Equipment for improving unit power generation and optimizing grid operation based on intelligent dispatching

Abstract

This invention relates to the field of energy management and discloses equipment for improving unit power generation and optimizing grid operation based on intelligent dispatching. The equipment includes: a data sensing module for collecting grid status data of the units and preprocessing it to obtain processed data; and a scale modeling module for constructing a unified mathematical model containing fast and slow variables using the processed data. A hierarchical Hamiltonian is used to describe the coupling relationship between fast and slow variables, embedding power balance equation constraints to ensure overall grid power conservation, and introducing Wasserstein distance relaxation unit start-up and shutdown integer constraints to handle non-convex optimization problems. The output is a multi-objective optimization mathematical model containing energy functions and constraints. By utilizing hierarchical Hamiltonian, dynamic characteristics such as synchronous generator rotor angle / angular velocity are incorporated into the same energy framework along with dispatching decision variables such as unit start-up and shutdown, and energy storage charging and discharging. Power balance equation constraints ensure overall grid energy conservation, while Wasserstein distance relaxation unit start-up and shutdown integer constraints are introduced.

Description

Technical Field

[0001] This invention relates to the field of energy management technology, specifically to equipment for improving unit power generation and optimizing grid operation based on intelligent dispatching. Background Technology

[0002] The power system is evolving from traditional centralized power generation to a smart grid with multi-dimensional interaction among power generation, grid, load, and storage. The large-scale grid connection of new energy sources (such as intermittent power sources like wind power and photovoltaics) has made the power grid exhibit dynamic characteristics at multiple time scales. The demand for coordinated control of fast variables (response time in milliseconds) such as synchronous generator rotor angle / angular velocity and slow variables (decision cycle in minutes) such as unit start-up and shutdown and energy storage charging and discharging has increased significantly.

[0003] Traditional intelligent dispatch-based equipment for boosting generator power generation and optimizing grid operation often employs a single time-scale modeling method and a centralized optimization architecture. This approach treats fast variables such as synchronous generator rotor angle / angular velocity separately from slow variables such as generator start-up and shutdown, and energy storage charging and discharging. This leads to a disconnect between dispatch strategies and grid dynamics. For example, when formulating generator start-up and power distribution strategies, it is impossible to perceive rotor angle stability constraints in real time, resulting in power angle oscillations or even generator disconnection. Furthermore, centralized optimization lags in responding to high-frequency disturbances and struggles to coordinate control requirements at different time scales, ultimately affecting the stability of generator operation. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a device for enhancing generator power generation and optimizing grid operation based on intelligent scheduling. This solves the problem that traditional devices for enhancing generator power generation and optimizing grid operation based on intelligent scheduling suffer from a disconnect between the scheduling strategy and the dynamic characteristics of the power grid, which affects the stability of generator operation.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a device for improving unit power generation and optimizing grid operation based on intelligent dispatching, comprising:

[0006] Data sensing module: used to collect grid status data of the generating unit and preprocess it to obtain processed data;

[0007] Scale modeling module: It is used to construct a unified mathematical model containing fast and slow variables using processed data. It adopts hierarchical Hamiltonian to describe the coupling relationship between fast and slow variables, embeds power balance equation constraints to ensure the power conservation of the entire network, introduces Wasserstein distance relaxation unit start-up and shutdown integer constraints to handle non-convex optimization problems, and outputs a multi-objective optimization mathematical model containing energy functions and constraints.

[0008] The optimization control module is used to solve the global optimization problem of slow variables using a multi-objective optimization mathematical model, decompose the network scheduling task through the distributed alternating direction multiplier method, output the unit combination strategy and energy storage charging and discharging strategy, and solve the real-time control problem of fast variables using the model predictive control algorithm based on the real-time collected data, and output the equipment control commands.

[0009] Intelligent execution module: used to convert unit combination strategy, energy storage charging and discharging strategy, and equipment control commands into control signals for power generation equipment, energy storage system and reactive power compensation device, and to perform corresponding operations;

[0010] Interaction module: Used for human-computer interaction, receiving set parameters and outputting power grid operation status information.

[0011] By adopting the above technical solution, the dynamic characteristics of synchronous generator rotor angle / angular velocity, along with scheduling decision variables such as unit start-up and shutdown, and energy storage charging and discharging, are incorporated into the same energy framework using hierarchical Hamiltonian. Power balance equation constraints ensure the conservation of energy across the entire grid. At the same time, Wasserstein distance relaxation unit start-up and shutdown integer constraints are introduced. Through the synergy of distributed algorithms and model predictive control, a seamless connection between global scheduling strategy and real-time dynamic control is achieved. This makes the unit combination strategy and energy storage charging and discharging strategy inherently include grid stability constraints, improving unit operation stability and grid disturbance resistance. It achieves the synergistic progress of multi-objective optimization and safe operation of smart grid, solving the problem that traditional smart scheduling-based unit power generation improvement and grid optimization equipment are disconnected from grid dynamic characteristics, affecting unit operation stability.

[0012] Preferably, the data sensing module includes a data acquisition unit and a preprocessing unit. The data acquisition unit is used to acquire grid status data of the generating units through acquisition devices, including fiber optic gyroscopes, electromagnetic speed sensors, voltage transformers, lidar anemometers, and smart meters. The grid status data includes synchronous generator rotor angle, angular velocity, node voltage amplitude, renewable energy output, load, generating unit start-up and shutdown status, and energy storage state of charge. The preprocessing unit is used to preprocess the grid status data to obtain processed data containing initial features of fast variables and basic parameters of slow variables. The preprocessing includes using wavelet transform to denoise the grid status data, using an inverse information entropy causal reasoning algorithm to label causal relationships, and using graph neural networks for manifold dimensionality reduction.

[0013] Preferably, the scaling modeling module includes a fast variable modeling unit, a slow variable modeling unit, and a constraint embedding unit. The fast variable modeling unit is used to extract fast variables from the processed data, construct a fast variable dynamic model based on the synchronous generator rotor motion equations, describe the state transition characteristics of the fast variables on the Riemann manifold using differential geometry methods, and output a fast variable dynamic model expression containing the fast variable state equations. The fast variables include the synchronous generator rotor angle and angular velocity, which characterize the dynamic characteristics of the power grid. The slow variable modeling unit is used to extract slow variables from the processed data, construct a multi-objective function containing a power generation cost function, a carbon emission function, and a power grid stability function, and output a multi-objective function expression containing weighted coefficients. The power generation cost function includes unit fuel cost and start-up and shutdown cost. The power grid stability function includes node voltage deviation and frequency deviation. The slow variables include the unit start-up and shutdown states and energy storage charge states that characterize scheduling decisions. The constraint embedding unit is used to embed power balance equation constraints into the dynamic model expression of the fast variables. The active and reactive power conservation conditions of the entire network are transformed into constraint equations through the Lagrange multiplier method. The Wasserstein distance relaxation unit start-up and shutdown integer constraints are introduced into the multi-objective function expression to relax discrete variables into continuous variables and add distance constraints to handle non-convex optimization problems. Based on the dynamic model expression of the fast variables and the multi-objective function expression, the hierarchical Hamiltonian is used to describe the energy coupling relationship between the fast and slow variables. The dynamic model expression is transformed into the Hamiltonian of the fast variables, and the multi-objective function expression is transformed into the Hamiltonian of the slow variables, forming a unified mathematical model containing the hierarchical Hamiltonian energy function and multi-objective constraints, and outputting a multi-objective optimization mathematical model.

[0014] Preferably, the expression for the hierarchical Hamiltonian is: ,in, The total Hamiltonian for fast and slow variables. For the slow variable Hamiltonian, For fast variables, Hamiltonian For the slow variable state vector, Unit start-up and shutdown status , In the state of energy storage charge, For fast variable state vectors, For the rotor angle of the synchronous generator, The rotor angular velocity, These are the conjugate momentum of the slow variable and the fast variable, respectively. The time-scale separation parameter indicates that the response speed of the fast variable is 100 times that of the slow variable;

[0015] The distance constraint expression is: ,in, This represents the expected distribution of the unit's start-up and shutdown states after relaxation. The distance constraint is the Dirac measure at 0 and 1, which relaxes discrete integer variables into continuous variables and controls integer bias.

[0016] Preferably, the slow variable Hamiltonian It is derived from a multi-objective function, and its expression is: ,in, This is the unit's no-load cost coefficient. This is the fuel cost coefficient. This is the start-stop cost coefficient. As a carbon emission factor, As the voltage deviation weight, Weighting coefficients for the target;

[0017] The fast variable Hamiltonian Based on the rotor motion equations of a synchronous generator, the covariant form is as follows: ,in, For the generator's moment of inertia, The covariant derivative characterizes the state transitions on the Riemannian manifold. The damping coefficient is... The rated angular velocity, For mechanical power input, It is for electromagnetic power output.

[0018] Preferably, the optimization control module includes a cloud optimization unit and an edge control unit. The cloud optimization unit is used to decompose the whole network scheduling task into regional sub-problems using the distributed alternating direction multiplier method, solve the slow variable global optimization problem, and output the unit combination strategy and the energy storage charging and discharging strategy. The slow variable global optimization problem includes multi-objective optimization such as minimizing power generation costs, controlling carbon emissions, and improving grid stability. The unit combination strategy includes the start-stop state sequence of the units and the active power output allocation. The energy storage charging and discharging strategy includes the charging and discharging power sequence of the energy storage device and the state of charge constraints. The edge control unit is used to perform rolling optimization of the fast variable dynamic model expression based on the fast variables in the real-time collected processing data using the model predictive control algorithm, and output the equipment control commands.

[0019] Preferably, the problem of solving the slow variable global optimization problem is characterized by the following objective function. ,in, For the power generation cost function, For carbon emission function, For the power grid stability function, The target weight coefficient;

[0020] The model predictive control algorithm ,in, It is a fast-variable state vector containing the rotor angle and angular velocity of the synchronous generator. The reference trajectory is calculated using the results of slow variable optimization. and These are the state and control weight matrices, respectively, for the prediction time domain. step.

[0021] Preferably, the intelligent execution module includes a power generation equipment control unit, an energy storage system control unit, a reactive power compensation control unit, and a work execution unit. The power generation equipment control unit is used to convert the unit combination strategy and equipment control commands into active power regulation signals and excitation voltage control signals of the synchronous generator in the unit. The energy storage system control unit is used to convert the energy storage charging and discharging strategy into charging and discharging power control signals of the energy storage device in the unit. The reactive power compensation control unit is used to convert equipment control commands into reactive power output regulation signals of the static synchronous compensator. The work execution unit is used to control the unit to perform corresponding operations based on the active power regulation signals, excitation voltage control signals, charging and discharging power control signals, and reactive power output regulation signals, including active / reactive power regulation of the synchronous generator, charging and discharging control of the energy storage device, and dynamic reactive power compensation of the static synchronous compensator.

[0022] Preferably, the interaction module includes a human-computer interaction unit and a communication unit. The human-computer interaction unit is used to receive set target weight parameters and display power grid operation status information through a three-dimensional visualization interface to perform human-computer interaction. The communication unit is used to transmit various data through communication equipment, including fiber optic transmission equipment and 5G communication equipment.

[0023] The method for improving unit power generation and optimizing grid operation based on intelligent dispatching, applied to the aforementioned equipment for improving unit power generation and optimizing grid operation based on intelligent dispatching, includes the following steps:

[0024] Data sensing: Collect grid status data from generating units and preprocess it to obtain processed data;

[0025] Scale modeling: A unified mathematical model containing fast and slow variables is constructed using processed data. Hierarchical Hamiltonian is used to describe the coupling relationship between fast and slow variables. Power balance equation constraints are embedded to ensure power conservation across the entire network. Wasserstein distance relaxation unit start-up and shutdown integer constraints are introduced to handle non-convex optimization problems. The output is a multi-objective optimization mathematical model containing energy functions and constraints.

[0026] Optimized control: The system uses a multi-objective optimization mathematical model to solve the global optimization problem of slow variables, decomposes the network scheduling task through the distributed alternating direction multiplier method, outputs unit combination strategy and energy storage charging and discharging strategy, and uses model predictive control algorithm based on real-time collected data to solve the real-time control problem of fast variables, outputting equipment control commands.

[0027] Intelligent execution: Transforms unit combination strategies, energy storage charging and discharging strategies, and equipment control commands into control signals for power generation equipment, energy storage systems, and reactive power compensation devices to perform corresponding operations;

[0028] Interaction: Enables human-computer interaction, receives set parameters, and outputs power grid operation status information.

[0029] This invention provides equipment for improving generator power generation and optimizing grid operation based on intelligent scheduling. It has the following beneficial effects:

[0030] 1. This invention utilizes hierarchical Hamiltonian to incorporate the dynamic characteristics of synchronous generator rotor angle / angular velocity, along with scheduling decision variables such as unit start-up and shutdown, and energy storage charging and discharging, into the same energy framework. It ensures the conservation of energy across the entire grid through power balance equation constraints, while introducing Wasserstein distance relaxation unit start-up and shutdown integer constraints. Through the synergy of distributed algorithms and model predictive control, it achieves seamless integration of global scheduling strategies and real-time dynamic control. This makes the unit combination strategy and energy storage charging and discharging strategy inherently include grid stability constraints, improving unit operation stability and grid anti-disturbance capability, and realizing the synergistic progress of multi-objective optimization and safe operation of the smart grid.

[0031] 2. This invention introduces Wasserstein distance constraints to relax discrete integer variables such as unit start-up and shutdown into continuous variables. By controlling integer deviation through probability distribution distance, it more accurately preserves the physical constraint characteristics. Combined with the distributed alternating direction multiplier method to decompose the entire network task, it reduces computational complexity while avoiding the communication bottleneck of centralized optimization. This improves the efficiency of solving multi-objective optimization problems with non-convex constraints and enhances the feasibility of the strategy.

[0032] 3. This invention solves the unit combination and energy storage strategy based on a multi-objective function, and uses distributed optimization to achieve economic allocation of network resources. During off-peak hours, it controls energy storage charging to absorb excess renewable energy, and discharges during peak hours to reduce the number of unit start-ups and shutdowns, so that the units operate in the high-efficiency range, effectively reducing the overall power generation cost. At the same time, the reactive power compensation control unit and active power dispatch coordinate to improve the power factor of the grid and reduce line losses, thus achieving a dual improvement in energy utilization efficiency and economic benefits.

[0033] 4. This invention utilizes the distributed alternating directional multiplier method to decompose the entire network scheduling task into regional subproblems, solves the global optimization problem of slow variables, and outputs a unit combination strategy that can achieve the optimal allocation of unit start-up and shutdown and active power output, avoiding unnecessary start-up and shutdown losses, and enabling units to operate in the high-efficiency range. At the same time, considering the goal of minimizing power generation costs, it optimizes the system by comprehensively considering factors such as unit fuel costs and start-up and shutdown costs, and performs charging and discharging at appropriate times. This helps to smooth power fluctuations, reduce frequent unit adjustments, further improve power generation efficiency, reduce power generation costs, and improve the overall economic benefits of the power grid. Attached Figure Description

[0034] Figure 1 This is a system architecture diagram of the intelligent scheduling-based unit power generation enhancement and grid optimization operation equipment proposed in this invention;

[0035] Figure 2 This is a flowchart of the method for improving unit power generation and optimizing grid operation based on intelligent scheduling proposed in this invention. Detailed Implementation

[0036] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] Please see the appendix Figure 1 This invention provides equipment for improving generator power generation and optimizing grid operation based on intelligent scheduling, including:

[0038] Data Sensing Module: This module collects and preprocesses grid status data from the generating units to obtain processed data. It includes a data acquisition unit and a preprocessing unit. The acquisition unit collects grid status data from the generating units using acquisition devices, including fiber optic gyroscopes, electromagnetic speed sensors, voltage transformers, lidar anemometers, and smart meters. The grid status data includes synchronous generator rotor angle, angular velocity, node voltage amplitude, renewable energy output, load, unit start-up / shutdown status, and energy storage state of charge. The preprocessing unit preprocesses the grid status data to obtain processed data containing initial features of fast variables and basic parameters of slow variables. Preprocessing includes denoising the grid status data using wavelet transform, labeling causal relationships using an inverse information entropy causal reasoning algorithm, and performing manifold dimensionality reduction using a graph neural network.

[0039] Specifically, the data perception module captures real-time state information of the entire power grid through a multi-source heterogeneous sensor network and extracts key features using a three-level progressive preprocessing architecture. Acquisition units are deployed at the generation side, transmission hubs, and load centers, acquiring rotor angular displacement data via fiber optic gyroscopes, synchronously acquiring angular velocity via electromagnetic speed sensors, and detecting voltage amplitude using voltage transformers. LiDAR anemometers configured at new energy power plants construct a three-dimensional wind speed field model, which, combined with load sampling data from smart meters, forms a multi-temporal and spatial scale state dataset. The preprocessing unit first performs multi-level decomposition of the original signal using a wavelet denoising subunit, suppressing high-frequency noise while preserving effective frequency bands. The inverse information entropy causal reasoning subunit calculates the causal strength between variables based on transfer entropy theory and identifies strong causal links by constructing a causal relationship graph. The graph neural network manifold dimensionality reduction subunit uses a graph attention mechanism to map high-dimensional state vectors to low-dimensional Riemannian manifolds, compressing feature dimensions while maintaining the main data features. This module ultimately outputs processed data containing initial features of fast variables and basic parameters of slow variables, providing high-precision input for the scale modeling module. After preprocessing by this module, the power grid state estimation error is significantly reduced, the model training convergence speed is effectively improved, and it effectively supports subsequent multi-timescale collaborative optimization.

[0040] The scaling modeling module is used to construct a unified mathematical model containing fast and slow variables from processed data. It employs a hierarchical Hamiltonian to describe the coupling relationship between fast and slow variables, embeds power balance equation constraints to ensure power conservation across the entire grid, and introduces Wasserstein distance relaxation unit start-up and shutdown integer constraints to handle non-convex optimization problems. The output is a multi-objective optimization mathematical model containing energy functions and constraints. The scaling modeling module includes a fast variable modeling unit, a slow variable modeling unit, and a constraint embedding unit. The fast variable modeling unit extracts fast variables from the processed data, constructs a fast variable dynamic model based on the synchronous generator rotor motion equations, describes the state transition characteristics of fast variables on the Riemannian manifold using differential geometry methods, and outputs a fast variable dynamic model expression containing the fast variable state equations. Fast variables include the synchronous generator rotor angle and angular velocity, which characterize the grid dynamics. The slow variable modeling unit extracts slow variables from the processed data, constructs a multi-objective function containing a generation cost function, a carbon emission function, and a grid stability function, and outputs... The system includes a multi-objective function expression with weighted coefficients, a generation cost function including unit fuel cost and start-up / shutdown cost, a grid stability function including node voltage deviation and frequency deviation, and slow variables including unit start-up / shutdown state and energy storage charge state characterizing dispatch decisions. A constraint embedding unit is used to embed power balance equation constraints into the fast variable dynamic model expression. The Lagrange multiplier method is used to transform the conservation conditions of active and reactive power across the entire network into constraint equations. Wasserstein distance relaxation and unit start-up / shutdown integer constraints are introduced into the multi-objective function expression, relaxing discrete variables into continuous variables and adding distance constraints to handle non-convex optimization problems. Based on the fast variable dynamic model expression and the multi-objective function expression, a hierarchical Hamiltonian is used to describe the energy coupling relationship between fast and slow variables, transforming the dynamic model expression into a fast variable Hamiltonian and the multi-objective function expression into a slow variable Hamiltonian, forming a unified mathematical model containing a hierarchical Hamiltonian energy function and multi-objective constraints. The system outputs a multi-objective optimization mathematical model.

[0041] The expression for layered Hamiltonians is: ,in, For the slow variable state vector, Unit start-up and shutdown status , In the state of energy storage charge, For fast variable state vectors, For the rotor angle of the synchronous generator, The rotor angular velocity, These are the conjugate momentum of the slow variable and the fast variable, respectively. The time-scale separation parameter indicates that the response speed of the fast variable is 100 times that of the slow variable;

[0042] The distance constraint expression is ,in, This represents the expected distribution of the unit's start-up and shutdown states after relaxation. The distance constraint is the Dirac measure at 0 and 1, which relaxes discrete integer variables into continuous variables and controls integer bias.

[0043] Slow variable Hamiltonian It is derived from a multi-objective function, and its expression is: ,in, This is the unit's no-load cost coefficient. This is the fuel cost coefficient. This is the start-stop cost coefficient. As a carbon emission factor, As the voltage deviation weight, Weighting coefficients for the target;

[0044] Fast variable Hamiltonian Based on the rotor motion equations of a synchronous generator, the covariant form is as follows: ,in, For the generator's moment of inertia, The covariant derivative characterizes the state transitions on the Riemannian manifold. The damping coefficient is... The rated angular velocity, For mechanical power input, It is for electromagnetic power output.

[0045] Specifically, the scaling modeling module takes processed data as input and constructs a unified mathematical model coupling fast and slow variables. The fast variable modeling unit extracts fast variables such as synchronous generator rotor angle and angular velocity, describes their state transitions on a Riemannian manifold based on the rotor motion equations combined with differential geometry methods, and outputs a fast variable dynamic model expression to accurately characterize the dynamic characteristics of the power grid. The slow variable modeling unit extracts slow variables such as unit start-up and shutdown, and energy storage state of charge, constructs a multi-objective function covering generation costs (including fuel and start-up costs), carbon emissions, and grid stability (voltage and frequency deviations), and outputs an expression with weighted coefficients to clarify the scheduling optimization objective.

[0046] The constraint embedding unit embeds power balance constraints (active and reactive power conservation) into the fast variable model using the Lagrange multiplier method, forming constraint equations. Wasserstein distance relaxation unit start-up and shutdown integer constraints are introduced to make discrete variables continuous and control deviations, handling non-convex optimization. Based on the hierarchical Hamiltonian framework, the slow variable state vector (unit start-up and shutdown, energy storage charging) and the fast variable state vector (rotor angle, angular velocity) are coupled through time-scale separation parameters (e.g., ε=0.01, the fast variable response is 100 times that of the slow variable) and transformed into slow and fast variable Hamiltonians, respectively. The slow variable Hamiltonian integrates the multi-objective function, while the fast variable Hamiltonian is constructed based on the covariant form of the rotor motion equation (including moment of inertia, damping coefficient, etc.), reflecting the state transition on the Riemannian manifold.

[0047] Ultimately, a unified model incorporating hierarchical Hamiltonian energy functions and multi-objective constraints is formed, outputting a multi-objective optimization mathematical model. This module achieves collaborative modeling of fast and slow variables through energy coupling and constraint embedding, providing a precise mathematical expression for optimal control. In practical applications, it effectively improves the modeling accuracy of multi-timescale scheduling, supports the efficient operation of subsequent cloud optimization and edge control, and demonstrates excellent performance in areas such as grid power conservation and non-convex optimization, laying the foundation for increased unit power generation and optimized grid operation.

[0048] The expression for the layered Hamiltonian, assuming a certain power grid scenario, is as follows: Taiwanese unit, Synchronous generator, unit start-up and shutdown status Energy storage state of charge synchronous generator rotor angle Rotor angular velocity As input. For hierarchical Hamiltonian expressions Slow variable state vector fast variable state vector , These are conjugate momentum, respectively. This expression represents the energy function related to the slow variable. Energy function related to fast variables The system couples the energy relationships between fast and slow variables in a hierarchical manner, outputting a hierarchical Hamiltonian that includes the energy coupling between fast and slow variables. This enables a unified modeling of the energy relationships between variables at different time scales in the power grid, providing a foundation for subsequent multi-objective optimization.

[0049] For distance constraint expressions , This represents the expected distribution of the unit's start-up and shutdown states after relaxation. When processing the unit's start-up and shutdown states, the originally discrete integer variables of unit start-up and shutdown are... (Values ​​can be 0 or 1) This distance constraint is used to relax the variable and transform it into a continuous variable, controlling for integer deviations. For example, this applies to unit start-up and shutdown states. After relaxation, the constraint controls the deviation from the discrete value, thereby relaxing the discrete variable into a continuous variable and controlling the integer deviation, which facilitates subsequent processing in the optimization algorithm and effectively handles the originally complex non-convex integer optimization problem.

[0050] For the slow variable Hamiltonian Input the start / stop status of the generator unit. Generator active power Node voltage And other related parameters, , The coefficients are denoted as . This expression transforms the multi-objective functions, such as power generation cost, carbon emissions, and grid stability, into slow-variable Hamiltonians, outputting an energy function value that includes multi-objective weighting. This achieves a unified expression of the multi-objective problem in grid dispatch in the form of an energy function, providing a foundation for subsequent optimization solutions.

[0051] For fast variable Hamiltonians Input the moment of inertia of the synchronous generator Mechanical power input Electromagnetic power output The parameters are constructed based on the synchronous generator rotor motion equation. This expression outputs the energy function value characterizing the dynamic characteristics of the synchronous generator rotor, describing the state transitions of fast variables on the Riemannian manifold. This enables energy function modeling of rapidly changing dynamic characteristics in the power grid, providing a basis for analyzing and controlling power grid dynamic processes.

[0052] The optimization control module is used to solve the global optimization problem of slow variables using a multi-objective optimization mathematical model. It decomposes the entire network scheduling task using the distributed alternating direction multiplier method, outputs unit combination strategies and energy storage charging / discharging strategies, and, based on real-time acquired processing data, uses model predictive control algorithms to solve the real-time control problem of fast variables, outputting equipment control commands. The optimization control module includes a cloud optimization unit and an edge control unit. The cloud optimization unit decomposes the entire network scheduling task into regional sub-problems using the distributed alternating direction multiplier method, solves the global optimization problem of slow variables, and outputs unit combination strategies and energy storage charging / discharging strategies. The global optimization problem of slow variables includes multi-objective optimization of minimizing power generation costs, controlling carbon emissions, and improving grid stability. The unit combination strategy includes the start-stop state sequence of units and active power output allocation, while the energy storage charging / discharging strategy includes the charging / discharging power sequence of energy storage devices and state of charge constraints. The edge control unit uses model predictive control algorithms to perform rolling optimization of the dynamic model expression of fast variables based on the real-time acquired processing data, outputting equipment control commands.

[0053] Solving the slow variable global optimization problem is characterized by the following objective function. ,in, For the power generation cost function, For carbon emission function, For the power grid stability function, The target weight coefficient;

[0054] Model predictive control algorithm ,in, It is a fast-variable state vector containing the rotor angle and angular velocity of the synchronous generator. The reference trajectory is calculated using the results of slow variable optimization. and These are the state and control weight matrices, respectively, for the prediction time domain. step.

[0055] Specifically, the optimization control module, based on the multi-objective optimization mathematical model output by the scale modeling module, achieves hierarchical optimization control of the power grid. The cloud-based optimization unit utilizes the distributed alternating direction multiplier method to decompose the complex scheduling task of the entire network into multiple relatively independent regional sub-problems. When dealing with slow-variable global optimization problems, multiple objectives are considered, including minimizing generation costs, controlling carbon emissions, and improving grid stability. (Generation cost function) Integrated unit start-up and shutdown status and those who have contributed This includes fuel costs, start-stop costs, etc.; carbon emission function Carbon emissions are calculated based on the unit's operating status and output; grid stability function. Through node voltage and frequency Measuring the stability of the power grid. This is achieved by adjusting the target weighting coefficients. By balancing different objectives, the unit combination strategy (including the unit start-up and shutdown state sequence and active power output allocation) and energy storage charging and discharging strategy (including the charging and discharging power sequence and state of charge constraints) are obtained, realizing the global optimization allocation of network resources.

[0056] The edge control unit utilizes fast variables from real-time acquired and processed data, such as the synchronous generator rotor angle and angular velocity. It employs a model predictive control algorithm to calculate a reference trajectory based on the optimization results of slow variables. To achieve the objective, a rolling optimization is performed on the fast-variable dynamic model expression. Within each control cycle, the state weight matrix is ​​used to optimize the model. and control weight matrix Trade-off fast variable state vector Deviation from reference trajectory and control input The size of the prediction time domain is Step by step, by continuously optimizing control inputs and output device control commands, such as generator excitation current adjustment and governor opening adjustment, real-time and precise control of the power grid's rapid dynamic processes can be achieved.

[0057] This optimized control module, through hierarchical collaborative optimization, grasps the overall optimization direction in the cloud and achieves rapid dynamic response at the edge, effectively improving the unit's power generation and the stability of grid operation. It has demonstrated good adaptability and optimization effect in actual grid dispatch, ensuring the safe, economical and efficient operation of the grid.

[0058] Specifically, regarding the representation of the objective function, suppose a regional power grid includes 3 thermal power units, 2 wind power units, and 1 energy storage device, and the unit start-up and shutdown state vectors are given. Active power output vector Node voltage vector System frequency As input, the objective function is characterized by introducing Wasserstein distance constraints to handle the integer variables of unit start-up and shutdown, relaxing the discrete start-up and shutdown states into continuous variables and controlling for integer deviations, and inputting the expected distribution of unit start-up and shutdown states. The output satisfies The relaxed variables enable efficient handling of nonconvex optimization problems.

[0059] Model predictive control algorithms use fast variable state vectors (Rotor angle and angular velocity of the two synchronous generators) and control input vector Using mechanical power and excitation voltage regulation as inputs, the reference trajectory is calculated based on the slow variable optimization results. By minimizing the objective function The algorithm outputs the optimal control sequence for the next 20 steps, and after rolling optimization, obtains the control commands for the current moment (such as excitation current adjustment and governor opening adjustment), achieving precise control of the power grid dynamic process. This algorithm describes the fast / slow variable coupling relationship through a hierarchical Hamiltonian system. The input includes a slow variable vector containing unit start-up / shutdown and energy storage status, and a fast variable vector containing rotor angle and angular velocity. The output is an integrated multi-objective energy function, achieving unified modeling and collaborative optimization of dynamic characteristics across multiple time scales.

[0060] The intelligent execution module is used to convert unit combination strategies, energy storage charging and discharging strategies, and equipment control commands into control signals for power generation equipment, energy storage systems, and reactive power compensation devices, and to perform corresponding operations. The intelligent execution module includes a power generation equipment control unit, an energy storage system control unit, a reactive power compensation control unit, and an operation execution unit. The power generation equipment control unit is used to convert unit combination strategies and equipment control commands into active power regulation signals and excitation voltage control signals for the synchronous generator in the unit. The energy storage system control unit is used to convert energy storage charging and discharging strategies into charging and discharging power control signals for the energy storage devices in the unit. The reactive power compensation control unit is used to convert equipment control commands into reactive power output regulation signals for the static synchronous compensator. The operation execution unit is used to control the unit to perform corresponding operations based on the active power regulation signals, excitation voltage control signals, charging and discharging power control signals, and reactive power output regulation signals, including active / reactive power regulation of the synchronous generator, charging and discharging control of the energy storage device, and dynamic reactive power compensation of the static synchronous compensator.

[0061] Specifically, the intelligent execution module receives the strategies and instructions output by the optimization control module and transforms them into operable control signals to drive equipment operation. The generator control unit receives unit start-up and shutdown and active power output allocation information from the unit combination strategy, as well as excitation regulation signals from equipment control instructions. Through intelligent algorithms and hardware interfaces, this information is transformed into active power regulation signals for the synchronous generator. These signals precisely adjust the input power of the prime mover based on the real-time operating status of the unit and the grid load demand, thereby changing the generator's active power output. Simultaneously, an excitation voltage control signal is generated, which, by adjusting the excitation system, changes the generator's reactive power output capacity and terminal voltage level to ensure grid voltage stability.

[0062] The energy storage system control unit acquires the energy storage charging and discharging strategy, which includes the charging and discharging power sequence and state of charge constraints. Based on this information, the control unit calculates and outputs the charging and discharging power control signals for the energy storage device, working in conjunction with the energy storage management system to precisely control the charging and discharging process of the energy storage device. During periods of low grid load, the control unit charges and stores electrical energy; during periods of high load or when the grid experiences a power deficit, the control unit discharges to supplement the grid's power, achieving peak shaving and valley filling of grid power and flexible energy allocation.

[0063] The reactive power compensation control unit receives equipment control commands and converts them into reactive power output regulation signals for the Static Synchronous Compensator (STATCOM). Based on the real-time reactive power demand and voltage conditions of the power grid, the reactive power output of the STATCOM is precisely controlled by adjusting parameters such as the converter firing angle within the STATCOM. When the grid voltage drops, the STATCOM outputs inductive reactive power to support voltage recovery; when the voltage is too high, it absorbs inductive reactive power to reduce the voltage, maintaining the grid voltage within a reasonable range and improving the power factor and power quality of the grid.

[0064] The operation execution unit integrates the aforementioned active power regulation signals, excitation voltage control signals, charge / discharge power control signals, and reactive power output regulation signals. Through a high-speed communication link and a reliable control interface, it transmits these signals to the corresponding power generation equipment, energy storage devices, and reactive power compensation devices. Based on the received signals, the equipment precisely executes operations such as active / reactive power regulation of synchronous generators, charge / discharge control of energy storage devices, and dynamic reactive power compensation of static synchronous compensators, achieving real-time control of the power grid's operating status.

[0065] This intelligent execution module ensures the implementation of optimization strategies and control commands through signal conversion and equipment control, improves the response speed and operational stability of the power grid, and enables the power grid to meet the requirements of power supply and demand balance and power quality more efficiently and reliably.

[0066] Interaction Module: Used for human-computer interaction, receiving set parameters and outputting power grid operation status information. The interaction module includes a human-computer interaction unit and a communication unit. The human-computer interaction unit receives set target weight parameters and displays power grid operation status information through a 3D visualization interface, enabling human-computer interaction. The communication unit transmits various data through communication equipment, including fiber optic transmission equipment and 5G communication equipment.

[0067] Specifically, the interaction module acts as a bridge between users and the power grid system, enabling convenient human-machine interaction and data transmission. The human-machine interaction unit provides a 3D visualization interface, built on advanced graphics rendering technology, to intuitively present the power grid's operating status. Users can directly operate on the interface, inputting set target weight parameters. For example, in the multi-objective optimization of the optimization control module, weight coefficients can be assigned to objectives such as power generation cost, carbon emissions, and power grid stability. Adjustments are made to meet different operational needs and policy orientations. Simultaneously, the interface displays real-time grid operation status information, including key data such as synchronous generator rotor angle and angular velocity, node voltage amplitude, unit start-up and shutdown status, and energy storage charge status, presented in the form of charts, curves, and virtual models. This allows users to comprehensively and clearly understand the grid operation, promptly identify potential problems, and make informed decisions.

[0068] The communication unit utilizes fiber optic transmission equipment and 5G communication equipment to achieve high-speed and stable data transmission. The fiber optic transmission equipment, leveraging the principle of optical signal transmission in optical fibers, and with its high bandwidth, low loss, and strong anti-interference capabilities, is responsible for rapidly and accurately transmitting large amounts of real-time monitoring data from the power grid, such as high-precision power grid status data collected by the data sensing module, from various monitoring points to the system's core processing unit. The 5G communication equipment, with its high reliability and low latency, supports real-time interaction between mobile terminals and the system. For example, maintenance personnel can remotely access the human-machine interface of the human-machine interface unit via a handheld terminal using the 5G network to set parameters and view status; simultaneously, it can also rapidly transmit system control commands to the intelligent execution module, ensuring that the equipment can respond and execute operations promptly.

[0069] The interaction module, through the collaborative work of the human-machine interaction unit and the communication unit, enables efficient and convenient information exchange between users and the power grid system, improves the flexibility and intelligence level of power grid operation and management, helps operators better monitor and control power grid operation, and ensures the safe, stable and economical operation of the power grid.

[0070] The data sensing module collects real-time grid status data from multiple heterogeneous sensors and preprocesses it through wavelet denoising, causal labeling, and manifold dimensionality reduction to provide high-precision input for subsequent modeling. The scale modeling module constructs a unified mathematical model containing fast and slow variables, quantifies their energy coupling relationship using a hierarchical Hamiltonian system, embeds power balance equation constraints to ensure power conservation across the entire grid, and introduces Wasserstein distance to relax the integer constraints of unit start-up and shutdown into continuous variables and control deviations, forming a multi-objective optimization mathematical model. The optimization control module decomposes the grid scheduling task using the distributed alternating direction multiplier method to solve the global optimization problem of slow variables, outputting unit combination strategies and energy storage charging and discharging strategies with dynamic stability constraints. At the same time, it performs rolling optimization of the fast variable dynamic model based on model predictive control algorithms and outputs equipment control commands such as excitation regulation. The intelligent execution module transforms the optimization strategies and control commands into synchronous generator power regulation signals, energy storage charging and discharging control signals, and reactive power compensation regulation signals to achieve precise equipment control. The interaction module supports parameter setting and status monitoring through a 3D visualization interface and high-speed communication link. Through the synergistic effect of various modules, the millisecond-level dynamic characteristics of the power grid and the minute-level scheduling decisions are unified in the hierarchical Hamiltonian energy framework. This achieves deep coupling between multi-timescale optimization strategies and the dynamic characteristics of the power grid, solving the problem of the disconnect between scheduling strategies and the dynamic characteristics of the power grid in traditional intelligent scheduling-based unit power generation improvement and power grid optimization operation equipment, which affects the stability of unit operation and improves the stability of unit operation and the safety margin of the power grid.

[0071] Please see the appendix Figure 2 The method for improving generator power generation and optimizing grid operation based on intelligent dispatching, applied to the aforementioned equipment for improving generator power generation and optimizing grid operation based on intelligent dispatching, includes the following steps:

[0072] Data sensing: Collect grid status data from generating units and preprocess it to obtain processed data;

[0073] Scale modeling: A unified mathematical model containing fast and slow variables is constructed using processed data. Hierarchical Hamiltonian is used to describe the coupling relationship between fast and slow variables. Power balance equation constraints are embedded to ensure power conservation across the entire network. Wasserstein distance relaxation unit start-up and shutdown integer constraints are introduced to handle non-convex optimization problems. The output is a multi-objective optimization mathematical model containing energy functions and constraints.

[0074] Optimized control: The system uses a multi-objective optimization mathematical model to solve the global optimization problem of slow variables, decomposes the network scheduling task through the distributed alternating direction multiplier method, outputs unit combination strategy and energy storage charging and discharging strategy, and uses model predictive control algorithm based on real-time collected data to solve the real-time control problem of fast variables, outputting equipment control commands.

[0075] Intelligent execution: Transforms unit combination strategies, energy storage charging and discharging strategies, and equipment control commands into control signals for power generation equipment, energy storage systems, and reactive power compensation devices to perform corresponding operations;

[0076] Interaction: Enables human-computer interaction, receives set parameters, and outputs power grid operation status information.

[0077] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A unit power generation capacity improvement and power grid optimal operation device based on intelligent scheduling, characterized in that, The application relates to a power grid operation optimization system and method. The data sensing module is used for collecting power grid state data of a unit and preprocessing the data to obtain processing data; The scale modeling module is used for constructing a unified mathematical model containing fast variables and slow variables by using the processing data, adopting hierarchical Hamilton to describe the coupling relationship between the fast variables and the slow variables, embedding a power balance equation constraint to make the total network power conservation, introducing a Wasserstein distance relaxation to process non-convex optimization problems, and outputting a multi-objective optimization mathematical model containing an energy function and a constraint condition; The optimization control module is used for solving a slow variable global optimization problem by using the multi-objective optimization mathematical model, decomposing the total network scheduling task by using a distributed alternating direction multiplier method, outputting a unit combination strategy and a storage energy charging and discharging strategy, and solving a fast variable real-time control problem by using a model predictive control algorithm based on real-time collected processing data, and outputting device control instructions; The intelligent execution module is used for converting the unit combination strategy, the storage energy charging and discharging strategy and the device control instructions into control signals of power generation equipment, storage energy systems and reactive power compensation devices, and performing corresponding operations; The interactive module is used for human-computer interaction, receiving setting parameters and outputting power grid operation state information. 2.The intelligent dispatching-based unit power generation capacity improvement and power grid optimal operation device according to claim 1, characterized in that: The data sensing module comprises an acquisition unit and a preprocessing unit, the acquisition unit is used for collecting power grid state data of a unit through an acquisition device, the acquisition device comprises a fiber-optic gyroscope, an electromagnetic speed sensor, a voltage transformer, a laser radar anemometer and a smart meter, the power grid state data comprises a synchronous generator rotor angle, an angular velocity, a node voltage amplitude, a new energy output, a load, a unit start-stop state and a storage energy state of charge, and the preprocessing unit is used for preprocessing the power grid state data to obtain processing data containing fast variable initial characteristics and slow variable basic parameters, the preprocessing comprises denoising the power grid state data by using a wavelet transform, marking a causal relationship by using an inverse information entropy causal inference algorithm and performing manifold dimension reduction processing by using a graph neural network. 3.The intelligent dispatching-based unit power generation capacity improvement and power grid optimal operation device according to claim 1, characterized in that: The scale modeling module comprises a fast variable modeling unit, a slow variable modeling unit and a constraint embedding unit, the fast variable modeling unit is configured to extract fast variables from the processing data, construct a fast variable dynamic model based on a synchronous generator rotor motion equation, describe state transition characteristics of the fast variables on a Riemannian manifold through a differential geometry method, and output a fast variable dynamic model expression containing a fast variable state equation, the fast variables comprising a synchronous generator rotor angle and angular velocity representing dynamic characteristics of a power grid, the slow variable modeling unit is configured to extract slow variables from the processing data, construct a multi-objective function containing a power generation cost function, a carbon emission function and a power grid stability function, and output a multi-objective function expression containing a weighting coefficient, the power generation cost function containing a unit fuel cost and a start-stop cost, and the power grid stability function containing a node voltage deviation and a frequency deviation, the slow variables comprising a unit start-stop state and an energy storage state of charge representing scheduling decisions, and the constraint embedding unit is configured to embed a power balance equation constraint in the fast variable dynamic model expression, convert a whole-network active power and reactive power conservation condition into a constraint equation through a Lagrange multiplier method, introduce a Wasserstein distance relaxation unit start-stop integer constraint into the multi-objective function expression, relax discrete variables into continuous variables and add a distance constraint, process a non-convex optimization problem, and based on the fast variable dynamic model expression and the multi-objective function expression, adopt a hierarchical Hamiltonian to describe an energy coupling relationship between the fast variables and the slow variables, convert the dynamic model expression into a fast variable Hamiltonian, convert the multi-objective function expression into a slow variable Hamiltonian, form a unified mathematical model containing a hierarchical Hamiltonian energy function and a multi-objective constraint condition, and output a multi-objective optimization mathematical model. 4.The intelligent dispatching-based unit power generation capacity improvement and power grid optimal operation device of claim 3, characterized in that: The layered Hamiltonian is expressed as wherein, is the total Hamiltonian of fast and slow variables, is the slow variable Hamiltonian, is the fast variable Hamiltonian, is the slow variable state vector, is the unit on-off state , is the energy storage state of charge, is the fast variable state vector, is the synchronous generator rotor angle, is the rotor angular velocity, is the conjugate momentum of slow and fast variables, respectively, is the time scale separation parameter, representing that the response speed of the fast variable is 100 times that of the slow variable; The distance constraint expression is wherein, is the desired distribution of the unit commitment state after relaxation, is the Dirac measure at 0 and 1, which relaxes the discrete integer variable to a continuous variable by the distance constraint and controls the integer deviation. 5.The intelligent dispatching-based unit power generation capacity improvement and power grid optimal operation device according to claim 4, characterized in that: the slow variable Hamiltonian From the multi-objective function, the expression is wherein, is the unit no-load cost coefficient, is the fuel cost coefficient, is the start-stop cost coefficient, is the carbon emission factor, is the voltage deviation weight, is the target weighting coefficient; The fast variable Hamiltonian Based on the rotor motion equation of synchronous generator, the covariant form is Where, is the moment of inertia of the generator, is the covariant derivative, representing the state transition on the Riemannian manifold, is the damping coefficient, is the rated angular velocity, is the mechanical power input, is the electromagnetic power output. 6.The intelligent dispatching based unit power generation capacity promotion and power grid optimal operation device according to claim 1, characterized in that: The optimization control module comprises a cloud optimization unit and an edge control unit, the cloud optimization unit is configured to decompose a whole-network scheduling task into regional sub-problems through a distributed alternating direction multiplier method, solve a slow variable global optimization problem, and output a unit combination strategy and an energy storage charging and discharging strategy, the slow variable global optimization problem comprising multi-objective optimization of power generation cost minimization, carbon emission control and power grid stability improvement, the unit combination strategy comprising a start-stop state sequence and an active power output distribution of units, and the energy storage charging and discharging strategy comprising a charging and discharging power sequence and a state of charge constraint of an energy storage device, and the edge control unit is configured to adopt a model predictive control algorithm to perform rolling optimization on the fast variable dynamic model expression based on fast variables in real-time collected processing data, and output a device control instruction. 7.The intelligent dispatching based unit power generation capacity promotion and power grid optimal operation device according to claim 6, characterized in that: The solving of the slowly varying variable global optimization problem is characterized by the following objective function wherein, is a power generation cost function, is a carbon emission function, is a power grid stability function, is a target weight coefficient; The model predictive control algorithm wherein, is a fast variable state vector comprising the synchronous generator rotor angle, angular velocity, is a reference trajectory calculated from the slow variable optimization result, and are state and control weight matrices, respectively, prediction horizon steps. 8.The intelligent dispatching based unit power generation capacity promotion and power grid optimal operation device according to claim 1, characterized in that: The intelligent execution module includes a power generation equipment control unit, an energy storage system control unit, a reactive power compensation control unit, and a work execution unit. The power generation equipment control unit is configured to convert a unit combination strategy and an equipment control instruction into an active power adjustment signal and an excitation voltage control signal of a synchronous generator in a unit. The energy storage system control unit is configured to convert an energy storage charging and discharging strategy into a charging and discharging power control signal of an energy storage device in the unit. The reactive power compensation control unit is configured to convert the equipment control instruction into a reactive power output adjustment signal of a static synchronous compensator. The work execution unit is configured to control the unit to perform corresponding work according to the active power adjustment signal, the excitation voltage control signal, the charging and discharging power control signal, and the reactive power output adjustment signal, including active / reactive power adjustment of the synchronous generator, charging and discharging control of the energy storage device, and dynamic reactive power compensation of the static synchronous compensator. 9.The intelligent dispatching based unit power generation capacity promotion and power grid optimal operation device according to claim 1, characterized in that: The interaction module includes a man-machine interaction unit and a communication unit. The man-machine interaction unit is configured to receive a set target weight parameter and display power grid operation state information through a three-dimensional visualization interface, and perform man-machine interaction. The communication unit is configured to transmit various data through a communication device. The communication device includes a fiber transmission device and a 5G communication device.

10. A method for improving the power generation of a generating unit and optimizing the operation of a power grid based on intelligent scheduling, characterized in that: The device for improving unit power generation and optimizing power grid operation based on intelligent scheduling according to any one of claims 1-9 comprises the following steps: Data sensing: collecting power grid state data of a unit and preprocessing the data to obtain processed data; Scale modeling: constructing a unified mathematical model containing fast variables and slow variables using the processed data, using hierarchical Hamilton to describe the coupling relationship between the fast variables and the slow variables, embedding power balance equation constraints to conserve power in the whole network, introducing Wasserstein distance relaxation to handle non-convex optimization problems, and outputting a multi-objective optimization mathematical model containing an energy function and constraint conditions; Optimization control: solving a slow variable global optimization problem using the multi-objective optimization mathematical model, decomposing the whole network scheduling task through a distributed alternating direction multiplier method, outputting a unit combination strategy and an energy storage charging and discharging strategy, and solving a fast variable real-time control problem using a model predictive control algorithm based on real-time collected processed data, and outputting an equipment control instruction; Intelligent execution: converting the unit combination strategy, the energy storage charging and discharging strategy, and the equipment control instruction into control signals of power generation equipment, energy storage systems, and reactive power compensation devices to perform corresponding work; Interaction: performing man-machine interaction, receiving set parameters, and outputting power grid operation state information.

Images