A power spot day-ahead clearing method considering load increment declaration constraint

Abstract

The application discloses a power spot day-ahead clearing method considering load increment declaration constraint and relates to the technical field of power system operation decision-making. The method comprises the following steps: acquiring declaration parameters, constructing boundary parameter sets containing physical increment and decrement output variables, and fusing the boundary parameter sets, a target function and mutual exclusion constraints into a system state control matrix; mapping discrete decision variables in the matrix into an energy physical model, taking the lowest energy ground state solution as the unit and load combination state benchmark; constructing a dynamic evolution system containing scheduling strategies and disturbance nodes according to the above-mentioned method, making the system evolve to a stable limit cycle state in a phase space through differential equation set constraints, and generating a final day-ahead clearing plan and node marginal electricity price. The application is used to solve the problems that a traditional algorithm is prone to combinatorial explosion when processing massive discrete variables and a static clearing model lacks dynamic evolution flexibility and robust safety boundary when coping with high-dimensional continuous disturbance.

Description

Technical Field

[0001] This invention relates to the field of power system operation decision-making technology, and more specifically, to a day-ahead clearing method for electricity spot markets that takes into account load increment reporting constraints. Background Technology

[0002] The new power system exhibits a physical characteristic of deep interaction between a high proportion of renewable energy and massive flexible loads. As a core hub for power market operation and system dispatch, day-ahead clearing aims to achieve globally economically optimal resource allocation on both the source and load sides while satisfying complex network topology and physical constraints on equipment operation. The influx of massive heterogeneous market participants places extremely high engineering demands on the solution dimensions and disturbance resistance capabilities of the day-ahead clearing algorithm.

[0003] Mainstream day-ahead joint clearing solutions in the industry typically construct mathematical models based on mixed-integer linear programming (MILP) and rely on traditional commercial solvers to perform numerical optimization through algorithms such as branch and bound. To address system uncertainty disturbances, existing solutions often introduce static robust optimization mechanisms, which passively mitigate power prediction bias by constructing a minimax adversarial model for the worst-case scenario and reserving large-scale static reserve capacity.

[0004] However, the aforementioned conventional technical approaches face profound limitations. First, when dealing with extremely large-scale discrete decision variables such as the start-up and shutdown of massive market participants and polymorphic mutually exclusive responses, conventional MILP models are prone to "combinatorial explosion" and local deadlock problems, making it difficult to complete high-dimensional solutions within an effective timeframe. Second, the static and conservative minimax defense mechanism lacks dynamic evolutionary elasticity in phase space, failing to efficiently cope with high-dimensional continuous disturbances, often resulting in severe and ineffective locking of system adjustment resources. Third, most existing methods sever the linkage logic between underlying physical security scheduling and market deviation settlement, making it difficult to support a three-in-one panoramic closed-loop system from complex constraint dimensionality reduction optimization, dynamic disturbance resistance to final financial settlement. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a day-ahead clearing method for electricity spot markets that considers load increment reporting constraints. This method maps a system control matrix containing state mutual exclusion logic to an energy physics model to obtain the lowest energy ground state, and uses a set of differential equations to guide the dynamic evolution system containing scheduling strategies and uncertain disturbances to converge to a stable limit cycle in phase space. This solves the problems that traditional clearing algorithms are prone to combinatorial explosion and local deadlock when dealing with massive discrete decision variables, and that static optimization models lack dynamic evolutionary elasticity and robust safety boundaries when facing high-dimensional continuous disturbances.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A day-ahead clearing method for electricity spot markets considering load increment declaration constraints includes the following steps: obtaining load increment declaration parameters and constructing a market boundary parameter set containing physical increase and decrease power variables based on the declaration parameters; fusing a joint clearing objective function, the parameter set, and network physical constraints containing state mutual exclusion logic into a system state control matrix; mapping discrete decision variables in the control matrix to corresponding constraints into an energy physics model, and using the state solution of the model in the lowest energy ground state as the unit and load combination state benchmark; generating a scheduling strategy based on the state benchmark and constructing a dynamic evolution system containing scheduling strategy nodes and uncertain disturbance nodes; and using preset differential equation system constraints to evolve the system to a stable limit cycle state in phase space, generating a final day-ahead clearing plan and node marginal electricity prices.

[0007] In a preferred embodiment, constructing a market boundary parameter set containing physical increase and decrease power output variables based on the declared parameters includes: extracting benchmark predicted load data, the upper limit of the declared capacity for increase and decrease power output, and the corresponding multi-tiered bidding data from the load increment declaration parameters; obtaining the dynamic operating parameters of the generator units and the power grid topology physical structure diagram; generating the increase and decrease power output economic coefficients corresponding to the physical increase and decrease power output variables based on the multi-tiered bidding data; and normalizing the dynamic operating parameters, the power grid topology physical structure diagram, and the increase and decrease power output economic coefficients to form the market boundary parameter set.

[0008] In a preferred embodiment, the joint clearing objective function includes the sum of the total operating cost of the system's physical generator units and the marginal cost of adjusting the load increase or decrease of the entire network; the joint clearing objective function is configured as follows: when a local node meets the preset condition of exceeding the marginal cost limit, the physical power reduction variable is assigned a first priority weight in the network physical constraints; when the system's marginal clearing cost is lower than a preset threshold, the physical power increase variable is assigned a second priority weight in the network physical constraints.

[0009] In a preferred embodiment, mapping the discrete decision variables and corresponding constraints in the control matrix to an energy physics model includes: dimensionality reduction mapping of the generator start-up and shutdown states and the power output increase and decrease states of market users to spin node variables in a lattice network topology; transformation of the topological security boundary and resource constraint boundary in the network physics constraints into interaction Hamiltonian energy penalty terms between adjacent spin node variables; and obtaining a global objective energy function characterizing the system state control matrix by superimposing the intrinsic energy of the spin node variables with the interaction Hamiltonian energy penalty terms.

[0010] In a preferred embodiment, the network physical constraints further include demand-side physical response boundary hard constraints. In a preferred embodiment, the step of using the state solution of the model in the lowest energy ground state as the unit and load combination state reference includes: initializing the initial ambient temperature field and annealing time step of the simulated annealing algorithm; inputting the energy physics model containing the global objective energy function into the simulated annealing algorithm, and reducing the ambient temperature field by introducing random disturbance terms and according to a preset decay rate; when the system evolution reaches the preset minimum temperature termination condition, extracting the global minimum energy distribution state of the energy physics model at this time, and decoding it into the unit and load combination state reference containing equipment start-stop control commands.

[0011] In a preferred embodiment, constructing a dynamic evolution system containing scheduling strategy nodes and uncertainty disturbance nodes includes: defining the scheduling strategy as a controlled variable in the dynamic evolution system, defining the acquired source load power uncertainty disturbance as an environmental excitation variable in the dynamic evolution system; and constructing a set of differential equations characterizing the dynamic nonlinear coupling relationship between the controlled variable and the environmental excitation variable through a population symbiotic evolution model.

[0012] In a preferred embodiment, the step of causing the system to evolve to a stable limit cycle state in phase space through preset differential equation constraints includes: performing time-series iterative solution of the differential equation system under the boundary constraints of the power grid cross-section transmission capacity including the power transfer distribution factor; monitoring the evolution trajectories of the controlled variables and environmental excitation variables until the evolution trajectories converge in phase space to a closed, non-intersecting stable limit cycle; extracting the coordinate eigenvalues ​​of the central gravity point of the stable limit cycle and mapping them to a set of clearing scheduling instructions with robust safety margins.

[0013] In a preferred embodiment, after generating the final pre-emptive clearing plan and node marginal electricity prices, the process further includes performing electricity data clearing operations: extracting the system energy price component, network congestion price component, and marginal line loss price component from the node marginal electricity prices; generating electricity clearing data for each node based on the physical increase / decrease in the final pre-emptive clearing plan and the node marginal electricity prices corresponding to each node, combined with the user's baseline forecast load data; the electricity clearing data includes: first clearing data calculated based on the forecast price for the electricity corresponding to the baseline forecast load data; and second clearing data calculated based on the difference between the corresponding node marginal electricity price and the declared price for the electricity corresponding to the physical increase / decrease in output.

[0014] The technical effects and advantages of the day-ahead clearing method for electricity spot markets that takes into account load increment reporting constraints in this invention are as follows: This invention effectively coordinates the bidirectional flexible adjustment potential of the load side by constructing a global market boundary parameter set that includes physical power output variables. By innovatively mapping the system state control matrix containing state mutual exclusion logic to an energy physics model, and by using the method of finding the lowest energy ground state solution, it successfully resolves the combinatorial explosion and local deadlock problems that are easily caused by massive discrete decision variables, and achieves efficient dimensionality reduction and optimization of complex unit and load combinations. Furthermore, this invention overcomes the fragility of traditional static clearing solutions by constructing a dynamic evolution system that includes scheduling strategies and uncertain disturbances, and using a system of differential equations to make it converge and evolve to a stable limit cycle state in phase space. This endows the system with dynamic elasticity and robust safety boundaries when facing high-dimensional continuous disturbances, and finally generates a day-ahead clearing plan and nodal marginal electricity price that takes into account both grid physical security and market economic benefits. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the day-ahead clearing method for electricity spot markets that takes into account load increment reporting constraints, provided in an embodiment of the present invention.

[0016] Figure 2 The simulation diagram of system energy evolution convergence is provided for an embodiment of the present invention.

[0017] Figure 3 The simulation diagram of the evolution of the two-dimensional phase space limit ring is provided for the embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1, Figure 1 The present invention provides a day-ahead clearing method for electricity spot markets that considers load increment reporting constraints, comprising the following steps: Step S1: Obtain the load increment declaration parameters, and construct a global market boundary parameter set containing physical increase power variables and physical decrease power variables based on the declaration parameters.

[0020] It should be noted that the load increment reporting parameters in this embodiment of the invention are collected in real time from the data terminals of each market-based user through the data interaction interface of the power trading center using a secure communication protocol. In specific operation, the system verifies the user identifier, time segment, and response intention by parsing the received structured messages, thereby ensuring the integrity and timeliness of the original reporting data.

[0021] In this embodiment, constructing a global market boundary parameter set containing physical increase and decrease power output variables based on the declared parameters includes: extracting the baseline predicted load data, the upper limit of the declared capacity for increase and decrease power output, and the corresponding multi-segment tiered pricing data from the load increment declaration parameters; and obtaining the dynamic operating parameters of the generator units and the physical structure diagram of the power grid topology. Specifically, to accurately depict the economic characteristics of users participating in the spot market, the aforementioned multi-segment tiered pricing data is organized in system memory using a data structure of a multi-dimensional dictionary or a composite object array. Taking the physical increase power output declaration of a certain market-oriented user as an example, its data structure is divided into multiple capacity ranges and corresponding price tags: for example, when the incremental capacity range is... The corresponding bid price is 250 yuan / MWh; when the incremental capacity range is The corresponding bid price is 320 yuan / MWh for the incremental capacity range of 20 MW to its upper limit; the corresponding bid price is 450 yuan / MWh. This multi-tiered data organization can truly reflect the objective physical law that the user-side response cost increases non-linearly with the increase of output depth.

[0022] To illustrate the data structure more intuitively, this embodiment provides a set of multi-tiered pricing data for typical market-based users participating in physical increases and decreases in power output. Specific parameters are shown in Table 1. This table displays the data of a large industrial user (assuming its baseline forecast load). The marginal cost of reporting for different deviation output ranges within a specific trading period (for 50MW).

[0023] Table 1

[0024] As shown in Table 1, by assigning tiered price tags to forward load increases and reverse load decreases within different capacity ranges, the algorithm can accurately call the economic coefficients of the corresponding ranges when constructing the objective function, providing high-precision underlying boundary parameters for bidirectional collaborative optimization of source and load. Simultaneously, the system synchronously acquires dynamic operating parameters of generator units (such as unit ramp rate, minimum start-up and shutdown time, etc.) and the physical structure diagram of the power grid topology (such as node admittance matrix, branch power flow safety margin, etc.) through the dedicated power dispatch network.

[0025] Furthermore, based on the multi-segment tiered pricing data, the economic coefficients for increased power output corresponding to the physical power output variable and the economic coefficients for decreased power output corresponding to the physical power output variable are generated. The dynamic operating parameters, the power grid topology physical structure diagram, and the economic coefficients for increased and decreased power output are normalized and combined to form the global market boundary parameter set. Because generator operating parameters (represented by megawatt-level power constraints), power grid topology parameters (represented by per-unit impedance parameters), and economic coefficients (represented by marginal costs in monetary units) have significant dimensional differences and large numerical variations, directly concatenating these multidimensional heterogeneous physical and economic parameters into the subsequent control matrix will lead to severe data bias during the energy physics model solution process, making the model prone to getting trapped in local optima or experiencing gradient explosion. Therefore, this embodiment uses an extreme value normalization method to uniformly map the feature parameters of different dimensions to a dimensionless feature space of [0, 1]. The calculation formula for the normalization process is as follows: (1) in, These are dimensionless parameter values ​​generated after normalization. The actual values ​​of the original input parameters, such as the declared price at a specific node or the unit's output limit; This refers to the minimum lower limit of this type of parameter allowed in the entire network market environment or historical operating data of the equipment; This represents the maximum allowed upper limit of this type of parameter under the constraints of the entire network's physical boundaries. After this mathematical transformation, all data with eliminated dimensional differences are horizontally concatenated into a feature vector, which is then combined to form the global market boundary parameter set.

[0026] Step S2: The joint clearing objective function, the parameter set, and the network physical constraints containing state mutual exclusion logic are fused into a system state control matrix.

[0027] It should be noted that traditional day-ahead power clearing methods often treat economic cost optimization and complex network physical operational boundaries as isolated modules, iteratively solving them alternately. When faced with massive amounts of heterogeneous source-load data and high-concurrency response requests, this fragmented modeling approach is prone to logical oscillations in the solutions generated by multidimensional variables during optimization iterations, and may even lead to local infeasible solution spaces where actual dispatch commands cannot be issued. This step, by constructing a high-dimensional system state control matrix, deeply integrates discrete equipment start-up and shutdown and state switching logic, continuous power regulation capabilities, and nonlinear economic operating costs at the tensor level. This achieves a rigorous dimensionality reduction transformation from multi-objective discrete constraints to a unified matrix space, thereby significantly improving the numerical stability and convergence accuracy of subsequent global optimization model calculations.

[0028] In this embodiment, the joint clearing objective function includes the sum of the total operating cost of the system's physical generator units and the marginal cost of adjusting the load increase / decrease across the entire network. The joint clearing objective function is configured as follows: when a local node meets a preset condition for exceeding the marginal cost limit, a first priority weight is assigned to the physical power reduction variable in the network physical constraints; when the system's marginal clearing cost is lower than a preset threshold, a second priority weight is assigned to the physical power increase variable in the network physical constraints. The specific formula expression for the joint clearing objective function with dynamic penalty weights is as follows: (2) in, This represents the total operating cost of the system within the global scheduling cycle; T is the total scheduling time period set, t is the specific time period; G is the generator set set, g is the specific generator set; The actual output of generator unit g during time period t is a continuous variable. , , These are the secondary, primary, and constant consumption cost coefficients of the generator set, respectively; U represents the set of market-oriented users participating in the market response, and u represents a specific user. and These are the multi-stage tiered marginal cost coefficients for physical reduction and increase in output force on the user side, respectively. and These are the actual scheduling values ​​of the physical reduction and physical increase forces for system decision-making, respectively. As the first priority weight, This is the second priority weight.

[0029] In the control logic of this objective function for the optimization direction, when the marginal cost of a local node exceeds the preset upper limit due to transmission corridor blockage or extreme power shortage, the control system automatically dynamically increases the Lagrange multiplier. (For example, by assigning it a higher penalty payoff coefficient), forcing the solver's optimization gradient to slide rapidly in the direction of increasing physical load reduction, in order to prioritize load reduction to alleviate power supply shortages; conversely, when a surge in renewable energy generation leads to oversupply in the system and a sharp drop in marginal liquidation costs below a preset trough threshold, the system is activated and significantly increased. The algorithm is guided to abandon the conventional strategy of reducing power output and prioritize scheduling users with adjustment capabilities to "fill the valley" and increase power consumption, so as to absorb the excess power at the lowest economic cost.

[0030] Furthermore, in the network physical constraints containing state mutual exclusion logic, the state mutual exclusion hard constraint model for any market user u in time period t is specifically expressed as follows: (3) in, This refers to the control state variable corresponding to the physical output force variable, and it is a binary variable; This refers to the control state variable corresponding to the physical output variable, and it is a binary variable. At the microscopic physical level of actual power system operation, the load response trajectory of electrical equipment must have a uniquely definite direction. If the aforementioned 0-1 binary mutual exclusion constraint is missing from the control matrix, when the electricity spot market encounters extreme negative electricity prices or violent price fluctuations caused by drastic fluctuations in new energy sources, the underlying mathematical solution engine is highly susceptible to falling into a simple numerical arbitrage trap. For example, the algorithm might, in order to forcibly balance the supply and demand equations in a purely mathematical sense and seek to maximize the objective function's profit, directly output the absurd physical instruction to the grid dispatch system: "requiring the same electrolytic aluminum smelter to both significantly increase its electricity consumption by 50 MW and simultaneously significantly decrease its electricity consumption by 40 MW within the same time period." Introducing this inequality constraint strictly ensures that any load node, at the same time point, can only be clearly in one of three mutually orthogonal physical states: "purely increasing output," "purely decreasing output," or "maintaining the baseline unchanged (both are 0)."

[0031] In addition, the network physical constraints also include hard constraints on the demand-side physical response boundary: (4) (5) in, For physical output force variables, This adds a physical force variable; The maximum allowable reduction in output force deviation rate, The maximum allowable output force deviation rate; This is the baseline predicted load extracted from the load increment reporting parameters.

[0032] After obtaining the joint clearing objective function and various network physical constraints, to meet the parallel tensor computation requirements of subsequent underlying algorithms, the system unifies the dimensionality reduction of discrete logic and continuous numerical values, fusing them into a system state control matrix. Specifically, the system combines the continuous variables of generator output, the continuous variables of load increase / decrease output, and the binary discrete variables of user state to form a global state column vector. Subsequently, the economic coefficients in the objective function are extracted to construct the target return row vector. The linearized partial derivative coefficients of the network physical boundary constraints and the mutual exclusion logic hard constraints are extracted to construct the physical constraint matrix. With logical constraint matrix The final fused system state control matrix The mathematical expression is as follows: (6) in, This is the fused system state control matrix; The economic parameter vector characterizing the gradient of the joint clearing objective function; It is a physical matrix that includes the boundary coefficients of unit ramping, network power flow, and demand-side physical response; To extract from Boolean coefficient matrix of discrete-state mutually exclusive logic; and These are dynamic penalty multipliers used to adjust the influence weights of the physical constraint submatrix and the logical constraint submatrix on the global system state, respectively.

[0033] This step establishes a complete control system by deeply coupling and encapsulating the multidimensional economic objective matrix, the Boolean state mutual exclusion space, and the continuous physical capacity boundary. This mutual exclusion response logic, which integrates discrete physical constraints with continuous economic optimization, completely eliminates ill-conditioned degradation and multi-directional conflict oscillations in the algorithm solution process from the underlying mathematical mechanism, fundamentally ensuring that the final day-ahead clearing dispatch instructions issued by the power grid dispatch center have absolute physical executability and security robustness.

[0034] Step S3: Map the discrete decision variables and corresponding constraints in the control matrix into an energy physics model, and use the state solution of the model in the lowest energy ground state as the unit and load combination state reference.

[0035] It should be noted that traditional day-ahead clearing methods for electricity spot markets based on mixed-integer linear programming (MILP) are prone to the "curse of dimensionality" caused by combinatorial explosion when dealing with a large number of discrete mutually exclusive variables, including thousands of generator start-ups and shutdowns and load-side increase / decrease responses. Conventional branch-and-bound methods often become excessively time-consuming or even fail completely due to solver memory overflow when solving such non-convex and discontinuous problems. This step innovatively introduces the Ising model concept from statistical physics, transforming the complex discrete power grid scheduling constraint problem into an energy minimization problem of the physical system.

[0036] In this embodiment, mapping the discrete decision variables and corresponding constraints in the control matrix to an energy physics model includes: mapping the start-stop state of the generator set and the power output increase and decrease states of market users to spin node variables in a lattice network topology in a dimension reduction manner; transforming the topological security boundary and resource constraint boundary in the network physics constraints into interaction Hamiltonian energy penalty terms between adjacent spin node variables; and obtaining a global objective energy function characterizing the system state control matrix by superimposing the intrinsic energy of the spin node variables with the interaction Hamiltonian energy penalty terms.

[0037] Specifically, the system state control matrix constructed in the aforementioned steps is first... Discrete state control variable vector (Including the unit's binary start / stop variables and the user's) and (etc.) are mapped to the spin node vector of the physical system through affine transformation. Its variable dimensionality reduction mapping formula is: (7) in, For matrix Any i-th binary discrete decision variable in the dataset; For the corresponding spin node variables after mapping, the power-on or response state ( ) corresponds to spin up ( ), stopped or unresponsive status ( ) corresponds to spin down ( Subsequently, in order to transfer the system state control matrix... The system is completely equivalently transformed into a physical energy expression, and a Hamiltonian model is constructed based on the theory of quadratic unconstrained binary optimization (QUBO). (8) in, Characterization matrix The global objective energy function; and These are the topologically related edges and the set of nodes, respectively. The coefficient of the Hamiltonian energy penalty term for interactions between nodes; Let be the eigenenergy external field bias coefficient of node i, which is given by matrix . The target profit row vector It is derived from this transformation. To achieve a rigorous transformation from the control matrix to the energy field, specific measures are taken for the matrix... Network physical constraints Using the quadratic penalty term This is transformed into an energy minimization problem, where B is the boundary constant vector corresponding to the physical constraints, representing rigid physical boundaries such as the maximum transmission capacity limit of the line, node load demand, and upper and lower limits of unit output; this is addressed by the state-mutually exclusive logic matrix. middle The hard constraint is transformed into a quadratic product penalty term. This ensures that a large energy penalty occurs when both are 1. Substituting the aforementioned mapping formula and expanding the polynomial, the coefficients of the energy penalty term of the interaction Hamiltonian are... The specific generation formula is as follows: (9) in, and For control matrix Dynamic penalty multipliers inherited in the middle; The coupling element in the i-th row and j-th column extracted after transposing and multiplying the physical constraint matrix represents the physical energy limit violation penalty caused by the power grid topology; This is the Boolean crossover penalty coefficient directly generated by the state mutual exclusion logic. Through the above rigorous algebraic isomorphism mapping, any erroneous scheduling combination that leads to excessive unit output or simultaneous increases or decreases in user output will be due to... The amplification effect causes the virtual physical energy of the local spin system to rise sharply, which is then naturally rejected by the system in subsequent evolution.

[0038] Further, the step of using the state solution of the model in the lowest energy ground state as the unit and load combination state reference includes: initializing the initial ambient temperature field and annealing time step of the simulated annealing algorithm; inputting the energy physics model containing the global objective energy function into the simulated annealing algorithm, reducing the ambient temperature field by introducing random disturbance terms and according to a preset decay rate; when the system evolution reaches the preset minimum temperature termination condition, extracting the global minimum energy distribution state of the converged energy physics model at this time, and decoding it into the unit and load combination state reference containing equipment start-up and shutdown control commands. In the specific algorithm execution process, the system first initializes the control parameters of the simulated annealing engine. To ensure that the system has sufficient thermodynamic fluctuation kinetic energy to escape the local energy optimum trap in the early stage of evolution, the initial ambient temperature field parameters are set as follows: Set the cooling decay rate corresponding to the annealing time step. During the evolution process, Markov chain Monte Carlo (MCMC) random fluctuations are introduced as random perturbations for spin flipping, and the Metropolis probability acceptance criterion is used to evaluate the system's state transitions. The ambient temperature field changes with time step k according to... An exponential smooth decay is performed. In this embodiment, the minimum temperature termination condition determination logic is specifically configured as follows: when the ambient temperature field drops to... If, within 500 consecutive iteration steps, the total rate of change of the global objective energy function of the system is less than 0.001%, then the physical evolution system is determined to have entered a "completely frozen" thermodynamic equilibrium state.

[0039] Figure 2 This is a simulation diagram illustrating the global target energy convergence evolution curve obtained by solving a simulated annealing method for an energy physics model containing 1500 discrete nodes under a typical scheduling day scenario in this embodiment. Figure 2 As shown, the horizontal axis represents the annealing evolution time step (MCMC iteration count), and the vertical axis represents the normalized global target energy of the system. In the initial high-temperature phase (iterations 0-2000), the system energy exhibits dramatic thermodynamic fluctuations, manifested as wide oscillations around the curve, which provides the algorithm with the kinetic energy to escape the trap of high-dimensional local optima. As the ambient temperature field decays exponentially (iterations 2000-8000), spin-flipping violations (such as breaches of unit output constraints) are subject to high penalties. With strict suppression, the total energy of the system rapidly decreases in a stepwise manner and undergoes continuous ground state collapse; finally, when the iteration reaches around 10,000 times, the energy curve smoothly tends to the horizontal minimum value, and the evolving system is completely frozen.

[0040] At this point, the final spin arrangement matrix of each microparticle in the entire lattice network is extracted as the global lowest energy distribution state. Subsequently, the spin matrix of the {-1,+1} configuration is accurately restored to the unit and load combination state reference composed of 0s and 1s using the inverse mapping decoding formula. The inverse mapping decoding formula is specifically as follows: (10) in, The final spin-optimal state variable of the i-th microscopic particle extracted when the system reaches thermodynamic equilibrium is given by a value of +1 or -1. The discrete optimal control command for the i-th device generated after decoding is a matrix element of the unit and load combination state reference, with a value of 1 (indicating the device is turned on or responds) or 0 (indicating the device is turned off or does not respond).

[0041] This step utilizes the idea of ​​natural collapse of the energy ground state to establish an isomorphic mapping between discrete mutual exclusion constraints and continuous physical energy. It completely breaks through the local deadlock and NP-hard dilemma that traditional commercial solvers are prone to when facing massive 0-1 mutual exclusion variables. It effectively resolves the combinatorial explosion problem faced by electricity spot clearing, significantly reduces the computational complexity of the model in terms of computing power solution, and greatly improves the solution efficiency and stability of multi-variable collaborative optimization for large-scale power grids.

[0042] Step S4: Generate a scheduling strategy based on the state benchmark, and construct a dynamic evolution system containing scheduling strategy nodes and uncertain disturbance nodes.

[0043] It should be noted that traditional day-ahead clearing methods for electricity spot markets typically employ robust optimization approaches when dealing with uncertainties in renewable energy sources such as wind and solar power. This traditional method relies on rigid, static minima-maximum reversal planes. To withstand extreme but rare "worst-case scenarios," it often reserves large and conservative reserve capacities, resulting in significant idle grid regulation capacity during normal periods and extremely poor economic efficiency. This step breaks away from the mindset of static worst-case scenarios, treating fluctuations in source and load power and the response of dispatch resources as dynamic ecological processes that evolve and mutually constrain each other in nature. This allows for a smoother and more flexible approach to balancing grid security and economy.

[0044] In this embodiment, constructing a dynamic evolutionary system containing scheduling strategy nodes and uncertainty disturbance nodes includes: defining the scheduling strategy as a controlled variable in the dynamic evolutionary system, and defining the acquired source load power uncertainty disturbance as an environmental excitation variable in the dynamic evolutionary system; constructing a set of differential equations characterizing the dynamic nonlinear coupling relationship between the controlled variable and the environmental excitation variable through a population symbiotic evolution model. The evolutionary system is constructed by deeply coupling the spatial topological physical network containing various scheduling strategy nodes and uncertainty disturbance nodes with the set of differential equations characterizing the nonlinear coupling relationship in the spatiotemporal dimension.

[0045] Specifically, the dynamic evolution system described in this embodiment is an equivalent mapping sandbox composed of a power grid spatial topology network (structural skeleton) and a nonlinear time evolution law (driving engine). At the static spatial topology level, the system maps grid-connected wind farms, photovoltaic power station access buses, and large load fluctuation distribution network access points in the actual power grid as the "uncertainty disturbance nodes"; simultaneously, it maps generator unit nodes in the operating state extracted based on the aforementioned combined state benchmark and market-based user nodes with increase / decrease response capabilities as the "scheduling strategy nodes". The above two types of nodes are connected through transmission line branches to construct the topological physical boundary of the system. On this topological skeleton, the system extracts the continuous flexible adjustment capacity inherent in the equipment in the active ground state as a controlled variable (i.e., scheduling strategy). At the dynamic temporal evolution level, the system draws on the Lotka-Volterra predator-prey symbiotic evolution equation from theoretical ecology, treating the power fluctuations generated by uncertain disturbance nodes as continuously proliferating "prey," and the absorption capacity released by scheduling strategy nodes as "predators" tracking and mitigating disturbances. A system-level dynamic nonlinear coupled differential equation set is established, the specific mathematical expression of which is as follows: (11) (12) in, The environmental excitation variable at time t represents the overall power uncertainty disturbance caused by the prediction deviation of wind and solar power output and load fluctuations in the entire network; Let be a controlled variable at time t, representing the available flexible scheduling policy capacity activated by the scheduling policy node; This represents the natural diffusion growth rate of the uncertainty disturbance under uncontrolled conditions. Environmental carrying capacity is the physical limit of the maximum power carrying capacity of a critical section of the power grid in a power system, which limits the endless spread of disturbances beyond the limit. This is the damping coefficient of the scheduling strategy for absorbing and mitigating uncertainties, reflecting the response damping of resources; The inherent standby cost or attenuation rate required to maintain the scheduling strategy; The conversion response coefficient, representing the sensitivity of the power grid frequency regulation or control system, is the factor that triggers an increase in the capacity of the dispatching strategy due to uncertainties and disturbances. Under the control physics logic of this set of differential equations, when sudden and drastic fluctuations in new energy sources lead to... When rising rapidly, the controlled variable It will be greatly stimulated and accelerated under its drive; with the adjustment capacity As the value increases, its inverse suppression effect on the disturbance term gradually becomes dominant, thus suppressing the disturbance; when the disturbance subsides, the equation automatically guides... The price will fall back to a low level to avoid wasting costs by maintaining redundancy reserves.

[0046] This step constructs a dynamic evolutionary system that integrates spatial physical nodes and the law of temporal symbiosis. It replaces the rigid and conservative minimax cutting plane confrontation in traditional robust optimization with a dynamic ecological evolution model. This allows the power system to find the symbiotic point on the time scale in the dynamic game between uncertain disturbances and scheduling resources. This not only ensures the dynamic stability of the system in the face of extreme fluctuations in mathematical principles, but also greatly releases the hidden flexible adjustment resources of the power grid, avoiding unnecessary idleness and waste of capacity.

[0047] Step S5: Using preset differential equation constraints, the system evolves to a stable limit cycle state in phase space, generating the final day-ahead clearing plan and nodal marginal electricity price.

[0048] It should be noted that traditional methods for solving electricity spot clearing (such as the interior-point method or the simplex method) often output a static, deterministic optimal solution. However, in new power systems with a high proportion of renewable energy sources such as wind and solar power, source-load fluctuations are continuous and severe. Static optimal solutions are prone to failure when faced with actual disturbances, leading to frequent restarts of the dispatch system, which consumes significant computing power and threatens grid security. This step introduces the limit cycle theory from nonlinear dynamics, moving away from pursuing a single vulnerable optimal point and instead seeking a dynamically stable interval that can accommodate disturbance fluctuations. This allows the power system to spontaneously maintain a safe periodic oscillation orbit when facing uncertainties within a certain range, relying on its flexible adjustment resources, thereby significantly improving the robustness of the clearing plan.

[0049] In this embodiment, the process of causing the system to evolve to a stable limit cycle state in phase space through preset constraints of a set of differential equations includes: performing time-series iterative solutions to the set of differential equations under the boundary constraints of the power grid cross-section transmission capacity, which includes a power transfer distribution factor; monitoring the evolution trajectories of the controlled variables and environmental excitation variables until the evolution trajectories converge to a closed, disjoint stable limit cycle in phase space; and extracting the coordinate eigenvalues ​​of the central gravity point of the stable limit cycle and mapping them to a set of clearing scheduling instructions with robust safety margins. Specifically, the system uses time as the evolution step and employs the fourth-order Runge-Kutta method to perform high-precision numerical integration on the set of differential equations constructed in the aforementioned steps. During the iteration process, the system monitors in real time the two-dimensional phase space state vector composed of environmental excitation variables (comprehensive power disturbances) and controlled variables (scheduling strategy capacity). The trajectory of the motion. According to the Poincaré-Bendixson Theorem, the geometric criterion for the generation of a stable limit cycle in a system is: the evolution trajectory of the state vector no longer diverges to infinity in phase space, nor collapses into a singularity, but eventually converges into a closed, non-self-intersecting phase with a fixed evolution period. Circular orbit In the middle, that is, satisfying .

[0050] Figure 3 The simulation diagram of the stable limit cycle evolution trajectory of the aforementioned dynamic evolution system in two-dimensional phase space is shown. Figure 3 As shown, the horizontal axis The vertical axis represents the power variable of the uncertainty disturbance. The capacity variable represents the scheduling policy triggered by the response. The dashed lines in the diagram represent the transient evolution trajectory of the system starting from four different severe initial conditions (including boundary scenarios such as sudden maximum disturbances and extreme lack of backup). The solid closed loop in the middle represents the final convergent stable limit cycle. It can be clearly observed that regardless of the initial non-equilibrium state of the system in phase space, under the guidance of the "predator-repulsion" constraint of the cross-coupled differential equations, all evolution trajectories eventually converge to the same closed orbit. This physically proves that the system can absorb and digest the impact of various high-dimensional source load fluctuations through the flexible dynamic release and recycling of resources.

[0051] Once the trajectory convergence to the stable limiting cycle is detected, the system will perform time-averaged integration on the circular trajectory over a complete cycle to calculate and extract the eigenvalues ​​of the geometric center gravitational point coordinates of the limiting cycle. The coordinate vector of the central gravitational point... The specific calculation formula is as follows: (13) in, The coordinate eigenvector of the central gravitational point of the stable limit cycle; This is the stable evolution period of the closed limit loop orbit; For a closed integral path of a limit cycle in phase space; Let t be the transient phase space coordinates of the system at time t. The system uses the extracted coordinate features of the central gravitational point as the optimal reference point, and maps them back to the power dimension of the physical power grid to generate the clearing and scheduling instruction set with robust safety margin. This instruction set represents the "optimal equilibrium stator" of the system in a dynamic adversarial environment, enabling the power grid to safely "swish back" to equilibrium based on the natural response of flexible resources, regardless of the direction of power disturbance during actual operation.

[0052] Furthermore, after generating the final pre-emptive clearing plan and the node marginal price, the process also includes performing electricity data settlement: extracting the system energy price component, network congestion price component, and marginal line loss price component from the node marginal price; generating electricity settlement data for each node based on the physical increase / decrease in the final pre-emptive clearing plan and the node marginal price corresponding to each node, combined with the user's benchmark forecast load data; the electricity settlement data includes: first settlement data calculated based on the forecast price for the electricity corresponding to the benchmark forecast load data; and second settlement data calculated based on the difference between the corresponding node marginal price and the declared price for the electricity corresponding to the physical increase / decrease in output. In a specific electricity market transaction settlement system, the Locational Marginal Pricing (LMP) is precisely decoupled into three parts: namely... Each corresponds to a unified energy benchmark price across the entire network. The congestion penalty price for the topology location of this node. And the marginal physical network loss price caused by line power transmission Based on this refined electricity pricing, the system generates a dual-track financial settlement bill for any participating market user u. The specific total financial cost for electricity settlement is detailed below. The calculation formula is as follows: (14) in, This represents the user's total settlement fee during time period t; the first term on the right side of the equation, within square brackets, represents the first settlement data. The electricity consumption corresponding to the baseline predicted load, The benchmark forecast price is the price stipulated in the medium- to long-term contract; the second term on the right side of the equation, within square brackets, represents the second liquidation data. and These represent the actual physical increase and decrease in power output as per the final clearing plan; Let $t$ be the marginal electricity price of the node during time period $t$. and These represent the tiered bids for increased and decreased power output initially submitted by the user. Using this formula, the baseline load is settled smoothly according to a fixed agreement, while the deviation in power output caused by the user responding to dispatch instructions is compensated as an incentive through the difference between the actual node value (LMP) and the expected cost (bid price) (i.e., market surplus).

[0053] This step provides a dynamic and elastic equilibrium state of the power system in the face of high-dimensional random uncertainty through the limit cycle solution method, which solves the vulnerability of traditional rigid dispatch from the underlying physical laws. At the same time, the refined settlement mechanism based on the marginal electricity price of the nodes essentially transforms the economic boundary conditions into control signals that drive the physical response of demand-side loads, completing the technical control closed loop from the issuance of power dispatch instructions to the precise execution of physical equipment. This ensures that the method provided by this invention can substantially improve the physical allocation efficiency of peak-shaving resources across the entire network and the overall operational reliability of the power grid.

[0054] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0055] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0056] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0057] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0058] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0059] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A day-ahead clearing method for electricity spot markets that considers load increment reporting constraints, characterized in that, include: Obtain the load increment declaration parameters, and construct a market boundary parameter set containing physical increase output variables and physical decrease output variables based on the declaration parameters; The joint clearing objective function, the parameter set, and the network physical constraints containing state mutual exclusion logic are integrated into a system state control matrix. Mapping the discrete decision variables and corresponding constraints in the control matrix into an energy physics model includes: reducing the dimension of the generator start-up and shutdown states and the power output increase and decrease states of market users to spin node variables in a lattice network topology; transforming the topological security boundary and resource constraint boundary in the network physics constraints into interaction Hamiltonian energy penalty terms between adjacent spin node variables; and obtaining the global objective energy function characterizing the system state control matrix by superimposing the intrinsic energy of the spin node variables and the interaction Hamiltonian energy penalty terms. The solution of the model in the lowest energy ground state is used as the unit and load combination state reference, including: initializing the initial ambient temperature field and annealing time step of the simulated annealing algorithm; inputting the energy physics model containing the global objective energy function into the simulated annealing algorithm, and reducing the ambient temperature field by introducing random disturbance terms and according to a preset decay rate; when the system evolution reaches the preset minimum temperature termination condition, extracting the global minimum energy distribution state of the energy physics model at this time, and decoding it into the unit and load combination state reference containing equipment start-up and shutdown control commands; Based on the state baseline, a scheduling strategy is generated, and a dynamic evolution system containing scheduling strategy nodes and uncertainty disturbance nodes is constructed, including: defining the scheduling strategy as a controlled variable in the dynamic evolution system, defining the acquired source load power uncertainty disturbance as an environmental excitation variable in the dynamic evolution system; and constructing a set of differential equations characterizing the dynamic nonlinear coupling relationship between the controlled variable and the environmental excitation variable through a population co-evolution model. The system is induced to evolve into a stable limit cycle state in phase space by using a set of pre-defined differential equations as constraints, generating a final day-ahead clearing plan and nodal marginal electricity prices. This process includes: performing time-series iterative solutions to the differential equations under the boundary constraints of the grid cross-section transmission capacity, which includes a power transfer distribution factor; monitoring the evolution trajectories of the controlled variables and environmental excitation variables until the evolution trajectories converge into a closed, disjoint stable limit cycle in phase space; and extracting the coordinate eigenvalues ​​of the central gravitational point of the stable limit cycle and mapping them to a clearing scheduling instruction set with robust safety margins.

2. The method according to claim 1, characterized in that, The construction of a market boundary parameter set based on the declared parameters, including physical increase and decrease force variables, includes: Extract the baseline predicted load data, the upper limit of the declared capacity for increasing or decreasing output, and the corresponding multi-tiered pricing data from the load increment declaration parameters; Obtain the dynamic operating parameters of the generator set and the physical structure diagram of the power grid topology; Based on the multi-stage tiered pricing data, generate the economic coefficients for the increase or decrease in output corresponding to the physical increase or decrease in output variables; The dynamic operating parameters, the power grid topology physical structure diagram, and the economic coefficients for increasing or decreasing power output are normalized and combined to form the market boundary parameter set.

3. The method according to claim 1, characterized in that, The joint clearing objective function includes the sum of the total operating cost of the system's physical generator units and the marginal cost of adjusting the overall network load increase or decrease; The joint clearing objective function is configured as follows: when a local node meets the preset condition of marginal cost exceeding the limit, the physical reduction force variable is assigned a first priority weight in the network physical constraint condition; when the system marginal clearing cost is lower than a preset threshold, the physical increase force variable is assigned a second priority weight in the network physical constraint condition.

4. The method according to claim 1, characterized in that, In the network physical constraints containing state mutual exclusion logic, the specific representation of the state mutual exclusion hard constraint model for any market user u in time period t is as follows: in, This refers to the control state variable corresponding to the physical output force variable, and it is a binary variable; It is the control state variable corresponding to the physical output force variable, and it is a binary variable.

5. The method according to claim 4, characterized in that, The network physical constraints also include hard constraints on the demand-side physical response boundary: in, For physical output force variables, This adds a physical force variable; The maximum allowable reduction in output force deviation rate, The maximum allowable output force deviation rate; This is the baseline predicted load extracted from the load increment reporting parameters.

6. The method according to claim 1, characterized in that, After generating the final pre-emptive clearing plan and nodal marginal electricity price, the process also includes performing electricity data clearing operations: Extract the system energy price component, network congestion price component, and marginal line loss price component from the marginal electricity price of the node; Based on the physical increase or decrease in power output in the final clearing plan and the corresponding marginal electricity price of each node, combined with the user's benchmark predicted load data, the power clearing data of each node is generated. The electricity settlement data includes: first settlement data calculated based on the forecast price for the electricity corresponding to the benchmark forecast load data; The second settlement data is calculated based on the difference between the marginal electricity price and the declared price at the corresponding node, for the electricity volume corresponding to the physical increase or decrease in output value.

Images