CN-122022416-A - Power resource scheduling method, system and storage medium embedded with time sequence coupling constraint

CN122022416ACN 122022416 ACN122022416 ACN 122022416ACN-122022416-A

Abstract

The application provides a power resource scheduling method, a system and a storage medium for embedding time sequence coupling constraint, which relate to the technical field of energy management and comprise the steps of obtaining a classical safety constraint economic scheduling model, carrying out recursive computation on the classical safety constraint economic scheduling model, wherein parameters of a neural network at each time point in a forward propagation process of the recursive computation process of the set neural network model are kept consistent, simultaneously, the output of a current time step computing unit is used as the input of a corresponding computing unit of the next time step to be connected with each computing unit, constraint conditions are embedded into the output end of the set neural network model, a gradient return information expressway is constructed, the set neural network model is trained by utilizing power training data, and the parameters of the neural network are adjusted to obtain the power scheduling model so as to output a power resource scheduling scheme. The application obviously improves the calculation efficiency through the coupling calculation unit, the parameter sharing and the recursion structure, and can quickly generate the power resource scheduling scheme.

Inventors

HAN JIAYU
YAN LEI
LI ZUYI

Assignees

浙江工业大学

Dates

Publication Date: 20260512
Application Date: 20260415

Claims (10)

1. The power resource scheduling method embedded with the time sequence coupling constraint is characterized by comprising the following steps of: acquiring a classical safety constraint economic dispatch model corresponding to the electric power resource; The method comprises the steps of carrying out recursive calculation on a classical safety constraint economic dispatch model by using a set neural network model, wherein parameters of a neural network at all time points in a forward propagation process of the set neural network model recursive calculation process are kept consistent, and simultaneously, taking output of a current time step calculation unit as input of a next time step corresponding calculation unit so as to connect all calculation units of the set neural network model; embedding constraint conditions of the classical safety constraint economic dispatch model into an output end of the set neural network model, and constructing a gradient return information expressway without passing through a neural network passage; Training the set neural network model by utilizing electric power training data, and adjusting the neural network parameters through a back propagation algorithm to obtain an electric power dispatching model; and outputting a power resource scheduling scheme based on the power scheduling model and the real-time power resource data.
2. The power resource scheduling method according to claim 1, wherein obtaining a security constraint scheduling model corresponding to a power resource includes: Constructing an objective function for minimizing the total power generation cost; the constraint conditions comprise power balance constraint, generator active output constraint, climbing constraint and thermal stability limit constraint of the power transmission line.
3. The power resource scheduling method according to claim 2, characterized by embedding constraints of the classical safety-constrained economic scheduling model into the computing unit in the set neural network model, further comprising: And generating a control factor through a neural network, wherein the control factor is used for indicating that the output of the power scheduling model meets the constraint condition.
4. The power resource scheduling method of claim 1, wherein generating the control factor through the neural network comprises: calculating the maximum possible output of the generator in the current time step, wherein the maximum possible output is a smaller value of the maximum output of the generator and the climbing capacity of the first unit, and the climbing capacity of the first unit is the sum of the output of the generator in the last time step and the allowable climbing amount; calculating the minimum possible output of the generator at the current time step, wherein the minimum possible output is a larger value of the minimum output of the generator and the climbing capacity of the second unit, and the climbing capacity of the second unit is the difference between the output of the generator at the last time step and the allowable climbing amount; calculating the actual active power output of the generator at the current time step, wherein the actual active power output is between the minimum possible output and the maximum possible output and is regulated by a control factor generated by a loss function; And calculating the relaxation power of the current time step, wherein the relaxation power is the sum of all load demands minus the sum of the output forces of all locked generators.
5. The power resource scheduling method of claim 4, wherein adjusting the neural network parameters by a back propagation algorithm comprises: determining association relation features between adjacent time steps according to the maximum possible output, the minimum possible output and a control factor; and generating gradient information of the loss function corresponding to the set neural network model according to the information characteristics of the time steps and the association relation characteristics between the adjacent time steps.
6. The power resource scheduling method of claim 5, further comprising: calculating a first partial derivative between the maximum possible output and the generator output of the previous time step; calculating a second partial derivative between the minimum possible output and the generator output of the previous time step; and calculating the control factor according to the difference between the first partial derivative and the second partial derivative.
7. A power resource scheduling system embedded with timing coupling constraints, comprising: The scheduling model acquisition module is used for acquiring a classical safety constraint economic scheduling model corresponding to the electric power resource; the system comprises a setting neural network model, a recursive computation module, a load demand and a generator active power output module, wherein the setting neural network model is used for carrying out recursive computation on the classical safety constraint economic dispatch model, parameters of the neural network at all time points in the forward propagation process of the setting neural network model recursive computation process are kept consistent, and meanwhile, the output of a current time step computation unit is used as the input of a next time step corresponding computation unit so as to be connected with all computation units of the setting neural network model; the constraint condition setting module is used for embedding constraint conditions of the classical safety constraint economic dispatch model into the output end of the set neural network model and constructing a gradient return information expressway without passing through a neural network passage; The model training module is used for training the set neural network model by utilizing the electric power training data, and adjusting the neural network parameters through a back propagation algorithm to obtain an electric power dispatching model; and the power resource scheduling module is used for outputting a power resource scheduling scheme based on the power scheduling model and the real-time power resource data.
8. An electronic device, comprising: A memory for storing a computer program; a processor for implementing the steps of the method according to any one of claims 1 to 6 when said computer program is executed.
9. A computer readable storage medium, characterized in that it has stored thereon a computer program which, when executed, implements the steps of the method according to any of claims 1 to 6.
10. A computer program product comprising a computer program which, when executed, implements the steps of the method according to any one of claims 1 to 6.

Description

Power resource scheduling method, system and storage medium embedded with time sequence coupling constraint Technical Field The application relates to the technical field of energy management, in particular to a power resource scheduling method, a power resource scheduling system, a power resource scheduling medium, a power resource scheduling device and a power resource scheduling program product, wherein the power resource scheduling system is embedded with time sequence coupling constraints. Background Along with the promotion of novel power system construction, new energy gradually becomes dominant power, and the complex variability of its output makes the degree of difficulty that keeps electric power system electric power electric quantity balanced increase, and the frequency of calling of Safe Constraint Economic Dispatch (SCED) constantly improves, has put forward higher demands to its solution rate and result feasibility. However, the prior art has the defects that firstly, the training efficiency of deep reinforcement learning is low, and the model result is difficult to guarantee. The method is suitable for continuous sequence decision-making, but has the problems that a simulation environment is difficult to restore a complex real scene, the design of a reward function is difficult, the training efficiency is low due to unordered exploration mechanisms, and the like in practical application, the end-to-end deep learning model is mostly used for single-time scale optimal power flow, the classical safety constraint scheduling model which has multiple time scales and time sequence coupling constraint is difficult to process, the application range is limited, and the RNN model has gradient explosion/disappearance problems, so that model training fails. Therefore, there is a need for an efficient and high quality SCED solving method to meet the complex and variable operating requirements of new power systems. Disclosure of Invention The application aims to provide a power resource scheduling method, a power resource scheduling system, a computer-readable storage medium and electronic equipment, which are embedded with time sequence coupling constraint, and can effectively realize safe constraint economic scheduling of power resources. In order to solve the technical problems, the application provides a power resource scheduling method embedded with time sequence coupling constraint, which comprises the following specific technical scheme: acquiring a classical safety constraint economic dispatch model corresponding to the electric power resource; The method comprises the steps of carrying out recursive calculation on a classical safety constraint economic dispatch model by using a set neural network model, wherein parameters of a neural network at all time points in a forward propagation process of the set neural network model recursive calculation process are kept consistent, and simultaneously, taking output of a current time step calculation unit as input of a next time step corresponding calculation unit so as to connect all calculation units of the set neural network model; embedding constraint conditions of the classical safety constraint economic dispatch model into an output end of the set neural network model, and constructing a gradient return information expressway without passing through a neural network passage; Training the set neural network model by utilizing electric power training data, and adjusting the neural network parameters through a back propagation algorithm to obtain an electric power dispatching model; and outputting a power resource scheduling scheme based on the power scheduling model and the real-time power resource data. Optionally, the obtaining a security constraint scheduling model corresponding to the power resource includes: Constructing an objective function for minimizing the total power generation cost; the constraint conditions comprise power balance constraint, generator active output constraint, climbing constraint and thermal stability limit constraint of the power transmission line. Optionally, after embedding the constraint condition of the classical safety constraint economic dispatch model into the calculation unit in the set neural network model, the method further includes: And generating a control factor through a neural network, wherein the control factor is used for indicating that the output of the power scheduling model meets the constraint condition. Optionally, generating the control factor through the neural network includes: calculating the maximum possible output of the generator in the current time step, wherein the maximum possible output is a smaller value of the maximum output of the generator and the climbing capacity of the first unit, and the climbing capacity of the first unit is the sum of the output of the generator in the last time step and the allowable climbing amount; calculating the minimum possible output of the generator at