Power market clearing scheme optimization method, system, equipment and medium

Abstract

The invention discloses an electricity market clearing scheme optimization method, system and device and a medium, the method comprises the steps that a multi-dimensional sequence data set is constructed according to collected historical data, and the historical data at least comprises intraday clearing electricity price data, electricity supply and demand data, meteorological data and market policy coding data; inputting the multi-dimensional sequence data set into a preset clearing prediction model for training, inputting clearing data, collected in real time, of the electricity market into the trained clearing prediction model, and outputting to obtain clearing electricity price prediction data; the clearing prediction model is designed to introduce a sparse attention mechanism and a feedforward neural network to form an Informer model of an encoder stacking structure so as to realize optimization of feature representation; and formulating a clearing scheme of the electricity market according to the clearing electricity price prediction data, thereby optimizing resource configuration.

Description

Technical Field

[0001] The present invention relates to the technical field of electric power, and in particular to an optimization method, system, device and medium for an electricity market clearing scheme. Background Art

[0002] With the global energy transformation and the gradual opening of the electricity market, the complexity and uncertainty of the electricity market are increasing day by day. As an important participant in the electricity market, electricity selling companies need to face a more volatile market environment and competitive pressure. Therefore, improving the accuracy and stability of electricity price forecasting is of great significance for the business decisions of electricity selling companies. In the operation of the electricity market, the clearing electricity price is the core signal of the supply and demand relationship in the electricity market, which directly affects the bidding strategies of power generators, the electricity consumption behaviors of users and the grid dispatching decisions. Through the competitive trading and price discovery mechanisms, the clearing scheme can reflect the real-time changes in the supply and demand of electricity and price fluctuations, providing a basis for decision-making for market participants. This helps to optimize the allocation of electric power resources, make resources flow towards the demand direction, and achieve the efficient utilization of resources.

[0003] The data dimensions relied on by the existing clearing schemes are relatively single. This single data prediction method makes the output results of the clearing prediction model have large errors, and the accuracy of the formulated clearing schemes is low, making it difficult to meet the requirements of the electricity market.

[0004] Therefore, how to optimize the clearing scheme of the electricity market and maintain the supply and demand balance of the electricity market has become a technical problem that needs to be urgently solved by those skilled in the art. Summary of the Invention

[0005] The present invention provides an optimization method, system, device and medium for an electricity market clearing scheme, which solves the problem of how to comprehensively consider the multi-dimensional factors affecting the electricity market clearing mechanism, and combine machine learning algorithms to realize multi-dimensional and all-round prediction and analysis of the clearing electricity price, generate reasonable and reliable clearing results, and ensure the stable operation of the power grid.

[0006] To solve the above technical problems, an embodiment of the present invention provides an optimization method for an electricity market clearing scheme, including:

[0007] Constructing a multi-dimensional sequence data set according to the collected historical data, where the historical data at least includes intraday clearing electricity price data, electricity supply and demand data, meteorological data and market policy coding data;

[0008] Input the multi-dimensional sequence dataset into a preset clearing prediction model for training, and input the real-time collected clearing data of the power market into the trained clearing prediction model to output the clearing price prediction data; the clearing prediction model is designed as an Informer model that introduces a sparse attention mechanism and a feed-forward neural network to form an encoder stack structure to optimize the feature representation;

[0009] Formulate a clearing plan for the power market according to the clearing price prediction data.

[0010] Further, constructing the multi-dimensional sequence dataset according to the collected historical data includes:

[0011] Arrange the collected intraday clearing price data, power supply and demand data, meteorological data, and market policy coding data at a preset time interval to obtain an initial sequence dataset;

[0012] Perform preprocessing operations such as data cleaning, timestamp alignment, and normalization on the initial sequence dataset, and analyze the characteristics of various types of data in the preprocessed initial sequence dataset to construct the multi-dimensional sequence dataset.

[0013] Further, the characteristic analysis of various types of data in the preprocessed initial sequence dataset to construct the multi-dimensional sequence dataset includes:

[0014] Calculate the mutual information between the intraday clearing price data and the power supply and demand data, meteorological data, and market policy coding data in the initial sequence dataset, and screen out a candidate feature set from the calculation results according to preset conditions;

[0015] Map each modal data in the candidate feature set to the same low-dimensional space through a pre-trained implicit neural representation model to generate corresponding embedded vector representations;

[0016] Perform clustering analysis on the embedded vector representations to construct the multi-dimensional sequence dataset according to the clustering results.

[0017] Further, the inputting the multi-dimensional sequence dataset into a preset clearing prediction model for training includes:

[0018] Analyze the output results of the clearing prediction model through the combined mean square error and mean absolute error to optimize the parameters of the clearing prediction model.

[0019] Further, the formulating a clearing plan for the power market according to the clearing price prediction data includes:

[0020] Construct a multi-objective optimization function considering line thermal stability constraints based on the intraday time-segmented clearing price prediction curve output by the clearing prediction model;

[0021] Use the decomposition and coordination algorithm to iteratively solve the multi-objective optimization function and output the optimal solution including the output plans of each unit in the power market, the power distribution of each line, and the capacity allocation of each resource. Formulate a clearing plan corresponding to the power market according to the optimal solution.

[0022] Further, the multi-objective optimization function includes:

[0023] The first sub-function with the maximum capacity adequacy of reserve resources in the power market as the goal and the second sub-function with the minimum transmission congestion risk as the goal.

[0024] Further, after formulating the clearing plan for the power market according to the clearing price prediction data, it further includes:

[0025] Extract the rising period data and falling period data in the intraday time-segmented clearing price prediction curve, and formulate the call priorities of each reserve resource in the power reserve capacity market with the rising period data and falling period data respectively;

[0026] Dynamically adjust the call priorities of each reserve resource based on the clearing plan.

[0027] Another embodiment of the present invention provides a power market clearing plan optimization system, including:

[0028] A data acquisition module for constructing a multi-dimensional sequence data set according to the collected historical data, where the historical data at least includes intraday clearing price data, power supply and demand data, meteorological data, and market policy coding data;

[0029] A clearing prediction module for inputting the multi-dimensional sequence data set into a preset clearing prediction model for training, and inputting the real-time collected clearing data of the power market into the trained clearing prediction model, and outputting the clearing price prediction data; the clearing prediction model is designed as an Informer model that introduces a sparse attention mechanism and a feed-forward neural network to form an encoder stacking structure to optimize the feature representation;

[0030] A clearing plan optimization module for formulating a clearing plan for the power market according to the clearing price prediction data.

[0031] Another embodiment of the present invention provides a computer device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the power market clearing plan optimization method as described above.

[0032] Another embodiment of the present invention provides a computer-readable storage medium storing a computer program. When the device where the computer-readable storage medium is located executes the computer program, the optimization method for the electricity market clearing scheme described above is implemented.

[0033] Compared with the prior art, the beneficial effects of the embodiments of the present invention are at least one of the following:

[0034] By integrating multi-dimensional data such as electricity price, supply and demand, weather, and policies, and performing data processing through feature engineering combining INR and mutual information, the embodiments of the present invention can enhance the adaptability of the model to complex market environments. By introducing an improved ProbSparse self-attention mechanism and self-attention distillation technology into the prediction model, the model can efficiently capture the long-term dependencies and short-term fluctuations in the clearing electricity price data, significantly improving the prediction accuracy. An accurate clearing scheme is formulated according to the clearing price curve output by the model to optimize resource allocation, reduce resource waste, and ensure the supply-demand balance of the electricity market. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 is a schematic flowchart of the optimization method for the electricity market clearing scheme in one embodiment of the present invention;

[0036] Figure 2 is a schematic flowchart of the operation process of the prediction model in one embodiment of the present invention;

[0037] Figure 3 is a schematic structural diagram of the optimization system for the electricity market clearing scheme in one embodiment of the present invention;

[0038] Figure 4 is a structural block diagram of a preferred embodiment of a computer device provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0039] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0040] In the description of this application, the terms "first", "second", "third", etc. are only used for descriptive purposes and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first", "second", "third", etc. may explicitly or implicitly include one or more of such features. In the description of this application, unless otherwise stated, the meaning of "a plurality" is two or more.

[0041] In the description of this application, it should be noted that unless otherwise clearly specified and limited, the terms "installed", "connected", and "coupled" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two components. The terms "vertical", "horizontal", "left", "right", "up", "down" and similar expressions used herein are only for illustrative purposes and do not indicate or imply that the device or component referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation of the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0042] In the description of this application, it should be noted that unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as commonly understood by those of ordinary skill in the technical field to which this invention belongs. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0043] An embodiment of the present invention provides a method for optimizing a power market clearing scheme. Specifically, please refer to Figure 1 , Figure 1 which shows a schematic flow chart of the method for optimizing a power market clearing scheme in one of the embodiments of the present invention, including the following steps:

[0044] S1. Construct a multi-dimensional sequence data set based on the collected historical data, where the historical data at least includes intraday clearing price data, power supply and demand data, meteorological data, and market policy coding data.

[0045] In this embodiment, the intraday clearing price (at a granularity of 15 minutes / hour) is obtained from the power trading center, and the fields include information such as timestamp, region ID, price value, and trading volume.

[0046] Through the power Internet of Things data collection system, such as the power grid operator system, relevant power supply and demand data including power generation, power consumption, load demand, etc. are obtained. The collection terminal can be a smart meter, sensor, etc., and the time granularity needs to be aligned with the electricity price data. The temperature, wind speed, precipitation, irradiance and other meteorological data in the relevant meteorological data service system are called, stored according to the longitude and latitude grid (such as 0.1°×0.1°), and the time granularity matches the power data. Obtain market policy information including market rules, policy documents, subsidy policies, etc. from relevant government departments or market operating institutions, encode them into structured tags, and associate them with timestamps.

[0047] The collected intraday clearing electricity price data, electricity supply and demand data, meteorological data and market policy coding data are arranged and sampled at preset time intervals to create an initial sequence data set. The data set D can be expressed as:

[0048] D={d1,d2,…d T}, d T =[p T ,s T ,w T ,m T ]

[0049] Among them, p T is the electricity price at time t; s T is the supply and demand indicator; w T is the weather characteristics; m T Code the market policy.

[0050] To ensure the quality of model input, this embodiment further performs preprocessing operations such as data cleaning, timestamp alignment, and normalization on the initial sequence data set. This includes filling missing data with historical mean values, or interpolation processing based on the volatility and periodicity of the data, as well as outlier detection processing. It is also necessary to convert data such as electricity price data, temperature, wind speed, etc. into a standard scale to eliminate dimensional differences between different data levels, so as to facilitate subsequent feature selection and dimensionality reduction processing.

[0051] The characteristics of various types of data in the preprocessed initial sequence data set are analyzed to construct the multidimensional sequence data set. Specifically, the characteristic analysis process aims to reduce the dimension of the initial sequence data set to improve the output accuracy of the model.

[0052] Preferably, the embodiment of the present invention measures the correlation between each feature (such as real-time load, renewable energy output, temperature, wind speed, etc.) and the target variable (clearing electricity price) by using mutual information. The larger the mutual information value, the stronger the correlation between the feature and the target variable. That is, in this embodiment, the intraday clearing electricity price data in the initial sequence data set is used as the target variable to calculate the mutual information between it and each data in the power supply and demand data, meteorological data and market policy coding data.

[0053] According to the preset conditions, the candidate feature set is screened from the calculation results. In this embodiment, the mutual information values are sorted from large to small, and a portion of the top-ranked features are selected as the candidate feature set. These features have a strong correlation with the target variable (clearing electricity price).

[0054] In some embodiments of the present invention, a Pearson correlation coefficient machine can be further used for feature analysis, and by comparing the results of mutual information analysis and Pearson correlation coefficient analysis, features that have both a strong nonlinear correlation (mutual information) with the target variable and a certain linear correlation (Pearson coefficient) can be screened out to improve the robustness of feature selection.

[0055] Implicit neural network representation (INR) is a powerful method to implicitly represent multidimensional data through neural networks, which learns the mapping from multidimensional coordinate space to signal space through neural networks. Based on this, in this embodiment, each modality data in the candidate feature set is mapped to the same low-dimensional space through a pre-trained implicit neural representation model to generate a corresponding unified embedding vector representation.

[0056] Regarding the construction process of the INR model, in this embodiment, multidimensional data including time coordinates, space coordinates, electricity prices, supply and demand, meteorological parameters, and policy codes are input into the input layer of the INR model, and a multi-layer perceptron with a sinusoidal activation function is designed as a hidden layer to perform high-frequency fitting of the data, and then the output layer is designed to output a unified embedded vector representation. The loss function is introduced during the training process for data reconstruction and fitting optimization.

[0057] It is worth noting that in some embodiments of the present invention, the attention mechanism is further introduced in the process of INR mapping to further integrate multimodal information. Specifically, according to the current market status and data characteristics, the weights of different modal data are dynamically adjusted so that the model can pay more attention to important information. For example, when market policies change, the weight of policy encoding data is increased so that the model can respond to policy changes more quickly.

[0058] Alternatively, information fusion is performed through a graph neural network (GNN), where data of different modalities are represented as nodes and edges in a graph structure, and the GNN is used to learn the feature representations of the nodes and edges. Exemplarily, meteorological data from different regions are represented as nodes in the graph, and the meteorological influence between regions is represented as edges. The GNN is used to learn the mutual influence of meteorological patterns between regions to improve the accuracy of prediction.

[0059] After outputting the embedded vector representation through the INR, this embodiment will further perform clustering analysis on the embedded vector representation to study the variation laws of variables such as clearing price and supply and demand in time and space. Preferably, this embodiment selects the K-means algorithm to calculate the cluster centers corresponding to various data points in the initial sequence dataset, and then combines the original data with the cluster labels to construct a multi-dimensional sequence dataset.

[0060] Through the combination of mutual information feature selection, feature mapping of the INR model, and clustering analysis, the embodiments of the present invention can achieve comprehensive modeling and analysis of complex multi-modal data.

[0061] S2. Input the multi-dimensional sequence dataset into a preset clearing prediction model for training, and input the real-time collected clearing data of the power market into the trained clearing prediction model to output the clearing price prediction data.

[0062] The clearing prediction model constructed in this embodiment is designed to introduce a sparse attention mechanism (ProbSparse self-attention) and a feed-forward neural network to form an Informer model with an encoder stacking structure to optimize the feature representation. It can be understood that the Informer model is an efficient long-sequence time series prediction model that combines the efficient sequence modeling ability of the Transformer and a special processing mechanism for long-sequence data, improving the model's feature extraction ability and the ability to accurately predict the fluctuation pattern of the clearing price. Specifically, for the training process of the clearing prediction model, please refer to Figure 2 as shown Figure 2 which shows a schematic diagram of the operation process of the prediction model in one of the embodiments of the present invention, including the following steps:

[0063] First, design the model structure to map the multi-dimensional sequence dataset into a high-dimensional vector to capture the time series pattern.

[0064] Embedding layer: Map the feature vector of each time point into an embedded vector with a feature dimension of d through a fully connected layer, and add absolute position encoding to provide the time series position information for the model. Preferably, d = 128 and the position encoding dimension = 64.

[0065] Encoder: ProbSparse self-attention and self-attention distillation are introduced in each layer. Specifically, ProbSparse self-attention is designed to retain the top-u key queries, such as Top-50%. After each layer of the encoder, a one-dimensional convolution (with a kernel size of 3 and a stride of 2) is used to reduce the dimension of the attention output, and the sequence length is halved layer by layer. A multi-layer stacked structure is formed to achieve feature abstraction and information fusion. Exemplarily, after the input embedding vector sequence, ProbSparse self-attention is introduced in sub-layer 1 to calculate the sparse attention weights and generate a context-aware sequence representation. In sub-layer 2, a feed-forward neural network (FFN) containing two fully-connected layers and ReLU activation functions is introduced to further extract non-linear features. Finally, after multi-layer stacking, the compressed high-dimensional features are output.

[0066] Generative decoder: Based on the positional encoding of the historical sequence and future timestamps, directly output the multi-step electricity price prediction sequence through a single forward propagation. In this embodiment, the input of the decoder is composed of two parts spliced together to ensure that the model can simultaneously perceive the historical pattern and future temporal position.

[0067] Specifically, take the features of the last L time points output by the encoder as the historical feature sequence. In this embodiment, the output of the encoder is represented as:

[0068] H enc ∈R L×d

[0069] where L is the final sequence length of the encoder, representing the size of the historical time window considered by the model, and R is the set of real numbers.

[0070] In this embodiment, by setting sin / cos functions with different frequencies, the model can perceive the absolute position of each time point in the future sequence to generate S positional encoding vectors PE future ∈R S×d , to predict the electricity price at the next S time points. The specific positional encoding is represented by the following formula:

[0071] PE (t，2i) =sin(t / 10000 2i / d )

[0072] PE (t，2i+1) =cos(t / 10000 2i / d )

[0073] where t represents the position of the current time step in the prediction sequence; i represents the dimension index of the feature vector.

[0074] Concatenate H enc and PE future along the time dimension to obtain the decoder input D in ∈R(L+S)×d Example: When predicting the electricity price for the next 24 hours with a data point every 15 minutes, the generated S = 96.

[0075] Further, the decoder fuses the encoder output and its own input by introducing two-stage attention, including: a self-attention layer for calculating the internal temporal dependencies of the decoder input D in to capture the internal dependencies within the future sequence, and an encoder-decoder attention layer that uses the encoder output H enc as key-value pairs and the decoder self-attention output as the query.

[0076] Further, the feed-forward output layer of the encoder maps the attention output to the electricity price prediction value through a fully connected layer which is expressed as:

[0077]

[0078] where W1 is the weight of the first fully connected layer; b1 is the bias term of the first layer; ReLU is the activation function; W2 is the weight of the second fully connected layer; b2 is the bias term of the second layer. W1 ∈ R d×4d and W2 ∈ R 4d×1 , and the output dimension is (L + S) × 1.

[0079] According to the above formula, only the prediction results for the next S time points will be retained and the first L ones will be discarded as prediction values.

[0080] Through the above construction process of the Informer model, in this embodiment, the output results of the Informer clearing prediction model are further analyzed by jointly using the mean square error (MSE) and the mean absolute error (MAE) to optimize the parameters of the Informer clearing prediction model. Specifically, the loss function φ is expressed as:

[0081]

[0082] where P ture is the true electricity price sequence; α and β are weight coefficients to balance the weights of MSE and MAE and control the contribution ratio of the two to the total loss. In some embodiments of the present invention, α and β can be set to 0.7 and 0.3 respectively.

[0083] During the optimization process, in this embodiment, the Adam optimizer is preferably used, and its learning rate can be set to 0.001, the batch size can be set to 64, and the number of training epochs can be set to 200. Specifically, examples of the model parameters are shown in Table 1:

[0084] Table 1 Examples of model parameters

[0085]

[0086] The model is verified by calculating the prediction error on the test set. Specifically, with a sampling interval of 15 minutes, the clearing data of the electricity market is collected in real time and annotated, and the annotated data is input into the trained clearing prediction model. The model is comprehensively evaluated through four indicators: Root Mean Square Error (RMSE), MAE, Direction Accuracy (DA), and Coefficient of Determination (R 2 ). The specific calculation process is as follows:

[0087] Root Mean Square Error (RMSE), the calculation process is expressed as follows:

[0088]

[0089] DA (Direction Accuracy) measures the correct rate of predicting the direction of electricity price change, and the calculation process is expressed as follows:

[0090]

[0091] R 2 (Coefficient of Determination) reflects the explanatory ability of the model for electricity price fluctuations, and the closer it is to 1, the better. The calculation process is expressed as follows:

[0092]

[0093] MAE (Mean Absolute Error) measures the absolute level of prediction error, and the calculation process is expressed as follows:

[0094]

[0095] In the above formulas, are the predicted values of the model at the t-th and (t - 1)-th moments respectively; P t , P t-1 are the true values at the t-th and (t - 1)-th moments respectively; is the average value of the true values.

[0096] In order to evaluate the performance of different methods in predicting the intraday clearing electricity price, in some embodiments of the present invention, multiple methods such as ARIMA, LSTM (Long Short-Term Memory Network), Transformer, and Informer are adopted, and a comparative test is carried out on the collected intraday clearing electricity price data set based on the selected target electricity price market. Specifically, please refer to Table 2 as follows:

[0097] Table 2 Evaluation and verification of clearing electricity price prediction with different models

[0098] Model RMSE MAE DA(%) <![CDATA[R 2 > ARIMA 12.45 9.87 61.2 0.43 LSTM 8.76 6.54 68.9 0.67 Transformer 7.23 5.32 73.5 0.78 Informer 5.89 4.21 82.4 0.89

[0099] As can be seen from Table 2 above, informer shows significant advantages in predicting the intraday clearing electricity price. Among them, through the powerful informer self-attention mechanism and self-attention distillation technology, informer achieves RMSE (5.89), MAE (4.21), and excellent R 2 (0.89), indicating that the model can explain 89% of the electricity price fluctuations and has high reliability.

[0100] S3. Formulate a clearing plan for the power market based on the clearing electricity price prediction data.

[0101] After obtaining the intraday sub-period clearing electricity price prediction curve output by the clearing prediction model, the embodiment of the present invention further constructs a multi-objective optimization function considering the line thermal stability constraint, combining the power market reserve capacity target and the transmission line reliability target.

[0102] Specifically, the objective function includes a first sub-function aiming at maximizing the capacity adequacy of the reserve resources in the power market and a second sub-function aiming at minimizing the transmission congestion risk.

[0103] Among them, the first sub-function ensures the system's ability to cope with sudden power deficits by maximizing the matching degree between the reserve capacity supply and demand. Then the first sub-function f1 is expressed as follows:

[0104]

[0105] Wherein, is the reserve capacity demand in period t, calculated according to the predicted electricity price; is the reserve capacity provided by unit i in period t; T is the total number of periods.

[0106] The second sub-function reduces the risk of cascading failures in the power grid by minimizing the proportion of line power flow exceeding the thermal stability limit. The smaller the value, the safer the operation of the power grid. Then the second sub-function f2 is expressed as follows:

[0107]

[0108] Wherein, is the thermal stability limit capacity of line l; P l,t is the transmission power of line l in period t; L is the total number of transmission lines in the power grid.

[0109] Then the multi-objective optimization function is expressed as follows:

[0110] min(w1f1 + w2f2)

[0111] Wherein, w1 and w2 are weight coefficients, which can be dynamically adjusted according to the predicted electricity price curve. For example, w1 is increased during high electricity price periods to strengthen safety and reserve.

[0112] The constraint conditions include power balance constraint, upper and lower limits of unit output constraint, unit ramp rate constraint, and line thermal stability constraint. The line thermal stability constraint is expressed as follows:

[0113]

[0114] This constraint ensures that the line power does not exceed its physical limit during bidirectional transmission, preventing failures caused by overload.

[0115] The decomposition and coordination algorithm is used to iteratively solve the multi-objective optimization function. Specifically, the above multi-objective function can be solved by setting the main problem and sub-problem decomposition, and combined with the alternating direction multiplier method for iterative operation to obtain the optimal solutions of the output plans of each unit, the power distribution of each line, and the capacity allocation of each resource. Exemplarily, such as the output values of each power generation unit (thermal power, hydropower, gas, new energy, etc.) in each time period (such as 15-minute intervals) in the power market, the reserved capacities of various reserve resources (energy storage, gas turbines, demand response, etc.) in each time period, and the power transmission values of each transmission line.

[0116] An clearing plan corresponding to the power market is formulated according to the optimal solutions, converted into corresponding instructions and sent to the corresponding control terminals in the power market, and the execution status is monitored in real time. Exemplarily, an output control instruction is sent to a thermal power unit, and the unit executes the output instruction and cooperates with the power grid dispatching center for real-time monitoring.

[0117] To cope with the volatility of the power market, the clearing priority of the power reserve capacity market is adjusted according to the clearing plan. Specifically, first, a priority trigger mechanism is set:

[0118] The rising period data and falling period data of the electricity price in the intraday time-period clearing electricity price prediction curve are extracted, and through weighted analysis, the call priorities of each reserve resource in the power reserve capacity market are formulated with the rising period data and falling period data respectively. In this embodiment, in the rising period, in the high clearing electricity price interval, battery energy storage and gas turbines are preferentially called in sequence for rapid response, and the call weight setting ratio of such resources is the largest in the rising period; in the falling period, in the low clearing electricity price interval, coal-fired units and demand response with low resource costs are preferentially called in sequence, and the call weight setting ratio of such resources is the largest in the falling period.

[0119] In the embodiment of the present invention, based on the clearing plan, the call priorities of each reserve resource are dynamically adjusted. Exemplarily, when abnormal situations occur in the market or the prediction results deviate greatly, in combination with real-time market data and prediction results, the clearing priority is optimized by dynamically adjusting the weights.

[0120] In summary, in the embodiment of the present invention, by using the INR model and mutual information to fuse multi-dimensional data such as electricity price, supply and demand, weather, and policies, various factors affecting the electricity market environment are fully considered; by using the Informer model to construct a clearing prediction model, and introducing the ProbSparse self-attention mechanism and self-attention distillation technology to set the multi-dimensional encoder stacking structure, the model can efficiently capture the long-term dependencies and short-term fluctuations in the clearing electricity price data, so as to improve the accuracy of the output structure; a clearing plan is formulated based on the accurate clearing electricity price curve, and the clearing priority of the electricity reserve capacity market is dynamically adjusted to optimize resource allocation, thus ensuring the supply-demand balance of the electricity market.

[0121] An embodiment of the present invention provides a system for optimizing the electricity market clearing plan. Specifically, please refer to Figure 3 , Figure 3 which shows a schematic structural diagram of the system for optimizing the electricity market clearing plan in one embodiment of the present invention, including:

[0122] A data acquisition module M1, configured to construct a multi-dimensional sequence data set according to the collected historical data, where the historical data at least includes intraday clearing electricity price data, electricity supply and demand data, meteorological data, and market policy coding data;

[0123] A clearing prediction module M2, configured to input the multi-dimensional sequence data set into a preset clearing prediction model for training, and input the clearing data of the electricity market collected in real time into the trained clearing prediction model, and output the clearing electricity price prediction data; the clearing prediction model is designed as an Informer model that introduces a sparse attention mechanism and a feed-forward neural network to form an encoder stacking structure to realize the optimization of feature representation;

[0124] A clearing plan optimization module M3, configured to formulate an electricity market clearing plan according to the clearing electricity price prediction data.

[0125] As Figure 4 shown, an embodiment of the present invention also provides a computer device, Figure 4 which is a structural block diagram of a preferred embodiment of a computer device provided by the present invention. The computer device includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the above-mentioned method for optimizing the electricity market clearing plan.

[0126] Preferably, the computer program may be divided into one or more modules / units (such as computer program 1, computer program 2, ……), and the one or more modules / units are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, and these instruction segments are used to describe the execution process of the computer program in the computer device.

[0127] The processor may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor, or the processor may also be any conventional processor. The processor is the control center of the terminal device and connects various parts of the terminal device through various interfaces and lines.

[0128] The memory mainly includes a program storage area and a data storage area. Among them, the program storage area may store an operating system, application programs required for at least one function, etc., and the data storage area may store relevant data, etc. In addition, the memory may be a high-speed random access memory, or may also be a non-volatile memory, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc., or the memory may also be other volatile solid-state storage devices.

[0129] It should be noted that the above terminal device may include, but is not limited to, a processor and a memory. Those skilled in the art can understand that Figure 4The structural block diagram is only an example of the terminal device and does not constitute a limitation on the terminal device. It may include more or fewer components than shown, or combine some components, or different components. Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it may include the processes of the embodiments of the above methods. Among them, the storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0130] Correspondingly, an embodiment of the present invention provides a computer-readable storage medium. The computer-readable storage medium includes a stored computer program. When the computer program runs, it controls the device where the computer-readable storage medium is located to execute the steps in the above method, for example Figure 1 the steps S1 to S3 described in

[0131] The technical features and technical effects of the power market clearing scheme optimization system proposed in the embodiments of the present invention are the same as those of the power market clearing scheme optimization method proposed in the embodiments of the present invention, and will not be elaborated here.

[0132] The above embodiments only represent several implementation manners of the present invention. The description is relatively specific and detailed, but it cannot be understood as a limitation on the scope of the patent of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the appended claims.

Claims

1. An optimization method for electricity market clearing scheme, characterized in that Including: Construct a multi-dimensional sequence data set based on the collected historical data, where the historical data at least includes intraday clearing electricity price data, power supply and demand data, meteorological data, and market policy coding data; Input the multi-dimensional sequence data set into a preset clearing prediction model for training, and input the real-time collected clearing data of the power market into the trained clearing prediction model to output the clearing electricity price prediction data; the clearing prediction model is designed as an Informer model that introduces a sparse attention mechanism and a feed-forward neural network to form an encoder stacking structure to optimize the feature representation; Formulate a clearing plan for the power market according to the clearing electricity price prediction data.

2. The optimization method for the electricity market clearing scheme according to claim 1, wherein The constructing a multi-dimensional sequence data set according to the collected historical data includes: Arrange the collected intraday clearing electricity price data, the power supply and demand data, the meteorological data, and the market policy coding data at a preset time interval to obtain an initial sequence data set; Perform preprocessing operations on the initial sequence data set, including data cleaning, timestamp alignment, and normalization, and perform characteristic analysis on various types of data in the preprocessed initial sequence data set to construct the multi-dimensional sequence data set.

3. The power market clearing scheme optimization method according to claim 2, characterized in that The performing characteristic analysis on various types of data in the preprocessed initial sequence data set to construct the multi-dimensional sequence data set includes: Calculate the mutual information between the intraday clearing electricity price data and the power supply and demand data, the meteorological data, and the market policy coding data in the initial sequence data set, and screen out a candidate feature set from the calculation results according to preset conditions; Map each modal data in the candidate feature set to the same low-dimensional space through a pre-trained implicit neural representation model to generate corresponding embedded vector representations; Perform clustering analysis on the embedded vector representations to construct the multi-dimensional sequence data set according to the clustering results.

4. The power market clearing scheme optimization method according to claim 1, characterized in that The inputting the multi-dimensional sequence data set into a preset clearing prediction model for training includes: Analyze the output results of the clearing prediction model through the combined mean square error and mean absolute error to optimize the parameters of the clearing prediction model.

5. The power market clearing scheme optimization method according to claim 1, characterized in that, The formulating a clearing plan for the power market according to the clearing electricity price prediction data includes: Construct a multi-objective optimization function considering line thermal stability constraints based on the intraday sub-period clearing electricity price prediction curve output by the clearing prediction model; Use the decomposition and coordination algorithm to iteratively solve the multi-objective optimization function to output the optimal solutions including the output plans of each unit in the power market, the power distribution of each line, and the capacity allocation of each resource, and formulate a clearing plan corresponding to the power market according to the optimal solutions.

6. The power market clearing scheme optimization method according to claim 5, wherein The multi-objective optimization function includes: A first sub-function with the goal of maximizing the capacity adequacy of reserve resources in the power market and a second sub-function with the goal of minimizing the transmission congestion risk.

7. The optimization method for the electricity market clearing plan according to claim 5, characterized in that, After the formulating a clearing plan for the power market according to the clearing electricity price prediction data, it further includes: Extract the rising period data and the falling period data from the intraday sub-period clearing electricity price prediction curve, and formulate the call priorities of each reserve resource in the power reserve capacity market with the rising period data and the falling period data respectively. Based on the clearing plan, dynamically adjust the call priorities of each backup resource.

8. An optimization system for electricity market clearing scheme, characterized in that, Including: A data acquisition module, configured to construct a multi-dimensional sequence data set according to the collected historical data, where the historical data at least includes intraday clearing price data, power supply and demand data, meteorological data, and market policy coding data; A clearing prediction module, configured to input the multi-dimensional sequence data set into a preset clearing prediction model for training, and input the clearing data of the power market collected in real time into the trained clearing prediction model, and output clearing price prediction data; the clearing prediction model is designed as an Informer model that introduces a sparse attention mechanism and a feed-forward neural network to form an encoder stacking structure to optimize the feature representation; A clearing plan optimization module, configured to formulate a clearing plan for the power market according to the clearing price prediction data.

9. A computer device, characterized in that, Including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, it implements the power market clearing plan optimization method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, and when the device where the computer-readable storage medium is located executes the computer program, it implements the power market clearing plan optimization method according to any one of claims 1 to 7.

Images