A method and device for simulating power system production based on power market transactions

Abstract

The present invention provides a method and device for simulating the production of an electric power system based on electric power market transactions, which obtains the historical price curve of electric power sellers and the historical load curve of electric power buyers; pre-processes the historical price curve and the historical load curve; classifies the pre-processed historical price curve and the historical load curve respectively and determines the predicted price curve and the predicted load curve; sorts the electric power sellers by priority; determines the power shortage according to the predicted load curve and the medium- and long-term contracts, and purchases the power corresponding to the power shortage according to the priority; determines the power purchase cost and the power purchase reliability index according to the medium- and long-term contracts, the predicted price curve, the predicted load curve and the unit characteristic index of the electric power sellers. The power quantity to be purchased, the power purchase cost and the power purchase reliability index are determined by introducing the data of electric power market transactions, so as to improve the reliability of the electric power system and the calculation accuracy of the power purchase cost.

Description

Technical Field

[0001] The present invention relates to the technical field of power system reliability analysis, and in particular to a power system production simulation method and device based on power market transactions. Background Art

[0002] With the reform of the electricity market and the advancement of electricity trading, the current form of electricity trading is mainly medium- and long-term electricity trading.

[0003] The power system needs to ensure real-time supply and demand balance. Decomposing the medium- and long-term contracted electricity into daily and hourly granularity will lead to deviations between the real-time operation of the power system and the dispatch plan. In order to solve the aforementioned deviations, the existing power system introduces the spot market to ensure real-time balance of electricity. However, the introduction of the spot market will increase the randomness and volatility of costs, which will lead to lower reliability of the calculated power system and lower accuracy of the power purchase cost. Summary of the invention

[0004] In view of this, an embodiment of the present invention provides a method and device for simulating power system production based on power market transactions to solve the problems of low accuracy of power system reliability and power purchase cost calculated by existing power system production simulation methods.

[0005] To achieve the above objectives, the embodiments of the present invention provide the following technical solutions:

[0006] A first aspect of an embodiment of the present invention discloses a method for simulating power system production based on power market transactions, the method comprising:

[0007] Obtaining the historical price curve of electricity sellers and the historical load curve of electricity buyers;

[0008] Preprocessing the historical price curve and the historical load curve respectively;

[0009] Classifying the preprocessed historical price curve and determining a predicted price curve according to the classification result, and classifying the preprocessed historical load curve and determining a predicted load curve according to the classification result;

[0010] According to the electricity selling price of the electricity selling user and the risk index of the power generation system, the electricity selling user is sorted according to the priority;

[0011] Determine the power shortage according to the predicted load curve and the pre-acquired medium- and long-term contracts, and purchase the power corresponding to the power shortage from the power seller user according to the priority;

[0012] The power purchase cost and the power purchase reliability index are determined based on the medium- and long-term contracts, the predicted price curve, the predicted load curve and the unit characteristic index of the power seller user.

[0013] Preferably, preprocessing the historical price curve and the historical load curve respectively includes:

[0014] Obtaining the highest historical price in the historical price curve, and obtaining the highest historical load in the historical load curve;

[0015] Normalizing the historical prices in the historical price curve according to the highest historical price, and normalizing the historical loads in the historical load curve according to the highest historical load;

[0016] According to a preset autoregressive moving average model, data processing is performed on the historical prices in the normalized historical price curve, and data processing is performed on the historical loads in the normalized historical load curve, wherein the data processing includes one or more of deleting bad data, restoring missing data, and repairing abnormal data.

[0017] Preferably, classifying the preprocessed historical price curve and determining the predicted price curve according to the classification result includes:

[0018] Extracting a preset price index from the preprocessed historical price curve;

[0019] According to the extracted preset price index, combined with the first sample data, a first neural network is constructed and trained to obtain a first clustering model;

[0020] Inputting the preprocessed historical price curve into the first clustering model for clustering to obtain multiple price patterns, each price pattern including at least one preprocessed historical price curve;

[0021] For each type of the price pattern, averaging the historical prices of the historical price curve in the price pattern to obtain a historical price average;

[0022] Determining a first weight of each type of the price pattern according to the number of the historical price curves in each type of the price pattern;

[0023] For each type of the price pattern, based on the historical price average corresponding to the price pattern and the deviation distribution of the historical prices of the historical price curve, a price curve corresponding to the price pattern is simulated;

[0024] Based on the first weight of each type of price pattern, the price curves corresponding to all the simulated price patterns are weighted averaged, and the highest historical price in the historical price curves is used to process the weighted average result to obtain a predicted price curve.

[0025] Preferably, classifying the preprocessed historical load curve and determining the predicted load curve according to the classification result includes:

[0026] Extracting a preset load index from the preprocessed historical load curve;

[0027] According to the extracted preset load index, combined with the second sample data, a second neural network is constructed and trained to obtain a second clustering model;

[0028] Inputting the preprocessed historical load curve into the second clustering model for clustering to obtain multiple types of load patterns, each type of load pattern includes at least one preprocessed historical load curve;

[0029] For each type of the load mode, averaging the historical loads of the historical load curves in the load mode to obtain a historical load average value;

[0030] determining a second weight for each type of the load pattern according to the number of the historical load curves in each type of the load pattern;

[0031] For each type of the load pattern, based on the historical load average value corresponding to the load pattern and the deviation distribution of the historical load of the historical load curve, a load curve corresponding to the load pattern is simulated;

[0032] Based on the second weight of each type of the load mode, the load curves corresponding to all the simulated load modes are weighted averaged, and the highest historical load in the historical load curves is used to process the weighted average result to obtain a predicted load curve.

[0033] Preferably, determining the power purchase cost and the power purchase reliability index according to the medium- and long-term contract, the predicted price curve, the predicted load curve and the unit characteristic index of the power seller user includes:

[0034] Determine the electricity purchase cost based on the medium- and long-term contracts and the predicted price curve;

[0035] The unit operating status is obtained by simulating the unit characteristic index of the electricity seller user;

[0036] Based on the available capacity of the units put into operation every hour by the power seller user, the total available capacity per hour is calculated;

[0037] The electricity of the medium- and long-term contracts and the total available capacity per hour are summed to obtain the actual electricity purchase per hour;

[0038] Calculate the difference between the actual power purchase amount per hour and the predicted load of the predicted load curve to obtain the power shortage per hour;

[0039] According to the hourly power shortage and power shortage hours, the average power shortage and power shortage probability are calculated;

[0040] Return to the step of simulating the operation status of the unit by using the unit characteristic index of the power seller user until the step is performed K times, and K average power shortages and K power shortage probabilities are obtained;

[0041] The average value of the K average power shortages is calculated to obtain the expected power shortage value, and the average value of the K power shortage probabilities is calculated to obtain the power shortage probability.

[0042] Preferably, the time granularity of the historical price curve is day granularity, and the total time scale of the historical price curve is one year; the time granularity of the historical load curve is hour granularity, and the total time scale of the historical load curve is one day.

[0043] A second aspect of an embodiment of the present invention discloses a power system production simulation device based on power market transactions, the device comprising:

[0044] An acquisition unit, used to acquire a historical price curve of a power seller user and a historical load curve of a power buyer user;

[0045] A preprocessing unit, used for preprocessing the historical price curve and the historical load curve respectively;

[0046] A prediction unit, configured to classify the preprocessed historical price curves and determine a predicted price curve according to the classification results, and to classify the preprocessed historical load curves and determine a predicted load curve according to the classification results;

[0047] A sorting unit, used to sort the power seller users by priority according to the power selling price of the power seller users and the risk index of the power generation system;

[0048] A processing unit, configured to determine the power shortage according to the predicted load curve and the pre-acquired medium- and long-term contracts, and purchase the power corresponding to the power shortage from the power seller user according to the priority;

[0049] A determination unit is used to determine the power purchase cost and the power purchase reliability index according to the medium- and long-term contract, the predicted price curve, the predicted load curve and the unit characteristic index of the power seller user.

[0050] Preferably, the pre-processing unit comprises:

[0051] An acquisition module, used to acquire the highest historical price in the historical price curve, and to acquire the highest historical load in the historical load curve;

[0052] a normalization module, configured to normalize the historical prices in the historical price curve according to the highest historical price, and to normalize the historical loads in the historical load curve according to the highest historical load;

[0053] A processing module is used to perform data processing on the historical prices in the normalized historical price curve and the historical loads in the normalized historical load curve according to a preset autoregressive sliding average model, wherein the data processing includes one or more of deleting bad data, restoring missing data and repairing abnormal data.

[0054] Preferably, the prediction unit for classifying the preprocessed historical price curve and determining the predicted price curve according to the classification result includes:

[0055] A first extraction module, used to extract a preset price index from the preprocessed historical price curve;

[0056] A first construction module is used to construct and train a first neural network to obtain a first clustering model based on the extracted preset price index and the first sample data;

[0057] A first clustering module, used for inputting the preprocessed historical price curve into the first clustering model for clustering to obtain multiple price patterns, each of which contains at least one preprocessed historical price curve;

[0058] A first calculation module is used for calculating the average of the historical prices of the historical price curve in each price mode to obtain the historical price average value;

[0059] A first determining module, configured to determine a first weight of each type of the price pattern according to the number of the historical price curves in each type of the price pattern;

[0060] A first prediction module is used for simulating, for each type of the price pattern, a price curve corresponding to the price pattern based on the historical price average corresponding to the price pattern and the deviation distribution of the historical prices of the historical price curve;

[0061] The first weighting module is used to perform weighted averaging of the price curves corresponding to all the simulated price patterns based on the first weight of each type of the price pattern, and use the highest historical price in the historical price curves to process the weighted average result to obtain a predicted price curve.

[0062] Preferably, the prediction unit for classifying the preprocessed historical load curve and determining the predicted load curve according to the classification result includes:

[0063] A second extraction module, used to extract a preset load index from the preprocessed historical load curve;

[0064] A second construction module is used to construct and train a second neural network to obtain a second clustering model based on the extracted preset load index and in combination with second sample data;

[0065] A second clustering module, used for inputting the preprocessed historical load curve into the second clustering model for clustering to obtain multiple types of load patterns, each type of load pattern including at least one preprocessed historical load curve;

[0066] A second calculation module is used to calculate the average value of the historical load of the historical load curve in the load mode for each type of the load mode to obtain the historical load average value;

[0067] A second determination module, configured to determine a second weight of each type of the load mode according to the number of the historical load curves in each type of the load mode;

[0068] A second prediction module is used for simulating, for each type of the load mode, a load curve corresponding to the load mode based on the historical load average value corresponding to the load mode and the deviation distribution of the historical load of the historical load curve;

[0069] The second weighting module is used to perform weighted averaging of the load curves corresponding to all the simulated load modes based on the second weight of each type of the load mode, and use the highest historical load in the historical load curves to process the weighted average result to obtain a predicted load curve.

[0070] Based on the above embodiment of the present invention, a power system production simulation method based on power market transactions is provided. The method is as follows: obtaining the historical price curve of the power seller user, and obtaining the historical load curve of the power buyer user; preprocessing the historical price curve and the historical load curve respectively; classifying the preprocessed historical price curve and determining the predicted price curve according to the classification result, and classifying the preprocessed historical load curve and determining the predicted load curve according to the classification result; sorting the power seller users by priority according to the power selling price and the risk index of the power generation system of the power seller user; determining the power shortage according to the predicted load curve and the pre-acquired medium- and long-term contracts, and purchasing the power corresponding to the power shortage according to the priority of the power seller user; determining the power purchase cost and the power purchase reliability index according to the medium- and long-term contracts, the predicted price curve, the predicted load curve and the unit characteristic index of the power seller user. This scheme determines the predicted price curve and the predicted load curve by introducing the data of power market transactions; and determines the power purchase cost and the power purchase reliability index to be purchased based on the predicted price curve, the predicted load curve and the unit characteristic index, so as to improve the reliability of the power system and the calculation accuracy of the power purchase cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0072] Figure 1 A flowchart of a power system production simulation method based on power market transactions provided by an embodiment of the present invention;

[0073] Figure 2 A flow chart of classifying historical price curves and determining predicted price curves provided by an embodiment of the present invention;

[0074] Figure 3 A flow chart for classifying historical load curves and determining predicted load curves provided by an embodiment of the present invention;

[0075] Figure 4 A flow chart for determining the electricity purchase cost and the electricity purchase reliability index provided by an embodiment of the present invention;

[0076] Figure 5 A structural block diagram of a power system production simulation device based on power market transactions provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0077] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0078] In this application, the terms "comprises", "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element.

[0079] As can be seen from the background technology, decomposing the medium- and long-term contract electricity into daily and hourly granularity will lead to deviations between the real-time operation of the power system and the dispatch plan. To solve this deviation, the existing power system introduces the spot market to ensure real-time power balance. However, the introduction of the spot market will increase the randomness and volatility of costs, which will lead to lower reliability of the calculated power system and lower accuracy of the power purchase cost.

[0080] Therefore, an embodiment of the present invention provides a method and device for simulating power system production based on power market transactions, which determines a predicted price curve and a predicted load curve by introducing data from power market transactions; and based on the predicted price curve, the predicted load curve and the unit characteristic indicators, determines the amount of electricity to be purchased, the power purchase cost and the power purchase reliability index, so as to improve the reliability of the power system and the calculation accuracy of the power purchase cost.

[0081] See also Figure 1 , shows a flow chart of a power system production simulation method based on power market transactions provided by an embodiment of the present invention, the power system production simulation method comprising:

[0082] Step S101: Acquire the historical price curve of the electricity seller user, and acquire the historical load curve of the electricity buyer user.

[0083] In the process of implementing step S101, the historical price curve of the electricity seller (that is, the seller user) is obtained. Specifically, the historical price curves of the electricity seller in previous years are obtained. The time granularity of each historical price curve is the day granularity, and the total time scale of each historical price curve is one year; that is, the historical price curve contains the historical prices of each day in a certain year (the historical prices of electricity sales).

[0084] Obtain the historical load curves of electricity purchasers (i.e., buyers, such as power grid companies). The time granularity of each historical load curve is hourly, and the total time scale of each historical load curve is one day; that is, the historical load curve contains the historical load of each hour in a certain day.

[0085] Step S102: pre-process the historical price curve and the historical load curve respectively.

[0086] In the specific implementation of step S102, the highest historical price in the historical price curve (that is, the highest historical price of the year) is obtained, and the highest historical load in the historical load curve (that is, the highest historical load of the day) is obtained.

[0087] For each historical price curve, the historical prices in the historical price curve are normalized according to the highest historical price of the historical price curve. Specifically, the historical prices in the historical price curve are normalized with the highest historical price as the base value.

[0088] For each historical load curve, the historical loads in the historical load curve are normalized according to the highest historical load of the historical load curve. Specifically, the historical loads in the historical load curve are normalized with the highest historical load as the base value.

[0089] In some specific embodiments, the historical price in the historical price curve and the historical load in the historical load curve are normalized by formula (1).

[0090]

[0091] In formula (1), when the historical prices in the historical price curve are normalized, P i * is the normalized historical price, P i is the unnormalized historical price, P imax is the highest historical price; when normalizing the historical load in the historical load curve, P i * is the normalized historical load, P i is the historical load without normalization, P imax The highest historical load.

[0092] According to a preset autoregressive moving average model, data processing is performed on the historical prices in the normalized historical price curve, and data processing is performed on the historical loads in the normalized historical load curve. The data processing includes one or more of deleting bad data, restoring missing data, and repairing abnormal data.

[0093] That is to say, by utilizing the autocorrelation characteristics and autoregressive sliding average, etc., the integrity of the historical prices in the normalized historical price curve (that is, the data integrity) and the integrity of the historical loads in the normalized historical load curve are identified; it is analyzed whether there are data omissions and bad data that do not conform to the autocorrelation characteristics in the normalized historical price curve and historical load curve; the bad data in the normalized historical price curve and historical load curve are deleted, and the missing data in the normalized historical price curve and historical load curve are restored.

[0094] It is understandable that there may be abnormal data in the normalized historical price curve and historical load curve. For example, under normal circumstances, the historical price changes in the normalized historical price curve are relatively stable. If the difference between the historical prices before and after a certain period of time is large, it indicates that there is a sudden change in the historical price. At this time, it can be preliminarily judged that there is abnormal data in the historical price curve.

[0095] For abnormal data in the normalized historical price curve and historical load curve, it is necessary to repair the abnormal data. For details on how to repair the abnormal data, please see the following content.

[0096] The way to repair abnormal data is:

[0097] An autoregressive moving average model is constructed in advance, and the expression of the constructed autoregressive moving average model is as shown in formula (2).

[0098]

[0099] In formula (2), {x t}(t=1,2,…) is the normalized historical price curve or historical load curve, p is the order of the autoregressive part (AR), q is the order of the moving average part (MA), is the autoregressive coefficient, θ i is the sliding average coefficient, a t The mean is 0 and the variance is σ a 2 Normally distributed white noise.

[0100] The AIC(p,q) function is used to determine the order of the autoregressive sliding model ARMA(p,q). The specific content of the AIC(p,q) function is shown in formula (3).

[0101]

[0102] In formula (3), N is {x t}The number of data in is the estimate of the variance of the noise in the autoregressive sliding model.

[0103] It can be understood that the autoregressive sliding model ARMA (p, q) is composed of a p-order autoregressive part and a q-order moving average part. The AIC (p, q) function given by the above formula (3) calculates the values ​​of p and q in the autoregressive sliding model ARMA (p, q), and the smallest value (p, q) is selected as the optimal order value, that is, the optimal autoregressive sliding model ARMA (p, q) is determined based on the smallest value (p, q).

[0104] The best autoregressive sliding model ARMA (p, q) determined is used to repair the abnormal data in the normalized historical price curve; similarly, the best autoregressive sliding model ARMA (p, q) determined is used to repair the abnormal data in the normalized historical load curve.

[0105] The specific repair method is: take the repair of abnormal data in the historical price curve as an example, assuming that the time corresponding to the abnormal data is t, the historical price before time t can be obtained through the historical price curve, and the historical price before time t and the best ARMA (p, q) are used to predict the historical price at time t, and the predicted price is used to replace the abnormal data corresponding to time t. The repair method for abnormal data of the historical load curve is similar, so it will not be repeated here.

[0106] Through the content in the above step S102, the preprocessing of the historical price curve and the historical load curve is completed.

[0107] Step S103: classify the preprocessed historical price curves and determine the predicted price curve according to the classification results, and classify the preprocessed historical load curves and determine the predicted load curve according to the classification results.

[0108] In the specific implementation of step S103, the pre-processed historical price curve is classified, and the predicted price curve is determined according to the classification result. The pre-processed historical load curve is classified, and the predicted load curve is determined according to the classification result.

[0109] The price in the predicted price curve is the predicted future daily electricity sales price (that is, the cost of spot transactions in the electricity market), and the load in the predicted load curve is the predicted future hourly load.

[0110] Step S104: sorting the power seller users by priority according to the power selling prices of the power seller users and the risk indicators of the power generation system.

[0111] In the specific implementation of step S104, the electricity selling price and power generation system risk index that can be provided by the electricity seller are obtained, and weights corresponding to the electricity selling price and risk are assigned respectively, the price weight is set to a, and the risk weight is set to b, where b=1-a.

[0112] Based on the electricity selling prices and power generation system risk indicators of various electricity sellers, as well as the set price weights and risk weights, the objective function min{aP+bR} is used to determine the electricity seller with the smallest total price and risk, and the priority of the electricity seller with the smallest total price and risk is set to the highest. The priorities of other electricity sellers decrease step by step in the order of price and risk sum from large to small. Among them, in the objective function min{aP+bR}, P is the electricity selling price, and R is the power generation system risk indicator.

[0113] Step S105: Determine the power shortage based on the predicted load curve and the pre-acquired medium- and long-term contracts, and purchase the power corresponding to the power shortage from the power seller user according to the priority.

[0114] In the specific implementation of step S105, after removing the power in the medium- and long-term contracts from the forecast load curve, the power shortage can be determined, and the power shortage indicates the power shortage of the power purchaser. The power corresponding to the power shortage is purchased from the power seller according to the priority of each power seller and the amount of power that can be provided.

[0115] For example: assuming the electricity shortage is 10MW·h, and the first priority (highest priority) electricity seller can only provide 5MW·h, then 5MW·h will be purchased from the highest priority electricity seller first, and the remaining electricity will be purchased from the next priority electricity sellers (such as the second and third priority) until the purchase is completed.

[0116] Step S106: Determine the power purchase cost and power purchase reliability index based on the medium- and long-term contracts, the predicted price curve, the predicted load curve and the unit characteristic index of the power seller user.

[0117] In the specific implementation of step S106, after purchasing the electricity corresponding to the electricity shortage from the electricity seller, the electricity purchase cost and electricity purchase reliability index are determined based on the medium- and long-term contracts, the predicted price curve, the predicted load curve and the unit characteristic index of the electricity seller.

[0118] In an embodiment of the present invention, a forecast price curve and a forecast load curve are determined by introducing data from electricity market transactions; and based on the forecast price curve, the forecast load curve and the unit characteristic index, the amount of electricity to be purchased, the electricity purchase cost and the electricity purchase reliability index are determined to improve the reliability of the power system and the calculation accuracy of the electricity purchase cost.

[0119] The above embodiments of the present invention Figure 1 The process of classifying the historical price curves and determining the predicted price curve involved in step S103 is described in detail in Figure 2 , shows a flow chart of classifying historical price curves and determining predicted price curves provided by an embodiment of the present invention, including the following steps:

[0120] Step S201: extracting a preset price index from the preprocessed historical price curve.

[0121] In the process of implementing step S201, preset price indicators are extracted from the preprocessed historical price curve. The preset price indicators include but are not limited to: annual price peak-to-valley difference, ratio of annual average price to annual maximum price, annual maximum price distribution, probability distribution of annual maximum price occurrence time, etc.

[0122] Step S202: Based on the extracted preset price index and in combination with the first sample data, a first neural network is constructed and trained to obtain a first clustering model.

[0123] In the specific implementation of step S202, a first neural network (eg, a SOM neural network) is constructed according to the extracted preset price index, and then the first sample data is used to train the first neural network to obtain a first clustering model.

[0124] The specific process of training the first neural network using the first sample data is:

[0125] The annual price curve of each electricity seller user is obtained from the sample data set and used as the first sample data, and the obtained first sample data is input into the first neural network.

[0126] The corresponding network structure is determined according to the first sample data input into the first neural network. It can be understood that the first sample data input into the first neural network is a vector, and the first sample data is recorded as X i =[x1,x2,...,x m ], where i is the number of the first sample data input, m is the dimension of the first sample data input, and X i The values ​​of each dimension of the first sample data of the input.

[0127] Calculate PCA extraction components to initialize network weights. The weight vector of each neuron is denoted as W i =[w1,w2,...,w n [, where i is the neuron number, n is the dimension of the weight vector, W i is the value of each dimension in the neuron’s weight vector.

[0128] Determine the number of iterations to train the first neural network.

[0129] Select the first sample data X that is input into the first neural network in sequence i , using Euclidean distance to calculate X i The distance between the weight vector of each neuron can be calculated by formula (4): i .

[0130]

[0131] Determine the weight vector of the nearest neuron as W min ; W min The neuron represented by is taken as the winning neuron, and the weight vector of the winning neuron and the neurons in its neighborhood is updated. Specifically, the weight vector is updated by formula (5).

[0132]

[0133] In formula (5), w i (t) is the value of the i-th dimension of the neuron weight vector at the t-th iteration, w i (t+1) is the value of the i-th dimension of the neuron weight vector at the t+1th iteration, ε(t) is the learning rate, which is adaptively adjusted as the number of iterations increases, and x i (t) is the value of the i-th dimension of the first sample data input at the t-th iteration.

[0134] Through the above method, the first neural network is repeatedly trained until the weight reaches the error requirement or satisfies the number of iterations, and the training of the first neural network is completed to obtain the first clustering model.

[0135] Step S203: input the preprocessed historical price curve into the first clustering model for clustering to obtain multiple price patterns.

[0136] In the specific implementation of step S203, the preprocessed historical price curve is input into the first clustering model for clustering to obtain a clustering result, which is a plurality of price patterns, and each price pattern contains at least one preprocessed historical price curve.

[0137] Step S204: For each type of price pattern, average the historical prices of the historical price curve in the price pattern to obtain the historical price average value.

[0138] In the specific implementation of step S204, for each type of price pattern, the historical prices of the historical price curve included in this type of price pattern are averaged to obtain the historical price average corresponding to this type of price pattern.

[0139] That is to say, the historical prices of the historical price curves belonging to the same type of price pattern are averaged to obtain the historical price average corresponding to each type of price pattern.

[0140] It can be understood that the historical price average corresponding to each type of price pattern is the cluster center, and the historical price average can also be called the typical annual price.

[0141] Step S205: Determine the first weight of each type of price pattern according to the number of historical price curves in each type of price pattern.

[0142] In the process of implementing step S205, the first weight of each type of price pattern is determined according to the number of historical price curves in each type of price pattern; specifically, the ratio of the number of historical price curves in each type of price pattern to the total number of all historical price curves is calculated, and the higher the ratio, the higher the first weight corresponding to the price pattern.

[0143] For example: Assume that after clustering all historical price curves, we get the first type of price pattern and the second type of price pattern, where the first type of price pattern contains 3 similar historical price curves and the second type of price pattern contains 5 similar historical price curves; calculate the ratio of the number of historical price curves in each type of price pattern to the total number of all historical price curves. The more historical price curves a price pattern contains, the greater its first weight is, that is, the first weight of the second type of price pattern is greater than the first weight of the first type of price pattern.

[0144] Step S206: For each type of price pattern, based on the historical price average corresponding to the price pattern and the deviation distribution of the historical prices of the historical price curve, simulate and obtain the price curve corresponding to the price pattern.

[0145] In the process of implementing step S206, for each type of price pattern, based on the historical price average corresponding to the price pattern and the historical price curve (normalized curve) in the price pattern, the Monte Carlo simulation method is used to simulate the price curve corresponding to the price pattern.

[0146] It can be understood that, through the above method, a corresponding price curve is simulated for each type of price model, that is, a simulation curve is generated for each type of price model.

[0147] Step S207: Based on the first weight of each type of price pattern, the price curves corresponding to all simulated price patterns are weighted averaged, and the highest historical price in the historical price curve is used to process the weighted average result to obtain a predicted price curve.

[0148] In the process of implementing step S207, after simulating the price curves corresponding to various price modes, the price curves corresponding to all the simulated price modes are weighted averaged according to the first weight of each price mode, and multiple simulated price curves are weighted averaged into a normalized price curve. The normalized price curve obtained by weighted average is the predicted price curve.

[0149] It should be noted that the historical price curve used in the above steps is a normalized curve, so the predicted price curve obtained by weighted averaging of multiple price curves is also a normalized curve; in order to restore the normalized predicted price curve to the actual predicted price curve, it is necessary to use the highest historical price in the historical price curve to transform the normalized predicted price curve (that is, the result of weighted averaging) and restore the normalized predicted price curve to the actual predicted price curve.

[0150] In an embodiment of the present invention, the historical price curves are clustered by using a neural network clustering method, and a predicted price curve is predicted based on the clustering results. In this way, the inherent change law of the historical price curve can be effectively excavated to ensure that the predicted price curve is representative and targeted. Furthermore, when the amount of electricity to be purchased, the electricity purchase cost and the electricity purchase reliability index are determined based on the predicted price curve, the predicted load curve and the unit characteristic index, the reliability of the power system and the calculation accuracy of the electricity purchase cost can be improved.

[0151] The above embodiments of the present invention Figure 1 The process of classifying the historical load curves and determining the predicted load curve involved in step S103 is described in detail in Figure 3 , shows a flow chart of classifying historical load curves and determining predicted load curves provided by an embodiment of the present invention, including the following steps:

[0152] Step S301: extracting a preset load index from the preprocessed historical load curve.

[0153] Step S302: Based on the extracted preset load index and in combination with the second sample data, a second neural network is constructed and trained to obtain a second clustering model.

[0154] In the specific implementation of step S302, a second neural network (such as a SOM neural network) is constructed according to the extracted preset load index, and then the second sample data is used to train the second neural network to obtain a second clustering model. For the specific process of training the second neural network to obtain the second clustering model, please refer to the above-mentioned embodiment of the present invention. Figure 2 The content of training the first neural network in step S302 is not repeated here.

[0155] Step S303: input the preprocessed historical load curve into the second clustering model for clustering to obtain multiple types of load patterns.

[0156] It should be noted that each type of load mode contains at least one preprocessed historical load curve.

[0157] Step S304: for each type of load pattern, average the historical loads of the historical load curves in the load pattern to obtain the historical load average value.

[0158] Step S305: Determine the second weight of each type of load mode according to the number of historical load curves in each type of load mode.

[0159] In the specific implementation of step S305, the ratio of the number of historical load curves in each type of load mode to the total number of all historical load curves is calculated, and the load mode with a higher ratio has a higher corresponding second weight. Figure 2 The content of step S205 is not repeated here.

[0160] Step S306: For each type of load pattern, based on the historical load average value corresponding to the load pattern and the deviation distribution of the historical load of the historical load curve, simulate and obtain the load curve corresponding to the load pattern.

[0161] In the process of implementing step S306, for each type of load pattern, based on the historical load average value corresponding to the load pattern and the historical load curve (normalized curve) of the load pattern, the Monte Carlo simulation method is used to simulate the load curve corresponding to the load pattern.

[0162] It can be understood that, through the above method, a corresponding load curve is simulated for each type of load mode, that is, a simulation curve is generated for each type of load mode.

[0163] Step S307: Based on the second weight of each type of load mode, the load curves corresponding to all simulated load modes are weighted averaged, and the highest historical load in the historical load curves is used to process the weighted average result to obtain a predicted load curve.

[0164] In the specific implementation of step S307, after simulating and obtaining the load curves corresponding to various load modes, the load curves corresponding to all simulated load modes are weighted averaged according to the second weight of each load mode, and multiple simulated load curves are weighted averaged into a normalized load curve, and the normalized load curve obtained by weighted average is the predicted load curve. Then, the highest historical load in the historical load curve is used to process the normalized predicted load curve (that is, the result of weighted average), and the normalized predicted load curve is restored to the actual predicted load curve.

[0165] It should be noted that the execution principle of step S301 to step S307 is the same as that of the above-mentioned embodiment of the present invention. Figure 2 The execution principles of each step are similar, and the details can be found in the above-mentioned embodiment of the present invention. Figure 2 The contents in will not be repeated here.

[0166] In an embodiment of the present invention, the historical load curves are clustered by using a neural network clustering method, and a predicted load curve is predicted based on the clustering results. In this way, the inherent variation law of the historical load curve can be effectively excavated to ensure that the predicted load curve is representative and targeted. Furthermore, when the amount of electricity to be purchased, the electricity purchase cost and the electricity purchase reliability index are determined based on the predicted price curve, the predicted load curve and the unit characteristic index, the reliability of the power system and the calculation accuracy of the electricity purchase cost can be improved.

[0167] The above-mentioned invention is implemented Figure 1 The process of determining the power purchase cost and the power purchase reliability index involved in step S106 is shown in Figure 4 , shows a flow chart of determining the power purchase cost and the power purchase reliability index provided by an embodiment of the present invention, including the following steps:

[0168] Step S401: Determine the electricity purchase cost based on the medium- and long-term contracts and the predicted price curve.

[0169] It should be noted that the medium- and long-term contract is a contract signed between the electricity seller and the electricity buyer, and the contract electricity purchase fee is agreed upon in the medium- and long-term contract; the predicted price curve includes the predicted daily electricity sales price.

[0170] In the specific implementation of step S401, the overall electricity purchase cost of the electricity purchaser is calculated using the contractual electricity purchase fee of the medium- and long-term contract and the predicted electricity sales price included in the predicted price curve.

[0171] Step S402: The unit operating status is simulated using the unit characteristic indicators of the power seller user.

[0172] In the specific implementation of step S402, the forced outage rate of the units of the power seller user is used to simulate the operation status of the units of the power seller user by using the Monte Carlo simulation method.

[0173] Specifically, for the electricity seller, the failure rate of each unit in operation is randomly sampled every hour, and the failure rate is compared with the forced shutdown rate of the unit. The operation status of the unit of the electricity seller can be determined by comparing the results. If the failure rate is greater than the forced shutdown rate, it means that the unit is faulty (cannot output normally), and if the failure rate is less than or equal to the forced shutdown rate, it means that the unit is operating normally.

[0174] The operating status of the power selling user's unit is shown in formula (6).

[0175]

[0176] In formula (6), R j is the failure rate of operating unit j, PF j is the forced outage rate of unit j, A j is the operating status of unit j, A j 0 indicates a fault, A j A value of 1 indicates normal operation.

[0177] Step S403: Based on the available capacity of the units put into operation by the power seller user every hour, the total available capacity per hour is calculated.

[0178] In the specific implementation of step S403, the available capacity of the units put into operation by the power seller user every hour is determined, and the available capacity of all units (for example, n units) put into operation every hour is summed up to obtain the total available capacity per hour.

[0179] It can be understood that the available capacity of the units put into operation every hour is as shown in formula (7).

[0180] P j =A j *S 额定 (7)

[0181] In formula (7), P j is the available capacity of unit j put into operation every hour, A j is the operating status of unit j, S 额定 is the rated power of unit j put into operation every hour.

[0182] Combined with formula (7), formula (8) is used to calculate the total available capacity per hour.

[0183]

[0184] In formula (8), S i is the total available capacity of the units of the electricity seller user in the i-th hour, and n is the total number of units put into operation every hour.

[0185] Step S404: sum the electricity of the medium- and long-term contracts and the total available capacity per hour to obtain the actual electricity purchase amount per hour.

[0186] In the specific implementation of step S404, taking into account the unit failure of the power seller, when calculating the actual power purchase amount per hour, the method of "power under medium- and long-term contracts + total available capacity per hour" is used to calculate the actual power purchase amount per hour.

[0187] Step S405: Calculate the difference between the actual power purchase amount per hour and the predicted load of the predicted load curve to obtain the power shortage per hour.

[0188] In the specific implementation process of step S405, the difference between the actual power purchase amount per hour and the predicted load of the predicted load curve is calculated to obtain the hourly power shortage within a certain time period (that is, the power shortage every hour), and the number of hours with power shortage in the time period is determined (that is, how many hours there is power shortage).

[0189] Step S406: Calculate the average power shortage and power shortage probability according to the power shortage and power shortage hours per hour.

[0190] In the specific implementation of step S406, according to the hourly power shortage in a certain time period, the average power shortage ENs in the time period is calculated in combination with formula (9): k .

[0191]

[0192] In formula (9), Ens t is the power shortage in the tth hour, T is the total number of hours in the time period, t represents the tth hour in the time period, and the value of t ranges from 1 to T.

[0193] According to the number of hours of power shortage in a certain time period, the power shortage probability Lolp in the time period is calculated by combining formula (10): k .

[0194]

[0195] In formula (10), h is the number of hours when there is a power shortage, and T is the total number of hours in the time period.

[0196] Step S407: Return to step S402 and execute until it is executed K times, to obtain K average power shortages and K power shortage probabilities.

[0197] In the process of implementing step S407, return to execute step S401 until it is executed K times, that is, repeat step S402 to step S406 K times. Each time step S402 to step S406 is executed, an average power shortage and a power shortage probability can be obtained. After executing step S402 to step S406 K times, K average power shortages and K power shortage probabilities are obtained.

[0198] Step S408: Calculate the average value of K average power shortages to obtain the power shortage expectation value, and calculate the average value of K power shortage probabilities to obtain the power shortage probability.

[0199] In the specific implementation of step S408, formula (11) is used to calculate the average value of K average power shortages to obtain the expected power shortage value EENS.

[0200]

[0201] In formula (11), ENs k is the kth average power shortage.

[0202] Using formula (12), the average value of K power shortage probabilities is calculated to obtain the power shortage probability LOLP.

[0203]

[0204] In formula (12), Lolp k is the kth power outage probability.

[0205] It can be understood that the specific value of K should be based on ensuring that the calculated expected value of power shortage and the probability of power shortage can converge. In the embodiment of the present invention, the specific value of K is not limited.

[0206] In the embodiment of the present invention, the power purchase cost is determined by using medium- and long-term contracts and forecast price curves, which can ensure the accuracy of the calculation of the power purchase cost. Based on the Monte Carlo simulation method, combined with the failure rate of each unit in operation of the power seller user, the unit operation status of the power seller user is determined and the power purchase reliability indicators such as the expected value of power shortage and the probability of power shortage are calculated accordingly, which ensures the accuracy of the calculation of the reliability of the power system.

[0207] Corresponding to the power system production simulation method based on power market transaction provided by the above embodiment of the present invention, see Figure 5, the embodiment of the present invention also provides a structural block diagram of a power system production simulation device based on power market transactions, the power system production simulation device includes: an acquisition unit 501, a preprocessing unit 502, a prediction unit 503, a sorting unit 504, a processing unit 505 and a determination unit 506;

[0208] The acquisition unit 501 is used to acquire the historical price curve of the electricity seller user and the historical load curve of the electricity buyer user.

[0209] In the specific implementation, the time granularity of the historical price curve is day granularity, and the total time scale of the historical price curve is one year; the time granularity of the historical load curve is hour granularity, and the total time scale of the historical load curve is one day.

[0210] The preprocessing unit 502 is used to preprocess the historical price curve and the historical load curve respectively.

[0211] The prediction unit 503 is used to classify the preprocessed historical price curves and determine the predicted price curve according to the classification results, and to classify the preprocessed historical load curves and determine the predicted load curve according to the classification results.

[0212] The sorting unit 504 is used to sort the power seller users by priority according to the power selling prices of the power seller users and the risk indicators of the power generation system.

[0213] The processing unit 505 is used to determine the power shortage according to the predicted load curve and the pre-acquired medium- and long-term contracts, and purchase the power corresponding to the power shortage from the power seller user according to the priority.

[0214] The determination unit 506 is used to determine the power purchase cost and the power purchase reliability index according to the medium- and long-term contracts, the predicted price curve, the predicted load curve and the unit characteristic index of the power seller user.

[0215] In an embodiment of the present invention, a forecast price curve and a forecast load curve are determined by introducing data from electricity market transactions; and based on the forecast price curve, the forecast load curve and the unit characteristic index, the amount of electricity to be purchased, the electricity purchase cost and the electricity purchase reliability index are determined to improve the reliability of the power system and the calculation accuracy of the electricity purchase cost.

[0216] Preferably, combined Figure 5 The content shown, the pre-processing unit 502 includes: an acquisition module, a normalization module and a processing module, and the execution principle of each module is as follows:

[0217] The acquisition module is used to obtain the highest historical price in the historical price curve and the highest historical load in the historical load curve.

[0218] The normalization module is used to normalize the historical prices in the historical price curve according to the highest historical price, and to normalize the historical loads in the historical load curve according to the highest historical load.

[0219] A processing module is used to perform data processing on the historical prices in the normalized historical price curve and on the historical loads in the normalized historical load curve according to a preset autoregressive sliding average model, wherein the data processing includes one or more of deleting bad data, restoring missing data and repairing abnormal data.

[0220] Preferably, combined Figure 5 The content shown is used for classifying the preprocessed historical price curve and determining the prediction unit 503 of the predicted price curve according to the classification result, including: a first extraction module, a first construction module, a first clustering module, a first calculation module, a first determination module, a first prediction module and a first weighting module. The execution principle of each module is as follows:

[0221] The first extraction module is used to extract a preset price index from the preprocessed historical price curve.

[0222] The first construction module is used to construct and train a first neural network to obtain a first clustering model based on the extracted preset price index and combined with the first sample data.

[0223] The first clustering module is used to input the preprocessed historical price curve into the first clustering model for clustering to obtain multiple price patterns, each of which contains at least one preprocessed historical price curve.

[0224] The first calculation module is used for calculating the average of the historical prices of the historical price curve in each price mode to obtain the average historical price.

[0225] The first determination module is used to determine the first weight of each type of price pattern according to the number of historical price curves in each type of price pattern.

[0226] The first prediction module is used to simulate and obtain a price curve corresponding to each type of price pattern based on the historical price average corresponding to the price pattern and the deviation distribution of the historical prices of the historical price curve.

[0227] The first weighting module is used to perform weighted average of the price curves corresponding to all simulated price patterns based on the first weight of each type of price pattern, and process the weighted average result using the highest historical price in the historical price curve to obtain a predicted price curve.

[0228] In an embodiment of the present invention, the historical price curves are clustered by using a neural network clustering method, and a predicted price curve is predicted based on the clustering results. In this way, the inherent change law of the historical price curve can be effectively excavated to ensure that the predicted price curve is representative and targeted. Furthermore, when the amount of electricity to be purchased, the electricity purchase cost and the electricity purchase reliability index are determined based on the predicted price curve, the predicted load curve and the unit characteristic index, the reliability of the power system and the calculation accuracy of the electricity purchase cost can be improved.

[0229] Preferably, combined Figure 5 The content shown is used for classifying the preprocessed historical load curve and determining the prediction unit 503 of the predicted load curve according to the classification result, including: a second extraction module, a second construction module, a second clustering module, a second calculation module, a second determination module, a second prediction module and a second weighting module. The execution principle of each module is as follows:

[0230] The second extraction module is used to extract the preset load index from the preprocessed historical load curve.

[0231] The second construction module is used to construct and train a second neural network to obtain a second clustering model based on the extracted preset load index and in combination with the second sample data.

[0232] The second clustering module is used to input the preprocessed historical load curve into the second clustering model for clustering to obtain multiple types of load patterns, each type of load pattern contains at least one preprocessed historical load curve.

[0233] The second calculation module is used to calculate the average value of the historical load of the historical load curve in the load mode for each type of load mode to obtain the historical load average value.

[0234] The second determination module is used to determine the second weight of each type of load mode according to the number of historical load curves in each type of load mode.

[0235] The second prediction module is used to simulate and obtain the load curve corresponding to each type of load pattern based on the historical load average value corresponding to the load pattern and the deviation distribution of the historical load of the historical load curve.

[0236] The second weighting module is used to perform weighted averaging of the load curves corresponding to all simulated load modes based on the second weight of each type of load mode, and process the weighted average result using the highest historical load in the historical load curves to obtain a predicted load curve.

[0237] In an embodiment of the present invention, the historical load curves are clustered by using a neural network clustering method, and a predicted load curve is predicted based on the clustering results. In this way, the inherent variation law of the historical load curve can be effectively excavated to ensure that the predicted load curve is representative and targeted. Furthermore, when the amount of electricity to be purchased, the electricity purchase cost and the electricity purchase reliability index are determined based on the predicted price curve, the predicted load curve and the unit characteristic index, the reliability of the power system and the calculation accuracy of the electricity purchase cost can be improved.

[0238] Preferably, combined Figure 5 The content shown, the determination unit 506 includes: a determination module, a simulation module, a first calculation module, a second calculation module, a third calculation module, a fourth calculation module, an execution module and a fifth calculation module, and the execution principle of each module is as follows:

[0239] The determination module is used to determine the electricity purchase cost based on the medium- and long-term contracts and the predicted price curve.

[0240] The simulation module is used to simulate the unit operating status using the unit characteristic indicators of the power seller user.

[0241] The first calculation module is used to calculate the total available capacity per hour based on the available capacity of the units put into operation by the power seller user every hour.

[0242] The second calculation module is used to sum the electricity of the medium- and long-term contracts and the total available capacity per hour to obtain the actual electricity purchase amount per hour.

[0243] The third calculation module is used to calculate the difference between the actual power purchase amount per hour and the predicted load of the predicted load curve to obtain the power shortage per hour.

[0244] The fourth calculation module is used to calculate the average power shortage and the power shortage probability according to the hourly power shortage and the number of power shortage hours.

[0245] An execution module is used to return to the execution simulation module until it is executed K times, and obtain K average power shortages and K power shortage probabilities;

[0246] The fifth calculation module is used to calculate the average value of K average power shortages to obtain the expected power shortage value, and calculate the average value of K power shortage probabilities to obtain the power shortage probability.

[0247] In the embodiment of the present invention, the power purchase cost is determined by using medium- and long-term contracts and forecast price curves, which can ensure the accuracy of the calculation of the power purchase cost. Based on the Monte Carlo simulation method, combined with the failure rate of each unit in operation of the power seller user, the unit operation status of the power seller user is determined and the power purchase reliability indicators such as the expected value of power shortage and the probability of power shortage are calculated accordingly, which ensures the accuracy of the calculation of the reliability of the power system.

[0248] In summary, the embodiments of the present invention provide a method and device for simulating power system production based on power market transactions, which determines a predicted price curve and a predicted load curve by introducing data from power market transactions; and based on the predicted price curve, the predicted load curve and the unit characteristic indicators, determines the amount of electricity to be purchased, the power purchase cost and the power purchase reliability index, thereby improving the reliability of the power system and the calculation accuracy of the power purchase cost.

[0249] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can refer to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiment. The system and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without creative work.

[0250] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0251] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for simulating power system production based on power market transactions, characterized in that: The method comprises: Obtaining the historical price curve of electricity sellers and the historical load curve of electricity buyers; Preprocessing the historical price curve and the historical load curve respectively; Classifying the preprocessed historical price curve and determining a predicted price curve according to the classification result, and classifying the preprocessed historical load curve and determining a predicted load curve according to the classification result; According to the electricity selling price of the electricity selling user and the risk index of the power generation system, the electricity selling user is sorted according to the priority; Determine the power shortage according to the predicted load curve and the pre-acquired medium- and long-term contracts, and purchase the power corresponding to the power shortage from the power seller user according to the priority; Determine the power purchase cost and power purchase reliability index according to the medium- and long-term contract, the predicted price curve, the predicted load curve and the unit characteristic index of the power seller user; According to the medium- and long-term contract, the predicted price curve, the predicted load curve and the unit characteristic index of the power seller user, the power purchase cost and the power purchase reliability index are determined, including: Determine the electricity purchase cost based on the medium- and long-term contracts and the predicted price curve; The unit operating status is obtained by simulating the unit characteristic index of the electricity seller user; Based on the available capacity of the units put into operation every hour by the power seller user, the total available capacity per hour is calculated; The electricity of the medium- and long-term contracts and the total available capacity per hour are summed to obtain the actual electricity purchase per hour; Calculate the difference between the actual power purchase amount per hour and the predicted load of the predicted load curve to obtain the power shortage per hour; According to the hourly power shortage and power shortage hours, the average power shortage and power shortage probability are calculated; Return to the step of simulating the operation status of the unit by using the unit characteristic index of the power seller user until the step is performed K times, and K average power shortages and K power shortage probabilities are obtained; The average value of the K average power shortages is calculated to obtain the expected power shortage value, and the average value of the K power shortage probabilities is calculated to obtain the power shortage probability.

2. The method according to claim 1, characterized in that Preprocessing the historical price curve and the historical load curve respectively includes: Obtaining the highest historical price in the historical price curve, and obtaining the highest historical load in the historical load curve; Normalizing the historical prices in the historical price curve according to the highest historical price, and normalizing the historical loads in the historical load curve according to the highest historical load; According to a preset autoregressive moving average model, data processing is performed on the historical prices in the normalized historical price curve, and data processing is performed on the historical loads in the normalized historical load curve, wherein the data processing includes one or more of deleting bad data, restoring missing data, and repairing abnormal data.

3. The method according to claim 1, characterized in that Classifying the preprocessed historical price curve and determining a predicted price curve according to the classification result, including: Extracting a preset price index from the preprocessed historical price curve; According to the extracted preset price index, combined with the first sample data, a first neural network is constructed and trained to obtain a first clustering model; Inputting the preprocessed historical price curve into the first clustering model for clustering to obtain multiple price patterns, each price pattern including at least one preprocessed historical price curve; For each type of the price pattern, averaging the historical prices of the historical price curve in the price pattern to obtain a historical price average; Determining a first weight of each type of the price pattern according to the number of the historical price curves in each type of the price pattern; For each type of the price pattern, based on the historical price average corresponding to the price pattern and the deviation distribution of the historical prices of the historical price curve, a price curve corresponding to the price pattern is simulated; Based on the first weight of each type of price pattern, the price curves corresponding to all the simulated price patterns are weighted averaged, and the highest historical price in the historical price curves is used to process the weighted average result to obtain a predicted price curve.

4. The method according to claim 1, characterized in that: Classifying the preprocessed historical load curve and determining a predicted load curve according to the classification result, including: Extracting a preset load index from the preprocessed historical load curve; According to the extracted preset load index, combined with the second sample data, a second neural network is constructed and trained to obtain a second clustering model; Inputting the preprocessed historical load curve into the second clustering model for clustering to obtain multiple types of load patterns, each type of load pattern includes at least one preprocessed historical load curve; For each type of the load mode, averaging the historical loads of the historical load curves in the load mode to obtain a historical load average value; determining a second weight for each type of the load pattern according to the number of the historical load curves in each type of the load pattern; For each type of the load pattern, based on the historical load average value corresponding to the load pattern and the deviation distribution of the historical load of the historical load curve, a load curve corresponding to the load pattern is simulated; Based on the second weight of each type of the load mode, the load curves corresponding to all the simulated load modes are weighted averaged, and the highest historical load in the historical load curves is used to process the weighted average result to obtain a predicted load curve.

5. The method according to any one of claims 1 to 4, characterized in that: The time granularity of the historical price curve is day granularity, and the total time scale of the historical price curve is one year; the time granularity of the historical load curve is hour granularity, and the total time scale of the historical load curve is one day.

6. A power system production simulation device based on power market transactions, characterized in that: The device comprises: An acquisition unit, used to acquire a historical price curve of a power seller user and a historical load curve of a power buyer user; A preprocessing unit, used for preprocessing the historical price curve and the historical load curve respectively; A prediction unit, configured to classify the preprocessed historical price curves and determine a predicted price curve according to the classification results, and to classify the preprocessed historical load curves and determine a predicted load curve according to the classification results; A sorting unit, used to sort the power seller users by priority according to the power selling price of the power seller users and the risk index of the power generation system; A processing unit, configured to determine the power shortage according to the predicted load curve and the pre-acquired medium- and long-term contracts, and purchase the power corresponding to the power shortage from the power seller user according to the priority; A determination unit, configured to determine the power purchase cost and the power purchase reliability index according to the medium- and long-term contract, the predicted price curve, the predicted load curve and the unit characteristic index of the power seller user; The determining unit comprises: A determination module, used to determine the electricity purchase cost according to the medium- and long-term contract and the predicted price curve; A simulation module, used to simulate the unit operating status using the unit characteristic index of the power seller user; A first calculation module is used to calculate the total available capacity per hour based on the available capacity of the units put into operation by the power seller user per hour; The second calculation module is used to sum the power of the medium- and long-term contracts and the total available capacity per hour to obtain the actual power purchase per hour; A third calculation module is used to calculate the difference between the actual power purchase amount per hour and the predicted load of the predicted load curve to obtain the power shortage per hour; A fourth calculation module is used to calculate the average power shortage and the power shortage probability according to the hourly power shortage and the number of power shortage hours; An execution module, used for returning to execute the simulation module until it is executed K times, and obtaining K average power shortages and K power shortage probabilities; The fifth calculation module is used to calculate the average value of the K average power shortages to obtain the expected power shortage value, and calculate the average value of the K power shortage probabilities to obtain the power shortage probability.

7. The device according to claim 6, characterized in that The pre-processing unit comprises: An acquisition module, used to acquire the highest historical price in the historical price curve, and to acquire the highest historical load in the historical load curve; a normalization module, configured to normalize the historical prices in the historical price curve according to the highest historical price, and to normalize the historical loads in the historical load curve according to the highest historical load; A processing module is used to perform data processing on the historical prices in the normalized historical price curve and the historical loads in the normalized historical load curve according to a preset autoregressive sliding average model, wherein the data processing includes one or more of deleting bad data, restoring missing data and repairing abnormal data.

8. The device according to claim 6, characterized in that The prediction unit for classifying the preprocessed historical price curve and determining the predicted price curve according to the classification result includes: A first extraction module, used to extract a preset price index from the preprocessed historical price curve; A first construction module is used to construct and train a first neural network to obtain a first clustering model based on the extracted preset price index and the first sample data; A first clustering module, used for inputting the preprocessed historical price curve into the first clustering model for clustering to obtain multiple price patterns, each of which contains at least one preprocessed historical price curve; A first calculation module is used for calculating the average of the historical prices of the historical price curve in each price mode to obtain the historical price average value; A first determining module, configured to determine a first weight of each type of the price pattern according to the number of the historical price curves in each type of the price pattern; A first prediction module is used for simulating, for each type of the price pattern, a price curve corresponding to the price pattern based on the historical price average corresponding to the price pattern and the deviation distribution of the historical prices of the historical price curve; The first weighting module is used to perform weighted averaging of the price curves corresponding to all the simulated price patterns based on the first weight of each type of the price pattern, and use the highest historical price in the historical price curves to process the weighted average result to obtain a predicted price curve.

9. The device according to claim 6, characterized in that The prediction unit for classifying the preprocessed historical load curve and determining the predicted load curve according to the classification result includes: A second extraction module, used to extract a preset load index from the preprocessed historical load curve; A second construction module is used to construct and train a second neural network to obtain a second clustering model based on the extracted preset load index and in combination with second sample data; A second clustering module, used for inputting the preprocessed historical load curve into the second clustering model for clustering to obtain multiple types of load patterns, each type of load pattern including at least one preprocessed historical load curve; A second calculation module is used to calculate the average value of the historical load of the historical load curve in the load mode for each type of the load mode to obtain the historical load average value; A second determination module, configured to determine a second weight of each type of the load mode according to the number of the historical load curves in each type of the load mode; A second prediction module is used for simulating, for each type of the load mode, a load curve corresponding to the load mode based on the historical load average value corresponding to the load mode and the deviation distribution of the historical load of the historical load curve; The second weighting module is used to perform weighted averaging of the load curves corresponding to all the simulated load modes based on the second weight of each type of the load mode, and use the highest historical load in the historical load curves to process the weighted average result to obtain a predicted load curve.

Images