A box type economic operation domain generation method and device based on a pre-trained model
By using a pre-trained model-based approach and deep neural network to optimize parameters, the problem of slow speed in solving box-shaped economic operating domains by traditional mathematical optimization methods is solved, achieving efficient solution results.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHEJIANG UNIV
- Filing Date
- 2026-03-09
- Publication Date
- 2026-07-10
AI Technical Summary
Traditional mathematical optimization methods are slow to compute when solving the economic operating domain of a box-type grid, making it difficult to meet the solution requirements in large-scale power grid scenarios.
A pre-trained model-based approach is adopted. By acquiring power samples of new energy sources and loads in the power grid, a mathematical solution model is established. Deep neural networks are then used for supervised learning to optimize the parameters of the pre-trained model, thereby improving the solution speed and effectiveness.
While maintaining accuracy, it significantly improves the computational efficiency of the box-shaped economic operating domain, enabling it to quickly approximate constraints.
Smart Images
Figure CN122366716A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the technical field of smart grid and artificial intelligence, specifically relating to a method and apparatus for generating a box-type economic operation domain based on a pre-trained model. Background Technology
[0002] Currently, renewable energy sources, represented by wind power and photovoltaics, are being integrated into the power system on an unprecedented scale and with unprecedented penetration, achieving high-proportion large-scale application. Meanwhile, electric vehicles, as the core driving force for green transformation in the transportation sector, are experiencing explosive growth in scale, driven by both policy incentives and technological advancements. Their widespread application not only brings enormous carbon emission reduction potential but also profoundly reshapes load-side characteristics, giving them the complex attributes of both energy storage potential and concentrated charging load. While this situation injects strong momentum into the decarbonization of the energy structure, it also brings unprecedented new challenges to the safe, stable, and efficient operation of the traditional power system—a sharp increase in uncertainty on both the source and load sides. On the one hand, renewable energy generation on the supply side is characterized by significant intermittency, volatility, and poor predictability, resulting in strong randomness in both time and space dimensions of its output. On the other hand, with the widespread adoption of electric vehicles, their charging demand and charging / discharging behavior are highly random and dispersed, exhibiting spatiotemporal coupling characteristics in time, space, and power, making the traditional "source follows load" balance model unsustainable. This two-way uncertainty, characterized by "uncontrollable sources and unpredictable loads," is amplified throughout the entire power generation, grid, load, and storage chain, profoundly altering the dynamic characteristics of the power system. It places extremely high demands on system planning, dispatching, operation, security, and market mechanism design, necessitating the development of advanced flexible resources, intelligent forecasting technologies, and new operation and control systems to address these challenges and provide support for building a highly resilient new power system with a high proportion of renewable energy.
[0003] Patent document CN116029464B discloses an accelerated solution method for the precise box-type economic operating domain of a power grid, which characterizes the impact of source-load dual-side uncertainties on power grid dispatch using the form of a box-type economic operating domain. However, this economic operating domain is constructed using mathematical optimization theory, resulting in a slow solution speed in large-scale power grid scenarios. Summary of the Invention
[0004] To address the issue of long computation time in solving box-shaped economic operating domains using traditional optimization methods, the present invention aims to provide a box-shaped economic operating domain generation method and apparatus based on a pre-trained model, which improves both the speed and effectiveness of solving the box-shaped economic operating domain.
[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for generating a box-shaped economic operating domain based on a pre-trained model, comprising the following steps: Obtain samples of renewable energy output ranges and load power samples from the power grid, and establish a mathematical solution model for the box-type economic operation domain; Based on the new energy output range samples and load power samples, a mathematical solver is used to solve the mathematical solution model of the box-type economic operation domain to obtain the box-type economic operation domain corresponding to each sample; the new energy output range and load power are extracted as pre-training samples, and the corresponding box-type economic operation domain is extracted as pre-training labels. Supervised learning of the pre-trained model relative to the pre-trained labels is performed using pre-trained samples to optimize the parameters of the pre-trained model; The box-shaped economic operating domain is solved using a pre-trained model with optimized parameters.
[0006] Furthermore, the new energy output range includes wind power output range and photovoltaic power output range, and the load power is the active load of the power system; The mathematical solution model for the box-shaped economic operation domain is a two-layer optimization model. The lower-layer model is an economic dispatch model, and the upper-layer model is constructed with the goal of solving the operating boundaries of each unit under uncertain conditions. The upper-layer model is an economic operation boundary solution model, which describes the uncertainty of new energy output in the form of interval numbers.
[0007] Furthermore, the method also includes: normalizing the pre-training samples and pre-training labels, and using the normalized pre-training samples to perform supervised learning of the pre-training model relative to the normalized pre-training labels. Normalize the renewable energy output range samples, load power samples, and box-type economic operation domain calculation labels. Then, use the normalized renewable energy output range samples and load power samples to perform supervised learning on the box-type economic operation domain mathematical solution model relative to the normalized box-type economic operation domain calculation labels.
[0008] Furthermore, the step of using pre-trained samples to perform supervised learning of the pre-trained model relative to the pre-trained labels to optimize the pre-trained model parameters includes: The pre-trained model is built based on a deep neural network. The input is the uncertainty range of new energy output and the load power value. The output is a box-shaped economic operation domain, that is, the upper and lower bounds of the economic operation of each unit under uncertainty. When supervising the model, a pre-training loss function is constructed based on the difference between the lower and upper bounds of the output economic operation domain and the lower and upper bounds of the actual economic operation domain contained in the pre-training labels. At the same time, a penalty loss function is introduced to penalize when the lower bound of the economic operation domain output by the model is greater than the upper bound, so as to ensure that the upper bound of the economic operation domain is always greater than the lower bound. Finally, the pre-trained model is trained with the goal of minimizing the sum of the two loss functions.
[0009] Furthermore, the process of solving the box-shaped economic operating domain using the pre-trained model with optimized parameters includes: The predicted new energy output range and load power are input into the pre-trained model with optimized parameters to calculate the box-type economic operation domain of the power system. Among them, the upper and lower limits of the output of each unit in the power system constitute the box-shaped economic operation domain of the power system.
[0010] The present invention also provides a box-shaped economic operating domain generation device based on a pre-trained model, comprising one or more processors for implementing the box-shaped economic operating domain generation method based on the pre-trained model.
[0011] The present invention also provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, is used to implement the aforementioned method for generating a box-shaped economic operating domain based on a pre-trained model.
[0012] The beneficial effects of this invention are as follows: It proposes a box-shaped economic operating domain method based on a pre-trained model, which can effectively solve the problem of low computational efficiency in traditional mathematical optimization methods. The method is validated in an improved IEEE 9 and IEEE 39-node test system. Simulation results show that the method can effectively approximate the box-constrained economic operating domain, significantly improving computational efficiency while maintaining accuracy. Attached Figure Description
[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0014] Figure 1 A flowchart of a box-shaped economic operating domain generation method based on a pre-trained model provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of a box-shaped economic operating domain generation device based on a pre-trained model, provided in an embodiment of the present invention. Detailed Implementation
[0015] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the scope of protection of this invention.
[0016] like Figure 1 As shown, this embodiment of the invention provides a method for generating a box-shaped economic operating domain based on a pre-trained model, which improves the solution speed and the effectiveness of the solution.
[0017] We obtain samples of renewable energy output ranges and load power samples from the power grid, and establish a mathematical solution model for the box-type economic operation domain.
[0018] Based on the new energy output range samples and load power samples, a mathematical solver is used to solve the mathematical solution model of the box-type economic operation domain, and the box-type economic operation domain corresponding to each sample is obtained; the new energy output range and load power are extracted as pre-training samples, and the corresponding box-type economic operation domain is extracted as pre-training labels.
[0019] Supervised learning relative to the pre-trained labels is performed on the pre-trained model using pre-trained samples to optimize the pre-trained model parameters.
[0020] The box-shaped economic operating domain is solved using a pre-trained model with optimized parameters.
[0021] As a preferred embodiment, the present invention further describes the box-shaped economic operating domain generation method based on the pre-trained model, including the following steps: S110: Obtain samples of new energy output ranges and load power samples of the power grid, and establish a mathematical solution model for the box-type economic operation domain.
[0022] In the embodiment, interval numbers are used to describe the uncertainty of new energy output, i.e. In the superscript, the negative sign "-" represents the lower limit of the interval, and the positive sign represents the upper limit of the interval.
[0023] This invention uses interval numbers to describe the uncertainty of wind power and solar power output and load demand, that is, using... To describe the uncertainty of wind and solar power output, the superscript "-" represents the lower limit of the interval, and the positive "+" represents the upper limit of the interval.
[0024] The mathematical solution model for the box-type economic operation domain of a power grid considering uncertainties on both the source and load sides can be expressed as the following two-level optimization model: The lower-level model includes: Objective function: (1); Constraints: (2); (3); (4); (5); In the above formula: Indicates dispatchable units Cost coefficient; Indicates dispatchable units exist Power during a given time period; , Indicates dispatchable units Minimum and maximum output power; New energy exist Maximum power generation during the period; Represents new energy exist Actual power consumption during the time period; Indicates dispatchable units Maximum downhill / climb rate; Indicates dispatchable units Maximum uphill speed; This indicates the quantity of new energy sources, dispatchable generating units, loads, and dispatch periods; Indicates load exist Power during a given time period; , Indicates connection with new energy Dispatchable units ,load The appropriate power transfer distribution coefficient; This represents the maximum transmission power of line l.
[0025] The upper-level model is the economic operation boundary solution model.
[0026] Objective function: (6).
[0027] When the objective function is When the solution model is obtained, the lower boundary of the economic operating domain can be obtained; when the objective function is When solving the model, the upper boundary of the economic operating domain can be obtained.
[0028] Constraints: ; in, These represent the lower and upper bounds of the uncertain range for new energy output, respectively. These are the lower and upper bounds of the load uncertainty interval, respectively.
[0029] S120, based on the new energy output range samples and load power samples, uses a mathematical solver to solve the mathematical model of the box-type economic operation domain, and obtains the box-type economic operation domain corresponding to each sample; extracts the new energy output range and load power as pre-training samples, and extracts the corresponding box-type economic operation domain as pre-training labels.
[0030] In this embodiment, the original two-level optimization model is transformed into a single-level mixed-integer linear programming problem that can be directly solved using a mathematical solver, through KKT conditions and the Big M method: ; ; ; ; ; ; ; Among them, matrix sum vector These represent the coefficient matrix and coefficient vector, respectively, which are adapted to the lower-level model in S110; For Lagrange multipliers; It is a sufficiently large constant; It is a 0-1 variable.
[0031] By solving the model using a mathematical solver, the upper and lower bounds of the economic operating domain for each unit can be obtained, ultimately yielding the box-shaped economic operating domain used as training labels.
[0032] S130 utilizes pre-trained samples to perform supervised learning of the pre-trained model relative to the pre-trained labels in order to optimize the parameters of the pre-trained model.
[0033] In this embodiment, the pre-training samples and pre-training labels are normalized, and the normalized pre-training samples are used to perform supervised learning of the pre-training model relative to the normalized pre-training labels.
[0034] Normalize the renewable energy output range samples, load power samples, and box-type economic operation domain calculation labels. Then, use the normalized renewable energy output range samples and load power samples to perform supervised learning on the box-type economic operation domain solution model relative to the normalized box-type economic operation domain calculation labels.
[0035] Finally, the data is normalized based on the following formula.
[0036] .
[0037] In this embodiment, the pre-trained model is built based on a deep neural network and is used to calculate and output the upper and lower bounds of the box-type economic operating domain based on the new energy output range and load power contained in the pre-trained samples.
[0038] In one embodiment, a pre-trained model is built based on a multilayer perceptron architecture, where the multilayer perceptron is a feedforward neural network composed of multiple layers of neurons. Each layer first performs a linear transformation, followed by a nonlinear activation function. For a single-layer network, its output can be expressed as: ; Where x is the input vector, W is the weight matrix, b is the bias vector, and σ is the activation function (e.g., ReLU or Sigmoid). The final output of the multilayer perceptron is computed by stacking multiple such structures.
[0039] In this embodiment, the loss function used during training consists of two parts: one is the accuracy loss, which is measured by mean squared error (MSE); the other is the constraint loss, which is used to ensure that the lower bound of the economic operating domain of the neural network output does not exceed its upper bound.
[0040] The expression for the accuracy loss is: ; in, Batch size This represents the actual value of the upper bound of the economic operating domain for the b-th sample in the batch at time t. This is the upper bound of the economic operating domain output by the neural network; Let be the actual value of the lower bound of the economic operating domain corresponding to the b-th sample in the batch at time t; This is the lower bound of the economic operating domain output by the neural network.
[0041] The expression for constraint loss is: .
[0042] The ultimate goal of training is to minimize the weighted sum of the two losses.
[0043] S140, by using the predicted new energy output range and the pre-trained model with optimized load power demand input parameters, the upper and lower bounds of the box-type economic operating domain can be quickly calculated, thus obtaining the final box-type economic operating domain.
[0044] As a preferred embodiment, the present invention also conducted the following comparative experiments on different systems, as shown in Tables 1 and 2 below. It can be seen that the present invention can significantly improve computational efficiency while maintaining accuracy.
[0045] Table 1: Comparison of Calculation Time .
[0046] Table 2: Calculation Accuracy .
[0047] This invention also provides a box-shaped economic operating domain generation device based on a pre-trained model, including one or more processors for implementing the box-shaped economic operating domain generation method based on the pre-trained model.
[0048] The embodiment of the pre-trained model-based box-type economic operating domain generation device of the present invention can be applied to any device with data processing capabilities, such as a computer. The device embodiment can be implemented in software, hardware, or a combination of both. Taking software implementation as an example, as a logical device, it is formed by the processor of any data processing device loading the corresponding computer program instructions from non-volatile memory into memory for execution. From a hardware perspective, such as... Figure 2 The diagram shown is a hardware structure diagram of any device with data processing capabilities, where the pre-trained model-based box-type economic operation domain generation device of this invention is located. (Except for...) Figure 2 In addition to the processor, memory, network interface, and non-volatile memory shown, any data processing device in the embodiment may also include other hardware depending on the actual function of the data processing device, which will not be described in detail here.
[0049] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.
[0050] This invention also provides a computer-readable storage medium storing a program thereon, which, when executed by a processor, is used to implement the aforementioned method for generating a box-shaped economic operating domain based on a pre-trained model.
[0051] The computer-readable storage medium can be an internal storage unit of any data processing device described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium can also be any data processing device, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc., equipped on the device. Furthermore, the computer-readable storage medium can include both internal storage units of any data processing device and external storage devices. The computer-readable storage medium is used to store the computer program and other programs and data required by the data processing device, and can also be used to temporarily store data that has been output or will be output.
[0052] The above embodiments are only used to illustrate the design concept and features of the present invention, and their purpose is to enable those skilled in the art to understand the content of the present invention and implement it accordingly. The protection scope of the present invention is not limited to the above embodiments. Therefore, all equivalent changes or modifications made based on the principles and design ideas disclosed in the present invention are within the protection scope of the present invention.
Claims
1. A method for generating a box-shaped economic operating domain based on a pre-trained model, characterized in that, Includes the following steps: Obtain samples of renewable energy output ranges and load power samples from the power grid, and establish a mathematical solution model for the box-type economic operation domain; Based on the new energy output range samples and load power samples, a mathematical solver is used to solve the mathematical solution model of the box-type economic operation domain to obtain the box-type economic operation domain corresponding to each sample; the new energy output range and load power are extracted as pre-training samples, and the corresponding box-type economic operation domain is extracted as pre-training labels. Supervised learning of the pre-trained model relative to the pre-trained labels is performed using pre-trained samples to optimize the parameters of the pre-trained model; The box-shaped economic operating domain is solved using a pre-trained model with optimized parameters.
2. The method for generating a box-shaped economic operating domain based on a pre-trained model according to claim 1, characterized in that, The new energy output range includes the wind power output range and the photovoltaic power output range, and the load power is the active power load of the power system; The mathematical solution model for the box-shaped economic operation domain is a two-layer optimization model. The lower-layer model is an economic dispatch model, and the upper-layer model is constructed with the goal of solving the operating boundaries of each unit under uncertain conditions. The upper-layer model is an economic operation boundary solution model, which describes the uncertainty of new energy output in the form of interval numbers.
3. The method for generating a box-shaped economic operating domain based on a pre-trained model according to claim 1, characterized in that, The method further includes: normalizing the pre-training samples and pre-training labels, and using the normalized pre-training samples to perform supervised learning of the pre-training model relative to the normalized pre-training labels. Normalize the renewable energy output range samples, load power samples, and box-type economic operation domain calculation labels. Then, use the normalized renewable energy output range samples and load power samples to perform supervised learning on the box-type economic operation domain mathematical solution model relative to the normalized box-type economic operation domain calculation labels.
4. The method for generating a box-shaped economic operating domain based on a pre-trained model according to claim 1, characterized in that, The method of using pre-trained samples to perform supervised learning on the pre-trained model relative to the pre-trained labels to optimize the pre-trained model parameters includes: the pre-trained model is built based on a deep neural network, the input is the uncertainty range of new energy output and the load power value, and the output is a box-shaped economic operation domain, that is, the upper and lower bounds of the economic operation of each unit under uncertainty; when supervising the model, a pre-training loss function is constructed based on the difference between the lower and upper bounds of the output economic operation domain and the lower and upper bounds of the actual economic operation domain contained in the pre-trained labels; at the same time, a penalty loss function is introduced to penalize when the lower bound of the economic operation domain output by the model is greater than the upper bound, so as to ensure that the upper bound of the economic operation domain is always greater than the lower bound; finally, the pre-trained model is trained with the goal of minimizing the sum of the two loss functions.
5. The method for generating a box-shaped economic operating domain based on a pre-trained model according to claim 1, characterized in that, The process of solving the box-shaped economic operating domain using a pre-trained model with optimized parameters includes: The predicted new energy output range and load power are input into the pre-trained model with optimized parameters to calculate the box-type economic operation domain of the power system, including the upper and lower bounds of the economic operation of each unit in the power system.
6. A box-shaped economic operating domain generation device based on a pre-trained model, characterized in that, It includes one or more processors for implementing the box-type economic operating domain generation method based on a pre-trained model as described in any one of claims 1-5.
7. A computer-readable storage medium having a program stored thereon, characterized in that, When executed by the processor, the program is used to implement the box-type economic operating domain generation method based on the pre-trained model as described in any one of claims 1-5.