid intelligent optimization algorithm, outputting a dispatch scheme that includes unit output plans and adjustable load response plans.

[0008] Optionally, in S1, the power supply side parameters include the upper limit of output, fuel consumption rate, and carbon emission coefficient; the electricity load side parameters include the predicted load curve; the renewable energy data includes the predicted renewable energy output curve; and the power grid physical constraint parameters include the power grid topology and equipment operation constraints.

[0009] Optionally, in S2, the economic cost includes the unit's fuel cost and start-up and shutdown cost, and the carbon emission cost is calculated from the unit's carbon emission coefficient and carbon price.

[0010] Optionally, in S2, the unit operation restrictions include upper and lower limits of unit output, ramp-up restrictions, and unit start-up and shutdown time restrictions; the safe operation restrictions include line power flow restrictions and node voltage safety restrictions.

[0011] Optionally, in S3, the demand-side response variables and constraints include constraints on the proportion of interruptible load reduction and the time period for peak-shifting load increase, which are set according to the real-time electricity price.

[0012] Optionally, in S4, the hybrid intelligent optimization algorithm includes a metaheuristic algorithm and a hybrid integer linear programming algorithm.

[0013] Optionally, the process of solving the problem using the hybrid intelligent optimization algorithm is as follows: first, the metaheuristic algorithm is used to perform a global search to obtain an initial solution, and then the hybrid integer linear programming algorithm is used to precisely optimize the initial solution.

[0014] Optionally, after S4, the method further includes: constructing a scheduling environment simulation model based on a reinforcement learning algorithm and training an agent, so that the agent can adjust the scheduling scheme online according to the real-time power grid operating status.

[0015] Optionally, the agent can be trained using a deep Q-network algorithm.

[0016] The present invention also provides a power system low-carbon operation optimization system for implementing the above method, comprising: The data acquisition and prediction module is used to collect multi-source operational data and predict power output. The optimization scheduling module, connected to the data acquisition and prediction module, is used to solve the optimization scheduling model and output a scheduling scheme. The demand response control module is connected to the optimization scheduling module and is used to interact with the user side, send incentive signals according to the optimization results and obtain load response feedback. The intelligent decision-making module, connected to the optimized scheduling module, evaluates the real-time power grid operating status based on reinforcement learning algorithms and adjusts the scheduling scheme online.

[0017] Compared with the prior art, the technical solution of the present invention has at least the following beneficial effects: 1) This invention makes carbon emission costs explicit as economic costs and incorporates them into the objective function of the optimization scheduling model, thereby imposing a direct economic penalty on high-emission units. Therefore, during the optimization process, the system will spontaneously guide the system to reduce the output of high-carbon emission units and increase the utilization of clean energy, thereby directly and effectively reducing the overall carbon emissions of the power system.

[0018] 2) Furthermore, the demand-side response mechanism is deeply integrated into the optimized scheduling model, enabling the model to simultaneously regulate power generation output and load consumption behavior, achieving source-load coordination. Peak shaving and valley filling through demand-side response reduce the dependence of peak loads on high-emission units, while smoothing load fluctuations, improving the grid's ability to absorb renewable energy fluctuations, and enhancing the flexibility and stability of system operation.

[0019] 3) Furthermore, a hybrid intelligent optimization algorithm combining metaheuristic algorithm and mixed-integer linear programming algorithm is adopted for solving the problem. First, the metaheuristic algorithm is used to perform a global fast search to obtain an initial solution, and then the mixed-integer linear programming algorithm is used to accurately optimize the initial solution. This overcomes the shortcomings of single algorithms in solving large-scale, nonlinear, complex scheduling models with integer variables, which are prone to getting trapped in local optima or having excessively long computation time. It significantly improves the solution speed and quality, so that this low-carbon operation optimization method can meet the actual power grid's requirements for computational efficiency.

[0020] 4) Furthermore, a simulation model of the scheduling environment is constructed based on the reinforcement learning algorithm and the agent is trained, enabling the agent to make optimal or near-optimal decisions under various system states. When unexpected situations such as prediction deviations in new energy output or load occur in actual operation, the agent can quickly adjust the scheduling scheme based on the real-time status, achieving a second-level response. This effectively makes up for the poor adaptive capability of traditional offline optimization models and greatly improves the robustness and reliability of the scheduling system in the face of uncertainty. Attached Figure Description

[0021] Figure 1 This is a flowchart of the power system low-carbon operation optimization method of the present invention.

[0022] Figure 2 This is a structural block diagram of the power system low-carbon operation optimization system of the present invention. Detailed Implementation

[0023] The following detailed description, in conjunction with the accompanying drawings and specific embodiments, provides a further detailed explanation of the low-carbon operation optimization method and system for power systems proposed in this invention. The advantages and features of this invention will become clearer from the following description. It should be noted that the accompanying drawings are in a very simplified form and use non-precise proportions, used only to facilitate and clearly illustrate the embodiments of this invention. Please refer to the accompanying drawings to make the objectives, features, and advantages of this invention more apparent and understandable. It should be understood that the structures, proportions, sizes, etc., depicted in the accompanying drawings are only for illustrative purposes to aid those skilled in the art and are not intended to limit the implementation conditions of this invention. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in proportions, or adjustments to the size, without affecting the effects and objectives achieved by this invention, should still fall within the scope of the technical content disclosed in this invention.

[0024] In terms of low-carbon dispatching in the power system, existing technologies have shortcomings such as the carbon emission costs not being quantified and incorporated into dispatching optimization, the demand-side response potential not being fully explored, and traditional algorithms and intelligent technologies being difficult to adapt to large-scale and complex low-carbon dispatching needs, which seriously limit the emission reduction efficiency and operational efficiency of the power system.

[0025] To address the lack of coordination and intelligence in existing low-carbon dispatching technologies, this invention comprehensively considers both generation-side carbon emissions and load-side demand response. First, it acquires multi-source operational data required for grid operation and generates 24-hour load and renewable energy output forecast curves using a predictive model. Next, it constructs an optimized dispatching model, incorporating carbon emissions into the objective function of the model through cost internalization. Once the load and renewable energy forecast data are ready, a hybrid intelligent optimization algorithm is used to solve the problem and output a dispatching scheme. If real-time optimization is required, a reinforcement learning agent further adjusts the dispatching scheme online. The final output dispatching scheme includes information such as the output and start-up / shutdown plans of each thermal power unit, and the execution plans for interruptible and peak-shifting loads, which is then distributed to power plants and users for execution through the energy management system.

[0026] Specifically, such as Figure 1 As shown, the present invention provides a method for optimizing the low-carbon operation of a power system, comprising at least the following steps: S1 collects multi-source operation data within the target power grid dispatch cycle.

[0027] Multi-source operational data includes power source parameters, load side parameters, renewable energy data, and grid physical constraint parameters. Power source parameters include output limits, fuel consumption rates, and carbon emission coefficients. Load side parameters include predicted load curves. Renewable energy data includes predicted renewable energy output curves (wind power, solar power, etc.). Grid physical constraint parameters include grid topology and equipment operating constraints. Based on this, a deep learning model is further used to predict load and renewable energy output, obtaining high-precision 24-hour predicted curves.

[0028] S2, Establish an optimized scheduling model based on the multi-source operation data.

[0029] The objective function considers both economic costs and carbon emission costs, summing them using pre-defined weighting coefficients. Economic costs include generator fuel costs and start-up / shutdown costs, while carbon emission costs are calculated based on generator carbon emission coefficients and carbon prices. Constraints on the optimization scheduling model include generator operation constraints, system power balance constraints, and safety operation constraints. Generator operation constraints include upper and lower limits on generator output, ramp-up limits, and generator start-up / shutdown time limits. Safety operation constraints include line power flow constraints and node voltage safety constraints. These constraints ensure that the optimization results meet the physical and safety requirements of the power grid.

[0030] S3, Introduce demand-side response variables and constraints into the optimized scheduling model.

[0031] Based on real-time electricity prices and user electricity consumption intentions, response models for interruptible loads and peak-shaving loads are set up. When the electricity price is higher than a threshold, interruptible load users reduce their load by an agreed proportion; when the electricity price is lower, peak-shaving loads (such as electric vehicle charging loads) increase their electricity consumption. Demand response participation constraints are established to ensure that the amount of load reduction or increase is within the user's allowable range and is included in the system's global power balance. By introducing demand response, peak-valley regulation of the load curve is achieved, thereby reducing reliance on high-emission units during peak hours. In specific implementation, interruptible loads of industrial users are reduced by no more than 15% during peak hours, and residential shiftable loads (such as electric vehicle charging) are started during off-peak hours at night, significantly narrowing the peak-valley difference. For example, in a test scenario, the total load during the evening peak was reduced by about 10% compared to the no-response scenario, while the nighttime load increased by about 5%. The narrowing of the peak-valley difference improved the unit combination and reduced the demand for coal-fired units to be activated during peak hours. This source-load coordination optimization effectively leverages the carbon reduction potential on the demand side.

[0032] S4. A hybrid intelligent optimization algorithm is used to solve the optimized scheduling model after introducing demand-side response variables and constraints.

[0033] The hybrid intelligent optimization algorithm comprises a metaheuristic algorithm and a mixed-integer linear programming (MILP) algorithm. First, a metaheuristic algorithm (e.g., Genetic Algorithm (GA) or Particle Swarm Optimization (PSO)) is used for a global search to quickly obtain a set of feasible initial solutions. Then, starting from this initial solution, the MILP algorithm is used to precisely optimize the unit combination and output, obtaining a near-optimal solution that satisfies the constraints. To address the residual uncertainties in renewable energy output and load forecasting, a certain safety margin is considered during the optimization scheduling model solution process. The final output includes the output plans for each unit and the response plans for each adjustable load. Testing shows that this hybrid intelligent optimization algorithm reduces computation time by approximately 30% while maintaining the quality of the scheduling scheme, and it is expected to be applied to scenarios requiring frequent calculations, such as daily planning, providing a feasible tool for practical power grid dispatching.

[0034] Following S4, the system further includes: constructing a scheduling environment simulation model based on a reinforcement learning algorithm and training an agent, enabling the agent to adjust the scheduling scheme online according to the real-time power grid operating status.

[0035] In applications requiring high real-time performance and adaptability, this method further constructs a scheduling environment simulation model based on reinforcement learning algorithms. The model uses the power grid's operating state as the environment, adjusting the scheduling scheme as the action, and negative overall operating costs as the reward to train the agent to learn the scheduling strategy. Through extensive training iterations, the agent can approximate the optimal scheduling strategy. In actual operation, when there is a deviation between prediction and reality or when unexpected events occur, the trained agent can quickly provide scheduling adjustment suggestions based on the new state, achieving second-level auxiliary decision support. In a specific embodiment, after training the agent using a deep Q-network algorithm for approximately 200 rounds, its cumulative reward convergence is significantly improved, and the corresponding daily operating cost differs from the traditional optimization result by less than 3%. This indicates that reinforcement learning technology can achieve online optimization and autonomous correction of scheduling decisions, improving the operational stability of the power grid in the face of uncertainty.

[0036] like Figure 2 As shown, this invention also provides a power system low-carbon operation optimization system for implementing the above method, comprising: a data acquisition and prediction module, an optimized scheduling module, a demand response control module, and an intelligent decision-making module. The functions of each module are as follows: the data acquisition and prediction module collects multi-source operating data and performs output prediction; the optimized scheduling module, connected to the data acquisition and prediction module, solves the optimized scheduling model and outputs a scheduling scheme; the demand response control module, connected to the optimized scheduling module, interacts with the user side, sends excitation signals based on the optimization results, and obtains load response feedback; the intelligent decision-making module, connected to the optimized scheduling module, evaluates the real-time power grid operating status based on a reinforcement learning algorithm and adjusts the scheduling scheme online.

[0037] The low-carbon operation optimization method of the present invention will be further explained below with reference to specific embodiments.

[0038] Example 1 The method of this invention is verified using the IEEE 39-node New England standard system. This system comprises 10 generators and 39 nodes, with a total load of approximately 6 GW. The carbon emission coefficients for coal-fired and gas-fired units are set at 1.0 tonnes / MWh and 0.5 tonnes / MWh, respectively, and the carbon trading price is set at 500 yuan / ton. First, an LSTM model is trained using historical data to predict the load and wind power output for the next 24 hours. The root mean square error of the prediction results is approximately 3%, which meets the dispatch requirements. Then, dispatch optimization is performed under the following scenarios: (a) economic dispatch only (without considering carbon costs and no demand response); (b) low-carbon dispatch considering carbon costs; (c) low-carbon dispatch + moderate demand response (assuming 50% load has response elasticity); (d) low-carbon dispatch + high-intensity demand response (100% load elasticity). A hybrid intelligent optimization algorithm is used to solve the optimization problem, calculating the unit output plan and the response adjustment amount of part of the load under each scenario.

[0039] Simulation results show that, compared to scenario (a), in scenario (b), due to the introduction of carbon costs, the output of high-emission coal-fired units generally decreases. For example, the output of unit number 2 drops from 680MW to around 500MW, reducing the total carbon emissions of the system by about 8%, but the total operating cost increases slightly by 1.5%. Scenario (c) introduces demand response on this basis, reducing peak load by an average of 10% and increasing off-peak load by 5%. For example, the peak load of node 4 (500MW) is reduced by 1.6MW, node 3 by 8.8MW, and node 2 by about 9MW during the off-peak period. The decrease in peak load causes some coal-fired units that would otherwise be in operation to shut down, further reducing carbon emissions to 15% lower than in scenario (a). At the same time, due to peak shaving and valley filling reducing the marginal power generation cost during peak load periods, the total operating cost of scenario (c) is about 0.5% lower than that of scenario (b). Scenario (d) assumes double the demand response potential, with peak load reduction and off-peak load increases approximately twice that of Scenario (c). Results show that carbon emissions are reduced by more than 20% relative to Scenario (a). However, excessive load shifting leads to a non-linear increase in wind power curtailment during certain periods, and user comfort may be significantly affected. Therefore, Scenario (c) is considered to achieve a good balance between emission reduction and cost. Furthermore, a reinforcement learning agent is applied to the scheduling environment simulated in Scenario (c). After training, the agent can adjust the output of some units based on real-time load and wind power changes, with the control effect deviating from the optimization result by less than 3%. Thus, the method of this invention successfully achieves significant carbon emission reduction (approximately 15%) with relatively low cost in this embodiment, validating its feasibility and effectiveness.

[0040] Example 2 A week-long simulation test was conducted on a real-world power grid in southern China (including 20 generators and a maximum annual load of 4GW). The system, deployed on a dispatch simulation platform, performs low-carbon dispatch calculations daily based on the next day's renewable energy output and load forecasts, and sends peak-shaving commands to the Demand Response platform via an interface. Test results show that compared to conventional dispatch, the low-carbon dispatch scheme cumulatively reduces CO2 emissions by approximately 5,000 tons, a reduction of about 12%; peak-shaving and valley-filling measures reduce the average daily load peak-valley difference by 8%, and reduce the operation of 3 coal-fired units during peak hours; meanwhile, system operating costs only increased by about 2%, demonstrating significant emission reduction effects while ensuring economic viability. Furthermore, in situations where actual solar irradiance falls short of expectations during certain peak load periods, resulting in lower-than-predicted photovoltaic output, the reinforcement learning module automatically issues suggestions to increase gas turbine output and initiate additional demand response, successfully avoiding voltage over-limits and power deficits, and improving flexibility and reliability in dealing with forecast errors and unforeseen circumstances.

[0041] The above embodiments demonstrate that the power system low-carbon operation optimization method and system based on demand response and intelligent optimization algorithms proposed in this invention can effectively reduce power system carbon emissions and achieve comprehensive optimization of economy and low carbon emissions with a slight increase in operating costs. The participation of demand-side resources improves the flexibility of grid dispatch and has a significant effect on peak shaving and valley filling. The introduction of artificial intelligence means that the dispatch strategy has self-learning and rapid decision-making capabilities, which can help the grid maintain safe and stable operation in scenarios with a high proportion of renewable energy access.

[0042] In summary, this invention comprehensively considers both carbon emissions from the power generation side and demand response from the load side. Through an improved optimization scheduling model and algorithm, it outputs a unified scheduling scheme that reduces grid carbon emissions, smooths peak flows, and ensures operational safety. The provided system is then used to implement this scheduling scheme in actual grid dispatching, thereby reducing fossil fuel consumption and promoting the use of clean energy.

[0043] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0044] Although the present invention has been described in detail through the preferred embodiments above, it should be understood that the above description should not be considered as a limitation of the present invention. Various modifications and substitutions to the present invention will be apparent to those skilled in the art after reading the above description. Therefore, the scope of protection of the present invention should be defined by the appended claims.

Claims

1. A low-carbon operation optimization method for a power system, characterized in that, Comprising the following steps: S1, collecting multi-source operation data in the target power grid dispatching period, the multi-source operation data including power supply side parameters, power consumption load side parameters, renewable energy data, and power grid physical constraint parameters; S2, establishing an optimization dispatching model based on the multi-source operation data, the objective function of the optimization dispatching model being a weighted sum of economic cost and carbon emission cost, the constraint conditions of the optimization dispatching model including unit operation constraints, system power balance constraints, and safe operation constraints; S3, introducing demand side response variables and constraints in the optimization dispatching model, establishing mathematical models of interruptible load and peak shifting load, and including power adjustment amounts of the interruptible load and the peak shifting load into system global power balance calculation; S4, solving the optimization dispatching model after introducing the demand side response variables and constraints by using a hybrid intelligent optimization algorithm, and outputting a dispatching scheme including unit output plan and adjustable load response plan.

2. The method of claim 1, wherein, In S1, the power supply side parameters include output upper limit, fuel consumption rate, and carbon emission coefficient; the power consumption load side parameters include predicted load curve; the renewable energy data include predicted renewable energy output curve; and the power grid physical constraint parameters include power grid topology and device operation constraints.

3. The method of claim 1, wherein, In S2, the economic cost includes unit fuel cost and start-stop cost, and the carbon emission cost is calculated by unit carbon emission coefficient and carbon price.

4. The method of claim 1, wherein, In S2, the unit operation constraints include unit output upper and lower limit constraints, ramping constraints, and unit start-stop time constraints; and the safe operation constraints include line flow constraints and node voltage safety constraints.

5. The method of claim 1, wherein, In S3, the demand side response variables and constraints include interruptible load reduction ratio constraint and peak shifting load increase period constraint set according to real-time electricity price.

6. The method of claim 1, wherein, In S4, the hybrid intelligent optimization algorithm includes meta-heuristic algorithm and mixed integer linear programming algorithm.

7. The method of claim 6, wherein, The solving process by using the hybrid intelligent optimization algorithm is as follows: firstly, the meta-heuristic algorithm is used for global search to obtain an initial solution, and then the mixed integer linear programming algorithm is used for accurate optimization of the initial solution.

8. The method of claim 1, wherein, After S4, a dispatching environment simulation model is constructed based on reinforcement learning algorithm, and an agent is trained, so that the agent can make online adjustment to the dispatching scheme according to real-time power grid operation state.

9. The method of claim 8, wherein, The method for training the agent is deep Q network algorithm.

10. A low-carbon operation optimization system for power systems, configured to implement the method of any one of claims 1-9, characterized in that, Comprise: a data acquisition and prediction module for acquiring multi-source operation data and performing output prediction; an optimization dispatching module connected with the data acquisition and prediction module, for solving the optimization dispatching model and outputting a dispatching scheme; a demand response control module connected with the optimization dispatching module, for interacting with user side, sending incentive signals according to optimization result, and obtaining load response feedback; an intelligent decision-making module connected with the optimization dispatching module, for evaluating real-time power grid operation state based on reinforcement learning algorithm and making online adjustment to the dispatching scheme.

Images