Market power early warning method and device based on neural network algorithm

Through the market power warning method based on neural network algorithm, the electricity price is predicted using factors such as power declaration data, weather and policies, and the market power of power generators is evaluated, which solves the problem of early warning of market power for power generators in the power market, and achieves a high-accurate market Force warning and electricity price stability.

CN118917871APending Publication Date: 2024-11-08CHINA SOUTHERN POWER GRID COMPANY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410929278.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-11
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

It is difficult for existing technology to effectively warn power generators to exercise market power in the power market, resulting in electricity price fluctuations and power outage risks. The existing market power index will be judged afterwards and cannot be used for early warning.

Method used

Using a market power early warning method based on neural network algorithm, the power market operation boundary conditions are constructed by obtaining power generation companies' power declaration data, off-site weather data and power policies, training the electricity price prediction model, predicting electricity prices and calculating market power monitoring indicators, and evaluating The possibility of power generation companies exercising market power and allocating attention.

Benefits of technology

It has achieved a high-accurate warning of the exercise of market power by power generators, predict electricity prices through various factors, evaluate market power risks, and help dispatching agencies reasonably arrange unit maintenance and operation strategies to reduce electricity price fluctuations and power outage risks.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118917871A_ABST
    Figure CN118917871A_ABST
Patent Text Reader

Abstract

The invention relates to a market power early warning method and device based on a neural network algorithm, and the method comprises the steps: obtaining power declaration data, off-site weather data, a current power policy and a power generation side new installation condition of a power generator, and constructing a power market operation boundary condition; taking the power market operation boundary condition as input data of the electricity price prediction model, and enabling the electricity price prediction model to output a predicted electricity price; inputting the predicted electricity price into a market power monitoring index of a power generator to obtain an index result; and according to the index result, determining the possibility that the power generator travels the market power, and distributing corresponding attention to the power generator. In the technical scheme, in order to obtain the predicted electricity price with high accuracy, various factors are considered, the market power monitoring index result of the power generator is calculated according to the predicted electricity price output by the electricity price prediction model, and the index result can evaluate the risk degree of market power of each power generator. And early warning is carried out on whether the power generator carries out market power or not by distributing the corresponding attention to the power generator.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of market power early warning, and in particular to a market power early warning method and device based on a neural network algorithm. Background Art

[0002] Market power (also known as market manipulation power) refers to the ability of power generators to change market prices and make them deviate from the price level under full market competition. In a broad sense, market power includes horizontal market power and vertical market power.

[0003] With the deepening of the market-oriented reform of the power industry, the power market has developed rapidly. In the power market, the use of market power can easily lead to extremely high electricity prices in the short term or even in the medium and long term. In extreme cases, power outages may even occur due to insufficient power supply. In order to ensure the normal operation of the power market, it is very important to warn the risk level of the power generation companies in the power market in exercising their market power and to monitor the market power.

[0004] In the relevant technology, although there are many market power indexes applied to other markets, they are less reliable for the power market. Moreover, most of these indicators are calculated after the fact to judge whether the power generator has exercised market power, which cannot play a warning role. Summary of the invention

[0005] In order to at least to some extent overcome the problem in the related art that it is difficult to provide market power warning for power generators in the power market, the present application provides a market power warning method and device based on a neural network algorithm.

[0006] The scheme of this application is as follows:

[0007] According to a first aspect of an embodiment of the present application, a market power early warning method based on a neural network algorithm is provided, comprising:

[0008] Obtain power declaration data from power generators, off-site weather data, current power policies, and new installed capacity on the power generation side to construct boundary conditions for power market operation;

[0009] Using the power market operation boundary conditions as input data of an electricity price prediction model, so that the electricity price prediction model outputs a predicted electricity price;

[0010] Inputting the predicted electricity price into the market power monitoring index of the power generator to obtain the index result;

[0011] Based on the results of the indicators, the possibility of the generator exercising market power is determined, and the corresponding attention is allocated to the generator.

[0012] Preferably, the method further comprises:

[0013] Generate unit maintenance and operation strategies based on predicted electricity prices.

[0014] Preferably, the predicted electricity price includes:

[0015] Intra-provincial electricity prices and outbound electricity prices.

[0016] Preferably, the method further comprises:

[0017] The electricity price prediction model is trained based on a BP neural network algorithm.

[0018] Preferably, the electricity price prediction model is trained based on a BP neural network algorithm, including:

[0019] Obtain power declaration data from power generators, off-site weather data, historical power policies, and new installed capacity on the power generation side as training data;

[0020] Constructing the input mode, output layer, network output, target output, transfer function from hidden layer to output layer and transfer function of the output layer of the BP neural network;

[0021] Reduce the error between the network output and the target output by adjusting the weights;

[0022] The training data is used as the input of the electricity price prediction model and it is iterated until the error between the network output and the target output is reduced to a preset requirement.

[0023] Preferably, the weights are adjusted so that the error between the network output and the target output is reduced to a preset value, including:

[0024] First adjust the weights from the hidden layer to the output layer, and then adjust the weights from the input layer to the hidden layer.

[0025] Preferably, after obtaining the power declaration data of the power generator, the method further comprises:

[0026] Calculate the average electricity reporting level of generators.

[0027] Preferably, the predicted electricity price is input into the market power monitoring index of the power generator to obtain the index result, including:

[0028] The predicted electricity price is input into the RSI index indicator of the power generator to obtain the RSI index of the power generator.

[0029] According to a second aspect of an embodiment of the present application, a market power early warning device based on a neural network algorithm is provided, comprising:

[0030] Processor and memory;

[0031] The processor and the memory are connected via a communication bus:

[0032] Wherein, the processor is used to call and execute the program stored in the memory;

[0033] The memory is used to store a program, and the program is at least used to execute a market power early warning method based on a neural network algorithm as described in any one of the above items.

[0034] The technical solution provided by this application may have the following beneficial effects:

[0035] The market power early warning method based on the neural network algorithm in the present application includes: obtaining the power declaration data of the power generator, off-site weather data, current power policy and new installed capacity on the power generation side, and constructing the boundary conditions of the power market operation; using the boundary conditions of the power market operation as the input data of the electricity price prediction model, so that the electricity price prediction model outputs the predicted electricity price; inputting the predicted electricity price into the market power monitoring index of the power generator to obtain the index result; according to the index result, determining the possibility of the power generator exercising market power, and allocating corresponding attention to the power generator.

[0036] In this technical solution, in order to obtain a highly accurate predicted electricity price, various factors are taken into account, such as the power declaration data of the power generator, off-site weather data, current electricity policy and new installed capacity on the power generation side. Then, the market power monitoring index results of the power generator are calculated based on the predicted electricity price output by the electricity price prediction model. The index results can assess the risk level of each power generator in exercising its market power, and issue early warnings on whether the power generator exercises its market power by allocating corresponding attention to the power generator.

[0037] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0039] Figure 1 It is a flowchart of a market power early warning method based on a neural network algorithm provided by an embodiment of the present application;

[0040] Figure 2 It is a structural diagram of a market power early warning device based on a neural network algorithm provided by an embodiment of the present application.

[0041] Reference numerals: processor-21; memory-22. DETAILED DESCRIPTION

[0042] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0043] Embodiment 1

[0044] A market power early warning method based on neural network algorithm, referring to Figure 1 ,include:

[0045] S11: Obtain the power declaration data of power generators, off-site weather data, current power policies and new installed capacity on the power generation side to construct the boundary conditions for power market operation;

[0046] It should be noted that after obtaining the power declaration data of the power generator, the method also includes:

[0047] Calculate the average electricity reporting level of generators.

[0048] S12: using the power market operation boundary conditions as input data of the power price prediction model, so that the power price prediction model outputs the predicted power price;

[0049] It should be noted that the predicted electricity price includes:

[0050] Intra-provincial electricity prices and outbound electricity prices.

[0051] S13: input the predicted electricity price into the market power monitoring index of the power generator to obtain the index result;

[0052] S14: Based on the indicator results, determine the possibility of the generator exercising market power and allocate corresponding attention to the generator.

[0053] The market power early warning method based on the neural network algorithm in this application starts from the average power declaration level of power generators, combines off-site weather data, current power policies and new installed capacity on the power generation side, constructs the boundary conditions for power market operation, and uses them as input data for the power price prediction model to predict the power prices within each province and the power delivery price. Since the neural network obtains the predicted data by training the input data, the accuracy of the predicted data is closely related to the input data. In order to obtain more accurate predicted values ​​of the power prices within each province and the power delivery price in the power market, it is necessary to consider multiple factors such as the power declaration data of power generators, off-site weather data, current power policies and new installed capacity on the power generation side.

[0054] It should be noted that the method also includes:

[0055] Generate unit maintenance and operation strategies based on predicted electricity prices.

[0056] In this technical solution, after the predicted electricity price is obtained using the electricity price prediction model, the dispatching agency is assisted in reasonably arranging unit maintenance and adjusting the operation strategy based on the predicted electricity price.

[0057] It should be noted that the method also includes:

[0058] The electricity price prediction model is trained based on the BP neural network algorithm.

[0059] The electricity price prediction model is trained based on the BP neural network algorithm, including:

[0060] Obtain power declaration data from power generators, off-site weather data, historical power policies, and new installed capacity on the power generation side as training data;

[0061] Construct the input mode, output layer, network output, target output, transfer function from hidden layer to output layer and transfer function of output layer of BP neural network;

[0062] Reduce the error between the network output and the target output by adjusting the weights;

[0063] The training data is used as the input of the electricity price prediction model and it is iterated until the error between the network output and the target output is reduced to the preset requirement.

[0064] The weights are adjusted to reduce the error between the network output and the target output to a preset value, including:

[0065] First adjust the weights from the hidden layer to the output layer, and then adjust the weights from the input layer to the hidden layer.

[0066] Neurons are the basic units of neural networks. Neurons first receive input, then perform certain mathematical operations and then generate an output.

[0067] Assume x1~xn are input signals from other neurons, w ij represents the connection weight from neuron j to neuron i, θ represents a value or called bias, then the relationship between the output and input of neuron i is expressed as:

[0068]

[0069] y i =f(net i )(2);

[0070] y i represents the output of neuron i, function f is called activation function or transfer function, and net is called net activation. If the threshold is regarded as the weight w of an input x0 of the neuron i0, then formula (1) can be simplified as:

[0071]

[0072] If X represents the input vector and W represents the weight vector, that is:

[0073] X=[x0 x1 x2…x n ](4);

[0074]

[0075] The output of the neuron can be expressed as a vector multiplication:

[0076] net i = = XW(6);

[0077] Y i =f(net i ) = f(XW)(7);

[0078] If the net activation net of a neuron is positive, the neuron is said to be in an activated state, and if the net activation net is negative, the neuron is said to be in an inhibited state.

[0079] According to the interconnection method of neurons in the network, the network structure of common neural networks can be mainly divided into the following three categories:

[0080] (1) Feedforward Neural Network

[0081] Feedforward network is also called forward network. This kind of network only has feedback signal in the training process, and in the classification process, data can only be transmitted forward until it reaches the output layer. There is no backward feedback signal between layers, so it is called feedforward network. BP neural network belongs to feedforward network. At present, in the practical application of artificial neural network, most of the neural network models use BP network and its variations.

[0082] (2) Feedback Neural Network

[0083] A feedback neural network is a neural network with feedback connections from output to input. Its structure is much more complex than that of a feedforward network. Typical feedback neural networks include: Elman network and Hopfield network

[0084] (3) Self-organizing neural network

[0085] A self-organizing neural network is a kind of unsupervised learning network. It self-organizes and adaptively changes the network parameters and structure by automatically finding the inherent laws and essentiality in samples.

[0086] The neural network algorithm is a general optimization algorithm, which has the following characteristics:

[0087] (1) The biggest difference between the neural network algorithm and the traditional parameter model method is that it is a data-driven adaptive technology that does not require any prior assumptions about the problem model. When the internal rules of solving the problem are unknown or difficult to describe, the neural network can obtain the implicit functional relationship between the data through learning and training of sample data. Therefore, the neural network method is particularly suitable for solving some problems that are difficult to explain using assumptions and existing theories, but have sufficient data and observation variables.

[0088] (2) Neural network technology has the ability to generalize. Generalization ability refers to the ability of a trained learning model to correctly respond to samples that appear in the future training set. Therefore, it is possible to predict future data outside the sample through historical data within the sample. Neural networks can obtain the rules hidden in the data through learning and training of input sample data, and use the learned rules to predict future data. Therefore, generalization ability makes neural networks an ideal prediction technology.

[0089] (3) Neural network is a function approximator with universal applicability. It can approximate any continuous function with arbitrary precision. When dealing with the same problem, the internal function form of neural network is more flexible and effective than traditional statistical methods. Traditional statistical prediction models cannot effectively estimate complex variable function relationships due to various limitations. The powerful function approximation ability of neural network provides an effective method for identifying internal functions of complex systems.

[0090] (4) Neural networks are nonlinear methods. Each neuron in a neural network can receive input from a large number of other neurons, and the input and output of each neuron are nonlinear. This mutual restriction and influence between neurons can achieve nonlinear mapping of the entire network from input state to output state space. Therefore, neural networks can handle some problems with very complex environmental information, unclear knowledge background, and unclear reasoning rules.

[0091] The market power early warning system based on the neural network algorithm uses the neural network algorithm to predict electricity prices. According to statistics, 80% to 90% of the neural network models use the BP network or its variants. This application also uses the BP neural network algorithm.

[0092] The main idea of ​​BP neural network algorithm is: for n input learning samples "x1, x2, .. xn", it is known that the corresponding m output samples are "1, 2, .. tm". The error between the actual output of the network (z1, z2, .., zm) and the target vector (t1, t2, .., tm) is used to modify its weights, so that zl (l = 1, 2, .., m) is as close as possible to the expected tl, that is, to minimize the sum of square errors of the network output layer.

[0093] The learning process of BP neural network mainly consists of four parts: forward propagation of input mode, reverse propagation of output error, cyclic memory training, and judgment of learning results. The learning process of this algorithm consists of forward propagation and reverse propagation. In the forward propagation process, the input information is processed layer by layer from the input layer through the hidden layer unit and transmitted to the output layer. The state of each layer of neurons only affects the state of the next layer of neurons. If the expected output cannot be obtained in the output layer, it will turn to reverse propagation, return the error signal along the original connection path, and reduce the error signal by modifying the weights of each layer of neurons, and then turn to the forward propagation process. Repeated iteration until the error is less than the given value. Feedforward neural networks using BP learning algorithm are usually called BP networks. BP network has strong nonlinear mapping ability. A 3-layer BP neural network can approximate any nonlinear function.

[0094] In this embodiment, the electricity price prediction model is trained based on the BP neural network algorithm, including:

[0095] Assume the input pattern of the network is x = (x1, x2, ..., xn) T , the output layer has h units, and the network output is z = (z1, z2, .., zm) T , the target output is t = (t1, t2.., tm) T , the transfer function from the hidden layer to the output layer is f, and the transfer function of the output layer is g. So we can get:

[0096]

[0097] In the formula, y i represents the output of the jth neuron in the hidden layer, w 0j =θ, x0=-1.

[0098]

[0099] In the formula, z k represents the output of the kth neuron in the output layer. At this time, the error between the network output and the target output is:

[0100]

[0101] The following steps are to adjust the weights to reduce ε. Since the negative gradient direction is the direction in which the function value decreases fastest, we can set a step size η and adjust η units along the negative gradient direction each time, that is, the weight adjustment each time is:

[0102]

[0103] Among them, η is called the learning rate in the neural network. It can be proved that if adjusted in this way, the error will gradually decrease. Therefore, the adjustment order of the BP neural network (back propagation) is:

[0104] (1) First adjust the weights from the hidden layer to the output layer. Let v k is the input of the kth neuron in the output layer, then

[0105]

[0106] w jk (t+1)=w jk (t)+ηδ k y j (14);

[0107] (2) The iterative formula for weight adjustment from the input layer to the hidden layer is:

[0108]

[0109] where u j Is the input of the th neuron in the hidden layer:

[0110]

[0111] The iterative formula for weight adjustment from the input layer to the hidden layer is:

[0112] w ij (t+1)=w ij (t)+ηδ j x i (17).

[0113] According to the above steps, after considering the comprehensive influencing factors such as the power declaration data of power generators, off-site weather data, historical power policies and new installed capacity on the power generation side, the boundary conditions for the operation of the power market can be constructed and used as training data for the electricity price prediction model. After iteration, the predicted electricity price can be obtained after the training is completed.

[0114] It should be noted that the predicted electricity price is input into the market power monitoring index of the power generator to obtain the index results, including:

[0115] The predicted electricity price is input into the RSI index indicator of the power company to obtain the RSI index of the power company.

[0116] The RSI index (surplus supply rate) refers to the ratio of the supply capacity and demand of other power generation companies except for a certain power generation company in a certain period of time in the power market. The RSI can accurately reflect the ability of a power generator in the southern power spot market to manipulate the market, thereby providing a basis for effective supervision. The smaller the RSI index of a power generation group, the stronger its ability to control market prices. Generally speaking, when the surplus supply rate index of a power generation group is between 1.2 and 1.5, the market is in a reasonable competitive state; when the index is less than 1, it indicates that the market demand will not be met without the power generation group, and the group has a strong control over the market price. From the perspective of the i-th power generation group, its RSI is defined as:

[0117]

[0118] In the formula, Q k is the maximum power generation capacity of the kth power generation group; D t and R t They are the spot market bidding space and backup demand during the trial operation period t.

[0119] In this technical solution, the RSI index of each power generator is calculated according to the predicted electricity price to evaluate the possibility of each power generator's market power utilization, thereby achieving the purpose of market power early warning.

[0120] In summary, in this technical solution, in order to obtain a highly accurate predicted electricity price, various factors are taken into consideration, such as the power declaration data of the power generator, off-site weather data, current electricity policy and new installed capacity on the power generation side. Then, the market power monitoring index results of the power generator are calculated based on the predicted electricity price output by the electricity price prediction model. The index results can assess the risk level of each power generator in exercising its market power, and issue early warnings on whether the power generator exercises its market power by allocating corresponding attention to the power generator.

[0121] Embodiment 2

[0122] A market power early warning device based on a neural network algorithm, referring to Figure 2 ,include:

[0123] Processor 21 and memory 22;

[0124] The processor 21 and the memory 22 are connected via a communication bus:

[0125] The processor 21 is used to call and execute the program stored in the memory 22;

[0126] The memory 22 is used to store a program, and the program is used to execute at least one of the market power early warning methods based on a neural network algorithm in the above embodiments.

[0127] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0128] It should be noted that, in the description of this application, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of this application, unless otherwise specified, the meaning of "plurality" refers to at least two.

[0129] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0130] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0131] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0132] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0133] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0134] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0135] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A market power early warning method based on a neural network algorithm, characterized in that: include: Obtain power declaration data from power generators, off-site weather data, current power policies, and new installed capacity on the power generation side to construct boundary conditions for power market operation; Using the power market operation boundary conditions as input data of an electricity price prediction model, so that the electricity price prediction model outputs a predicted electricity price; Inputting the predicted electricity price into the market power monitoring index of the power generator to obtain the index result; Based on the results of the indicators, the possibility of the generator exercising market power is determined, and the corresponding attention is allocated to the generator.

2. The method according to claim 1, characterized in that The method further comprises: Generate unit maintenance and operation strategies based on predicted electricity prices.

3. The method according to claim 1, characterized in that The predicted electricity price includes: Intra-provincial electricity prices and outbound electricity prices.

4. The method according to claim 1, characterized in that: The method further comprises: The electricity price prediction model is trained based on a BP neural network algorithm.

5. The method according to claim 1, characterized in that The electricity price prediction model is trained based on the BP neural network algorithm, including: Obtain power declaration data from power generators, off-site weather data, historical power policies, and new installed capacity on the power generation side as training data; Constructing the input mode, output layer, network output, target output, transfer function from hidden layer to output layer and transfer function of the output layer of the BP neural network; Reduce the error between the network output and the target output by adjusting the weights; The training data is used as the input of the electricity price prediction model and it is iterated until the error between the network output and the target output is reduced to a preset requirement.

6. The method according to claim 1, characterized in that The error between the network output and the target output is reduced to a preset value by adjusting the weights, including: First adjust the weights from the hidden layer to the output layer, and then adjust the weights from the input layer to the hidden layer.

7. The method according to claim 1, characterized in that After obtaining the power declaration data of the power generator, the method further includes: Calculate the average electricity reporting level of generators.

8. The method according to claim 1, characterized in that The predicted electricity price is input into the market power monitoring index of the power generator to obtain the index results, including: The predicted electricity price is input into the RSI index indicator of the power generator to obtain the RSI index of the power generator.

9. A market power early warning device based on a neural network algorithm, characterized in that: include: Processor and memory; The processor and the memory are connected via a communication bus: Wherein, the processor is used to call and execute the program stored in the memory; The memory is used to store a program, and the program is at least used to execute a market power early warning method based on a neural network algorithm as described in any one of claims 1-8.