A power price fluctuation prediction method and device

Abstract

The application provides a power price fluctuation prediction method and device, and relates to the technical field of power systems. The method comprises the following steps: obtaining historical operation data and operation condition data of a power system; performing delay embedding on the operation data to construct a Hankel matrix, dividing the Hankel matrix into two-dimensional subspace patches and converting the two-dimensional subspace patches into tokens, and processing the tokens by using an encoder to obtain a latent state; encoding the operation condition data into a condition vector; based on a conditionally affine conditional Koopman operator, performing transition on the latent state by using the condition vector to obtain a latent state at a future time; and decoding the latent state by using a quantile prediction head to obtain a conditional distribution prediction result of power price fluctuation in a future period. In the application, the operation condition is taken as a known input for prediction, a conditionally affine conditional Koopman operator is used to realize fast linear replay to support counterfactual reasoning, and a quantile prediction head is used to output the conditional distribution of power price fluctuation, so that the accuracy, interpretability and calculation efficiency of power price prediction in an event-driven scenario are improved.

Description

Technical Field

[0001] This application relates to the field of power system technology, and in particular to a method and apparatus for predicting electricity price fluctuations. Background Technology

[0002] In the context of the electricity spot market, the dispatching, operation, and safety control of the power system are highly dependent on accurate perception of the supply-demand balance. The supply-demand balance of the power system is jointly determined by load demand, power output, unit combination, and grid constraints. Imbalances directly reflect deviations in system frequency from rated values, potentially triggering serious safety incidents such as frequency overruns, under-frequency load shedding, and even cascading failures, jeopardizing the safe and stable operation of the power system. Therefore, accurate perception of the power system's supply-demand situation is crucial for ensuring the safe operation of the power grid. Electricity prices, as a comprehensive signal of the supply-demand situation, directly reflect the system's supply-demand balance. However, traditional supply-demand situation perception methods typically rely solely on load and power output trends, failing to effectively perceive sudden changes in the supply-demand situation under event-driven scenarios. Therefore, there is an urgent need for a method that can use operating conditions as known inputs, possess an interpretable dynamic structure, and output the distribution of supply-demand situation conditions to improve the perception capability of the power system's supply-demand situation in event-driven scenarios, providing technical support for the safe operation of the power grid. Summary of the Invention

[0003] This application aims to at least partially address one of the technical problems in the related art.

[0004] Therefore, the first aspect of this application proposes a method for predicting electricity price fluctuations, including: The system acquires operating data and operating condition data of the power system within a historical period. The operating data includes electricity price data, load data, and power output data. The operating condition data includes unit start-up and shutdown status, load shedding commands, and reactive power regulation. Delayed embedding is performed on the running data to construct a Hankel matrix. The Hankel matrix is ​​divided into multiple two-dimensional subspace patches and converted into tokens. The latent states are obtained by the encoder. The unit start-up and shutdown status, the load shedding command, and the reactive power regulation are encoded into a condition vector; Based on the conditional Koopman operator of control affine, the latent state is transitioned using the conditional vector to obtain the latent state at the future time. The conditional Koopman operator is a linear superposition of the basic operator and the corresponding operators of each conditional component. The basic operator and the corresponding operators of each conditional component are selected from the pattern library and obtained by weighted combination of pattern weights output by event gating. By decoding the latent state at the future time using a quantile prediction head, the conditional distribution prediction result of electricity price fluctuations in the future period is obtained.

[0005] In some embodiments of this application, the conditional Koopman operator is represented as:

[0006] in, Conditional vector The corresponding conditional Koopman operator, For zero-conditional fundamental operators, Let j be the conditional component at time t. For the j-th conditional component The corresponding operator.

[0007] In some embodiments of this application, the transition of the latent state is represented as follows:

[0008] in, Let be the latent state at time t. Let be the latent state at time t+1. Inject a matrix into the event, Let t be the event token at time t, and b be the bias term.

[0009] In some embodiments of this application, the event token is obtained by compression after step or pulse detection of the load shedding command and trip event, and the event token includes at least the event type, amplitude, start and end time and duration.

[0010] In some embodiments of this application, the pattern weight α_r(t) of the r-th pattern operator in the pattern library and the pattern output of the event gating satisfy:

[0011] in, Conditional vector The corresponding conditional Koopman operator, For the basic operator of the r-th pattern, For the operator corresponding to the j-th conditional component in the r-th mode, the mode weight is... Let t be the probability that the power system is in the r-th market operation mode at time t.

[0012] In some embodiments of this application, the latent states at multiple future time points are obtained by multiplying the conditional Koopman operator segment by segment:

[0013] in, Let be the latent state at time t+h, and h be the prediction step. Let represent the condition vector at time t+k. Conditional vector The corresponding conditional Koopman operator.

[0014] In some embodiments of this application, the method further includes: By replacing the conditional vector with the hypothetical conditional vector and linearly replaying the latent state based on the conditional Koopman operator, the counterfactual electricity price fluctuation trajectory is obtained.

[0015] A second aspect of this application provides an electricity price fluctuation prediction device, comprising: The acquisition module is used to acquire the operating data and operating condition data of the power system in historical time periods. The operating data includes electricity price data, load data and power output data. The operating condition data includes unit start-up and shutdown status, load shedding commands and reactive power regulation. The first encoding module is used to perform delayed embedding on the running data to construct a Hankel matrix, divide the Hankel matrix into multiple two-dimensional subspace patches and convert them into tokens, and process them by the encoder to obtain the latent state; The second encoding module is used to encode the unit start-up and shutdown status, the load shedding command and the reactive power adjustment into a condition vector; The prediction module is used to transition the latent state based on the conditional Kupman operator of control affines, using the conditional vector to obtain the latent state at the future time. The conditional Kupman operator is a linear superposition of the basic operator and the corresponding operators of each conditional component. The basic operator and the corresponding operators of each conditional component are selected from the pattern library and obtained by weighted combination of pattern weights output by event gating. The decoding module is used to decode the latent state at the future time using the quantile prediction head to obtain the conditional distribution prediction result of electricity price fluctuations in the future period.

[0016] A third aspect of this application provides an electronic device, including: a processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method described in the first aspect above.

[0017] A fourth aspect of this application provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method described in the first aspect above.

[0018] The electricity price fluctuation prediction method provided in this application encodes operating conditions such as unit start-up and shutdown states, load shedding commands, and reactive power regulation into condition vectors. Based on these vectors, a conditional Koopman operator is constructed that controls the condition vectors affinely. This allows the electricity price prediction model to explicitly utilize known operating commands as input, significantly improving the accuracy of supply and demand situation awareness in event-driven scenarios. The conditional Koopman operator is a linear superposition of the basic operator and the corresponding operators of each condition component. This structure ensures a clear linear dependency of latent state evolution on the condition vectors. When it is necessary to deduce supply and demand changes under different operating commands, only the condition vectors need to be replaced and the operators linearly combined again. The corresponding evolution trajectory can be obtained through a single forward calculation, eliminating the need for repeated iterations or retraining, thus significantly reducing the computational cost of deduction. Furthermore, the encoder converts the Hankel matrix constructed by delayed embedding into latent states, preserving the high-dimensional dynamic information of the system. The quantile prediction head directly decodes the latent states, outputting the conditional distribution prediction results of future supply and demand situations. This effectively characterizes the extreme fluctuation range that may be caused by supply and demand imbalances, providing a more comprehensive situational awareness basis for power system dispatch and operation.

[0019] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0020] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 A flowchart illustrating a method for predicting electricity price fluctuations provided in this application embodiment; Figure 2 This is a schematic diagram of a conditional Koopman transition mechanism proposed in an embodiment of this application; Figure 3 This is a schematic diagram of an electricity price fluctuation prediction device provided in an embodiment of this application. Detailed Implementation

[0021] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0022] Specifically, the electricity price fluctuation prediction method and apparatus of this application are described below with reference to the accompanying drawings.

[0023] Figure 1 This is a flowchart illustrating a method for predicting electricity price fluctuations provided in an embodiment of this application. Figure 1 As shown, the electricity price fluctuation forecasting method may include the following steps: Step 101: Obtain the power system's operating data and operating condition data within the historical time period. The operating data includes electricity price data, load data, and power output data. The operating condition data includes unit start-up and shutdown status, load shedding commands, and reactive power regulation.

[0024] The electricity price data, load data, and power output data are all in continuous time series format. For example, the operational data and operational condition data can be unified to a 15-minute or 1-hour resolution, and the electricity price channel is logarithmically or quantilely normalized according to market standards to reduce the distortion of the embedding geometry caused by heavy tails. The operational data and operational condition data are merged on a unified time grid and supplemented with channel type embeddings, which identify whether a token belongs to a price channel, load channel, output channel, control channel, or status channel.

[0025] Step 102: Perform delayed embedding on the running data to construct a Hankel matrix, divide the Hankel matrix into multiple two-dimensional subspace patches and convert them into tokens, and process them by the encoder to obtain the latent state.

[0026] In some embodiments of this application, each continuous signal channel can be... Constructing the delay vector The embedding dimension m=500 and the delay step τ=1. The L delay vectors are stacked into a Hankel matrix. L = T - m + 1. The Hankel matrix is ​​divided into several non-overlapping two-dimensional subspace patches (default size p × q = 25 × 50). Each patch is flattened and converted into a token by linear projection. Sine position encoding and channel type embedding are added, and after 1D average pooling downsampling, it is fed into the encoder Transformer to obtain the latent state. .

[0027] Step 103: Encode the unit start-up and shutdown status, load shedding command and reactive power regulation into a condition vector.

[0028] In some embodiments of this application, the start-up and shutdown status of the generating units is obtained by looking up a learnable state embedding table to obtain a state token. Homogeneous generating units are first clustered into a group and the aggregated status is maintained. The load shedding command is compressed into an event token (including event type, amplitude, start and end time, and duration) through step or pulse detection. The reactive power regulation is embedded as a continuous quantity with a delay and an additional control flag. The state token and event token are fed into the encoder in parallel with the continuous token in the form of a condition vector. The output is aggregated into a condition vector according to the time step. .

[0029] Step 104: Based on the conditional Kupman operator of control affine, the latent state is transitioned using the conditional vector to obtain the latent state at the future time. The conditional Kupman operator is a linear superposition of the basic operator and the corresponding operators of each conditional component. The basic operator and the corresponding operators of each conditional component are selected from the pattern library and obtained by weighted combination of the pattern weights output by the event gating.

[0030] In some embodiments of this application, the conditional Koopman operator can be expressed as:

[0031] in, Conditional vector The corresponding conditional Koopman operator, For zero-conditional fundamental operators, Let j be the conditional component at time t (e.g., unit start-up / shutdown status, load shedding, reactive power regulation, load, wind and solar power output, fuel price, etc.). For the j-th conditional component The corresponding operator. This formula linearly superimposes the condition vectors onto the basic transitions, making the latent state evolution affine with respect to the condition vectors.

[0032] Figure 2 This is a schematic diagram of a conditional Koopman transition mechanism proposed in an embodiment of this application. The transition of a latent state is represented as follows:

[0033] in, Let be the latent state at time t. Let be the latent state at time t+1. Inject a matrix into the event, Let b be the event token at time t, and b be the bias term. This is used to re-anchor the event token. The latent state is conditionally transitioned using this formula, simultaneously incorporating the runtime conditions and discrete events into the evolution of the latent state.

[0034] In some embodiments, the event token is obtained by compressing the load shedding command and trip event after step or pulse detection. The event token includes at least the event type, amplitude, start and end times, and duration.

[0035] In some embodiments of this application, when it is necessary to predict the latent states at multiple future time points, the latent states at multiple future time points can be obtained by multiplying the conditional Koopman operator piecewise:

[0036] in, Let be the latent state at time t+h, and h be the prediction step. Let represent the condition vector at time t+k. Conditional vector The corresponding conditional Koopman operator. When the condition vector remains approximately constant during the prediction period, the operator is exponentially used. A forward calculation is performed to obtain the full prediction.

[0037] Among them, the first in the pattern library Pattern weights of pattern operators and event gating outputs It can be satisfied:

[0038] in, Conditional vector The corresponding conditional Koopman operator, For the basic operator of the r-th pattern, For the operator corresponding to the j-th conditional component in the r-th mode, the mode weight is... Let be the probability that the power system is in the r-th market operation mode at time t. Market operation modes can include the periodicity and trends of the power market. This formula is a weighted combination of the mode operators based on the mode weights output by the event gating system. It also provides interpretable probabilities of the current market operation mode of the power system. For example, the pattern library can pre-train R≈6~10 pattern operators, and an event-gated classifier. Output mode weights .

[0039] Step 105: Decode the latent state at future time points using the quantile prediction head to obtain the conditional distribution prediction results of electricity price fluctuations in the future period.

[0040] For example, the quantile prediction head can output a future electricity price quantile curve for each quantile τ∈{0.05,0.1,0.25,0.5,0.75,0.9,0.95}, and train it using pinball loss. Simultaneously, the pattern classification head is connected in parallel with the quantile prediction head to output the probability curves of each market operating pattern for the next h steps.

[0041] In one implementation, quantile loss can be used to train the quantile prediction head to power balance the residual terms. As a physical soft constraint, it is added to the total loss, and is represented by the Koopman consistency loss. The transitions of the constrained latent states conform to a linear structure.

[0042] in,

[0043] in, This represents the active power output of the g-th generator unit at time t. This represents the system load at time t. Let represent the network loss power at time t. This formula incorporates the power balance residual as a physical soft constraint into the training loss, enabling the latent state to remember the fundamental physical fact of charge conservation and suppressing the occurrence of passive electricity in counterfactual deduction.

[0044]

[0045] in, This represents the Koopman consistency loss. This formula constrains the actual evolution of the latent state to be consistent with the linear transition given by the conditional Koopman operator, and is the source of interpretability and counterfactual reliability.

[0046] In some embodiments of this application, before decoding the latent state at a future time in step 105, cross-channel cross-attention can be performed on the latent state at the future time. Through cross-channel cross-attention, the latent state of the price channel reads the latent states of the load channel, the output channel, and the unit status channel to obtain the latent state after coupling between channels. The coupled latent state is then sent to the quantile prediction head for decoding, and the final prediction result of the electricity price fluctuation is output.

[0047] In some embodiments of this application, the condition vector can be replaced with a hypothetical condition vector, and the latent state can be linearly replayed based on the conditional Koopman operator to obtain the counterfactual electricity price fluctuation trajectory, which can be used for fault contingency plan review and market rule change impact assessment.

[0048] By implementing the embodiments of this application, operating conditions such as unit start-up and shutdown states, load shedding commands, and reactive power regulation are encoded into condition vectors. A conditional Koopman operator, which is an affine control of the condition vectors, is then constructed. This allows the electricity price prediction model to explicitly utilize known operating commands as input, significantly improving the accuracy of supply and demand situation awareness in event-driven scenarios. The conditional Koopman operator is a linear superposition of the basic operator and the corresponding operators of each condition component. This structure ensures a clear linear dependency of latent state evolution on the condition vectors. When it is necessary to predict supply and demand changes under different operating commands, only the condition vectors need to be replaced and the operators linearly combined again. The corresponding evolution trajectory can be obtained through a single forward calculation, without repeated iterations or retraining, thus significantly reducing the computational cost of prediction. Furthermore, the encoder converts the Hankel matrix constructed by delayed embedding into latent states, preserving the high-dimensional dynamic information of the system. The quantile prediction head directly decodes the latent states, outputting the conditional distribution prediction results of future supply and demand situations. This effectively characterizes the extreme fluctuation range that may be caused by supply and demand imbalances, providing a more comprehensive situational awareness basis for power system dispatch and operation.

[0049] Figure 3 This is a schematic diagram of an electricity price fluctuation prediction device provided in an embodiment of this application. Figure 3As shown, the electricity price fluctuation prediction device may include: an acquisition module 301, a first encoding module 302, a second encoding module 303, a prediction module 304, and a decoding module 305.

[0050] The acquisition module 301 is used to acquire the operating data and operating condition data of the power system in historical time periods. The operating data includes electricity price data, load data and power output data, and the operating condition data includes unit start-up and shutdown status, load shedding command and reactive power regulation. The first encoding module 302 is used to perform delayed embedding on the running data to construct a Hankel matrix, divide the Hankel matrix into multiple two-dimensional subspace patches and convert them into tokens, and obtain the latent state through encoder processing; The second encoding module 303 is used to encode the unit start-up and shutdown status, load shedding command and reactive power regulation into a condition vector; The prediction module 304 is used to use the conditional Kupman operator based on control affines to transition the latent state using the conditional vector to obtain the latent state at the future time. The conditional Kupman operator is a linear superposition of the basic operator and the corresponding operators of each conditional component. The basic operator and the corresponding operators of each conditional component are selected from the pattern library and obtained by weighted combination of the pattern weights output by the event gating. The decoding module 305 is used to decode the latent state at future time points through the quantile prediction head to obtain the conditional distribution prediction result of electricity price fluctuations in future periods.

[0051] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0052] To implement the above embodiments, this application also proposes an electronic device, including: a processor and a memory communicatively connected to the processor; the memory stores computer execution instructions; the processor executes the computer execution instructions stored in the memory to implement the method provided in the foregoing embodiments.

[0053] To implement the above embodiments, this application also proposes a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the methods provided in the foregoing embodiments.

[0054] To implement the above embodiments, this application also proposes a computer program product, including a computer program that, when executed by a processor, implements the methods provided in the foregoing embodiments.

[0055] In the foregoing descriptions of the embodiments, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0056] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0057] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0058] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0059] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0060] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0061] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0062] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A method for predicting electricity price fluctuations, characterized in that, Includes the following steps: The system acquires operating data and operating condition data of the power system within a historical period. The operating data includes electricity price data, load data, and power output data. The operating condition data includes unit start-up and shutdown status, load shedding commands, and reactive power regulation. Delayed embedding is performed on the running data to construct a Hankel matrix. The Hankel matrix is ​​divided into multiple two-dimensional subspace patches and converted into tokens. The latent states are obtained by the encoder. The unit start-up and shutdown status, the load shedding command, and the reactive power regulation are encoded into a condition vector; Based on the conditional Koopman operator of control affine, the latent state is transitioned using the conditional vector to obtain the latent state at the future time. The conditional Koopman operator is a linear superposition of the basic operator and the corresponding operators of each conditional component. The basic operator and the corresponding operators of each conditional component are selected from the pattern library and obtained by weighted combination of pattern weights output by event gating. By decoding the latent state at the future time using a quantile prediction head, the conditional distribution prediction result of electricity price fluctuations in the future period is obtained.

2. The method according to claim 1, characterized in that, The conditional Koopman operator is expressed as follows: in, Conditional vector The corresponding conditional Koopman operator, For zero-conditional fundamental operators, Let j be the conditional component at time t. For the j-th conditional component The corresponding operator.

3. The method according to claim 2, characterized in that, The transition of the latent state is represented as follows: in, Let be the latent state at time t. Let be the latent state at time t+1. Inject a matrix into the event. Let t be the event token at time t, and b be the bias term.

4. The method according to claim 3, characterized in that, The event token is obtained by compressing the load shedding command and trip event after step or pulse detection. The event token includes at least the event type, amplitude, start and end time and duration.

5. The method according to claim 3, characterized in that, The r-th pattern operator in the pattern library and the pattern weight α_r(t) output by the event gating satisfy the following: in, Conditional vector The corresponding conditional Koopman operator, For the basic operator of the r-th pattern, For the operator corresponding to the j-th conditional component in the r-th mode, the mode weight is... Let t be the probability that the power system is in the r-th market operation mode at time t.

6. The method according to claim 3, characterized in that, The latent states at multiple future time points are obtained by multiplying the conditional Koopman operator piece by piece: in, Let be the latent state at time t+h, and h be the prediction step. Let represent the condition vector at time t+k. Conditional vector The corresponding conditional Koopman operator.

7. The method according to claim 1, characterized in that, The method further includes: By replacing the conditional vector with the hypothetical conditional vector and linearly replaying the latent state based on the conditional Koopman operator, the counterfactual electricity price fluctuation trajectory is obtained.

8. A device for predicting electricity price fluctuations, characterized in that, include: The acquisition module is used to acquire the operating data and operating condition data of the power system in historical time periods. The operating data includes electricity price data, load data and power output data. The operating condition data includes unit start-up and shutdown status, load shedding commands and reactive power regulation. The first encoding module is used to perform delayed embedding on the running data to construct a Hankel matrix, divide the Hankel matrix into multiple two-dimensional subspace patches and convert them into tokens, and process them by the encoder to obtain the latent state; The second encoding module is used to encode the unit start-up and shutdown status, the load shedding command and the reactive power adjustment into a condition vector; The prediction module is used to transition the latent state based on the conditional Kupman operator of control affines, using the conditional vector to obtain the latent state at the future time. The conditional Kupman operator is a linear superposition of the basic operator and the corresponding operators of each conditional component. The basic operator and the corresponding operators of each conditional component are selected from the pattern library and obtained by weighted combination of pattern weights output by event gating. The decoding module is used to decode the latent state at the future time using the quantile prediction head to obtain the conditional distribution prediction result of electricity price fluctuations in the future period.

9. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-7.

Images