Dynamic rolling prediction-based reference power-frequency modulation bidding capacity determination method

Abstract

The invention discloses a reference power-frequency modulation bidding capacity determination method based on dynamic rolling prediction, and the method is characterized in that the method comprises the steps: constructing and training an Informer model, predicting the electricity price based on the Informer model, and enabling the electricity price to comprise an electric energy price and a frequency modulation price; and according to the electricity price predicted by the Informer model, a rolling optimization strategy is adopted, and the electric energy market reference power and the frequency modulation bidding capacity of each time period are determined by taking the maximum energy storage income as the target. According to the method, the long-term dependency relationship in the electricity price sequence can be effectively captured by using the Informer model, the prediction efficiency is improved, and the electric energy market reference power or the frequency modulation market bidding capacity of energy storage in the next time period is determined through online rolling optimization by comprehensively considering the operation parameters such as the charge state based on the accurate electricity price prediction result.

Description

Technical Field

[0001] The present invention belongs to the field of energy storage optimization, and particularly relates to a method for determining the reference power - frequency regulation bidding capacity based on dynamic rolling prediction. Background Art

[0002] Due to the high upfront construction cost of energy storage, energy storage power stations generally face problems such as long investment recovery periods and low rates of return. Although there is a large room for future cost reduction, at the current stage, with the still high battery cost, the mode of energy storage only participating in the electricity energy market and making profits through peak - valley price differences has a ceiling and cannot well stimulate the development of energy storage. In addition, due to the fast and accurate response ability of energy storage, it is very suitable for participating in auxiliary services such as power grid frequency regulation and ramping.

[0003] Furthermore, existing research is mostly limited to the short - term coordination problems of energy storage in the day - ahead electricity energy, spinning reserve, and frequency regulation markets, without considering the changes in intraday real - time market prices. Summary of the Invention

[0004] Aiming at the deficiencies of the existing technology, the present invention provides a method for determining the reference power - frequency regulation bidding capacity based on dynamic rolling prediction.

[0005] In a first aspect, an embodiment of the present invention provides a method for determining the reference power - frequency regulation bidding capacity based on dynamic rolling prediction, the method comprising:

[0006] Construct and train an Informer model, and predict electricity prices based on the Informer model, where the electricity prices include electricity energy prices and frequency regulation prices;

[0007] According to the electricity prices predicted by the Informer model, adopt a rolling optimization strategy to determine the reference power of the electricity energy market and the frequency regulation bidding capacity for each period with the goal of maximizing the energy storage revenue.

[0008] In a second aspect, an embodiment of the present invention provides an electronic device, comprising:

[0009] At least one processor; and

[0010] A memory communicatively connected to the at least one processor; wherein,

[0011] The memory stores one or more computer programs executable by the at least one processor, and the one or more computer programs are executed by the at least one processor so that the at least one processor can execute the above - mentioned method for determining the reference power - frequency regulation bidding capacity based on dynamic rolling prediction.

[0012] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned method for determining the benchmark power - frequency regulation bidding capacity based on dynamic rolling prediction is implemented.

[0013] In a fourth aspect, an embodiment of the present invention provides a computer program product, including a computer program / instructions. When the computer program / instructions are executed by a processor, the above-mentioned method for determining the benchmark power - frequency regulation bidding capacity based on dynamic rolling prediction is implemented.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0015] The present invention provides a method for determining the benchmark power - frequency regulation bidding capacity based on dynamic rolling prediction. The method of the present invention uses the Informer model to not only effectively capture the long-term dependence relationship in the electricity price sequence, improve the prediction efficiency, but also more accurately fit the fluctuation trend of the electricity price according to the periodic information in the electricity price, thereby improving the prediction accuracy. Based on the accurate electricity price prediction results, considering operation parameters such as the state of charge, the benchmark power in the electricity energy market or the frequency regulation market bidding capacity of the energy storage for the next time period is determined by online rolling optimization. Description of the Drawings

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0017] Figure 1 It is a flowchart of the method for determining the benchmark power - frequency regulation bidding capacity based on dynamic rolling prediction provided by the embodiment of the present invention;

[0018] Figure 2 It is a schematic block diagram of the rolling optimization model provided by the embodiment of the present invention;

[0019] Figure 3 It is a schematic diagram of the rolling optimization process provided by the embodiment of the present invention;

[0020] Figure 4 It is a schematic diagram of an electronic device provided by the embodiment of the present invention. Detailed Embodiments

[0021] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0022] It should be noted that, without conflict, the features in the following embodiments and implementation manners can be combined with each other.

[0023] The present invention provides a method for determining the benchmark power-frequency regulation bidding capacity based on dynamic rolling prediction, and the method includes:

[0024] Step S1, construct and train an Informer model, and predict electricity prices based on the Informer model, where the electricity prices include the electricity energy price and the frequency regulation price.

[0025] It should be noted that when the energy storage power station operator participates in the electricity market in a real-time response mode, as a price taker, the energy storage only needs to report the charge-discharge power in the electricity energy market or the bidding capacity in the frequency regulation market of the energy storage in the next time period before the time. In order to pursue its own maximum interests, the energy storage can make decisions according to the electricity energy market price or the frequency regulation price in the next time period, but these two types of prices are not disclosed before the time. Therefore, considering predicting these two types of electricity prices provides a reference for formulating a reasonable operation strategy for the energy storage before the time, which is beneficial to optimizing the energy storage's own power and reducing operation risks. In order to ensure that the reference value does not deviate much from the actual value, accurate prediction of electricity prices is particularly important. Generally speaking, the electricity price sequence has obvious periodicity, such as daily cycle, monthly cycle, seasonal cycle, etc. The Informer model has a strong non-linear fitting ability and can learn the complex non-linear relationships in the electricity price data. After enhancing with cycle information, the model can better adapt to the complex change patterns of electricity prices, more accurately fit the peak value, valley value and fluctuation trend of electricity prices, thereby improving the prediction accuracy. In addition, the electricity price sequence also has long-term correlation, that is, the current electricity price change is not only affected by the recent price, but may also be related to the price trend in the past for a long time. The self-attention mechanism of the Informer model can effectively capture the long-term dependence relationship in the electricity price sequence. And for the long-sequence data of the electricity price sequence, traditional prediction methods may have problems such as low calculation efficiency. The Informer model can greatly reduce the calculation complexity by adopting technologies such as the probabilistic sparse self-attention mechanism, and can greatly improve the prediction efficiency while ensuring the prediction accuracy.

[0026] Specifically, the step S1 includes:

[0027] Step S101: Obtain the historical electricity price data sequence, calculate its mean and standard deviation, identify the abnormal historical electricity price data, and then correct the abnormal historical electricity price data using the mean substitution method; and perform normalization.

[0028] P = {p 1 , p 2 , …, p N} (1)

[0029]

[0030] In the formula, P = {p 1 , p 2 , …, p N} represents the historical electricity price data sequence, N represents the number of historical electricity price data, μ p is the mean, and σ p is the standard deviation.

[0031] In this example, the data that satisfies |p i - μ p | > kσ p is regarded as abnormal data, where k = 3.

[0032] After that, perform normalization processing on the data, map the data to the [0, 1] interval, as follows:

[0033]

[0034] In the formula, p max , p min are the maximum and minimum values in the historical electricity price data respectively.

[0035] Step S102: Perform discrete Fourier transform on the normalized historical electricity price data sequence to obtain the electricity price sequence embedding features; the expression is as follows:

[0036]

[0037] P norm = {p 1,norm , p 2,norm , …, p n,norm …, p N,norm} (6)

[0038] In the formula, P DFT (k) represents performing discrete Fourier transform on the normalized historical electricity price data, and p n,norm represents the nth normalized historical electricity price data sequence;

[0039] According to P DFT (k), determine the daily cycle characteristics and calculate the daily average electricity price Standard deviation σ of daily electricity price d As the eigenvector of the daily cycle The expression is as follows:

[0040]

[0041] Step S103: Use trigonometric functions to perform feature embedding on the normalized historical electricity price data sequence to obtain periodic information embedding features.

[0042] Specifically, the position information in the normalized historical electricity price data sequence is encoded into the vector space through sine and cosine functions. Using the periodicity of trigonometric functions, the embeddings at different positions have different periodic variations, which is convenient for the model to capture the position information in the sequence, as follows:

[0043]

[0044] In the formula, pos represents the position of the data point in the entire sequence. For example, for a sequence with 24×30 = 720 data points, the value range of pos is 0 to 719; i represents the dimension index, and the value range is 0 to (d model / 2)-1, generating an embedding vector with multiple dimensions for each position; d model is the model dimension. Then, the periodic information of the electricity price sequence is embedded into the model. Similar to the embedding of the historical electricity price sequence, the periodic information is encoded using the periodicity of trigonometric functions, as follows:

[0045]

[0046] In the formula, PE (pos,2i) represents the position information of a certain data point in the electricity price sequence, and PE period (t, 2i) represents the periodic information of a certain data point in the electricity price sequence, t represents the sequence point, and T represents the period.

[0047] Step S104: Construct an Informer model, which consists of an encoder and a decoder.

[0048] Step S105: Concatenate the electricity price sequence embedding features and the periodic information embedding features as input features to train the Informer model.

[0049] Furthermore, there is often a certain error between the predicted value obtained by the encoder and the decoder and the true value. It is necessary to select an appropriate loss function to quantify this error, such as the mean square error:

[0050]

[0051] In the formula, is the predicted value, y t is the true value. According to the obtained loss function value, the parameters of the Informer model are updated using an optimization algorithm, and the parameter values of the Informer model are adjusted to minimize the loss function. The above process will be iterated multiple times, and the parameters of the Informer model will be updated each time, enabling the Informer model to gradually learn the patterns in the electricity price data.

[0052] Step S106, evaluate the trained Informer model using the mean absolute percentage error, and the expression is as follows:

[0053]

[0054] After multiple iterations of training and evaluation, the performance of the obtained model has reached a satisfactory level, and the trained model can be used for electricity price prediction for the next 24 hours, thus completing the entire electricity price prediction process. The prediction process is as Figure 1 shown.

[0055] Step S2, according to the electricity price predicted by the Informer model, adopt a rolling optimization strategy to determine the benchmark power of the electricity energy market and the frequency regulation bidding capacity for each time period with the goal of maximizing the energy storage benefit.

[0056] Before the hour, the energy storage needs to determine the benchmark power of the electricity energy market or the frequency regulation market bidding capacity. For the electricity energy market, the hourly benefit of the energy storage in the electricity energy market is:

[0057]

[0058] In the formula, represents the price of the electricity energy market, represents the benchmark power in the electricity energy market, a positive value indicates discharging, and a negative value indicates charging.

[0059] For the frequency regulation market, assuming the PJM frequency regulation market as the background, in this market, the frequency regulation signals are divided into traditional frequency regulation signals and dynamic frequency regulation signals, which are called RegA signals and RegD signals respectively. RegA is a frequency regulation signal designed for traditional frequency regulation resources with ramp rate limitations such as thermal power and hydropower units, while RegD is a frequency regulation signal designed for frequency regulation resources with fast ramp capabilities such as battery energy storage. Considering the characteristics of electrochemical energy storage, it is more suitable to respond to the frequently changing RegD signal. On the one hand, because the energy storage has almost no ramp constraint and can instantaneously output the power required by the system. On the other hand, the energy storage itself has limited stored energy and is suitable for responding to the RegD signal that is generally energy-neutral within a certain period (one hour).

[0060] The reference power in the electricity energy market changes hourly. Therefore, if the charging and discharging power of the energy storage within a certain hour is known, the change in the state of charge (SOC) of the energy storage in subsequent periods can be calculated. However, the RegD signal to which the energy storage responds changes every 2 seconds, and the frequency regulation instructions are highly random, making it difficult to accurately predict the changes in the frequency regulation instructions in the future for some time. This is not conducive to the energy storage to calculate and update its SOC in real time during the process of responding to frequency regulation. In addition, the total power of the energy storage is limited. The degree of participation of the energy storage in the electricity energy market within a certain hour will affect its revenue in the frequency regulation market, and vice versa, the degree of participation in the frequency regulation market will also affect its revenue in the electricity energy market. Therefore, the energy storage needs to determine the charging and discharging plan for subsequent periods based on the prices in the future for some time. The energy storage aims to maximize its own benefits, and the objective function is as follows:

[0061]

[0062] In the formula, is the revenue of the energy storage in the electricity energy market at time t, is the revenue in the frequency regulation market, represents the price in the frequency regulation market, represents the winning frequency regulation capacity of the energy storage at time t. The revenue in the frequency regulation market is not only related to the frequency regulation price and frequency regulation capacity but also related to the performance index S of the energy storage during the process of responding to the frequency regulation instructions t is related. The frequency regulation performance index has the same weight in three aspects: regulation accuracy, response time, and regulation speed. Due to the fast response characteristics of the energy storage, the response to the frequency regulation instructions is instantaneous. Therefore, the response time and regulation speed can be regarded as full score values. The PJM performance index can be appropriately simplified, and the calculation formula of this index is as follows:

[0063]

[0064] In the formula, is the winning frequency regulation capacity of the energy storage at time t, I is the total number of frequency regulation instructions within this period. Since the frequency regulation instructions are sent every 2 seconds, 1800 frequency regulation instructions will be sent sequentially within one hour. Therefore, I = 1800; r t,i is the i-th frequency regulation instruction, and its value range is -1 to 1; is the output power of the energy storage for the i-th frequency regulation instruction, is the average value of the absolute values of all frequency regulation instructions within time t; δ is the weight coefficient, and the value here is 2 / 3.

[0065] Since the FM performance indicators are calculated retrospectively, generally speaking, the regulation accuracy is determined by finding their maximum values in a 5-minute rolling window. Therefore, before the event, the energy storage is uncertain about its future FM performance indicators. The subsequent FM performance indicators of the energy storage can be determined by averaging the historical FM performance indicators of the current day. This indicator is only used to calculate the revenue of the energy storage before the event, and the actual performance indicators are calculated retrospectively.

[0066] When the energy storage optimizes its subsequent charging and discharging plans according to the predicted price, it also needs to meet the corresponding constraints, which are as follows:

[0067]

[0068] In the formula, Δt is 1 hour, ε is a 0 / 1 variable, and E t represents the SOC of the energy storage at each time period. represents the charging power of the energy storage at time t. represents the discharging power of the energy storage at time t. represents the output power of the energy storage, with a positive value for discharging and a negative value for charging. ε represents a 0-1 variable, and η is the charge-discharge efficiency of the energy storage. This formula represents the dynamic change of the SOC of the energy storage at each time period. Since the overall RegD signal responded by the energy storage is energy-neutral within one hour, the change of its SOC during this time period is ignored when the energy storage participates in frequency modulation, which simplifies the constraint model.

[0069]

[0070]

[0071] E L ≤E t ≤E H (24)

[0072] E t =E t+23 =0.5·E max (25)

[0073] Equation (21) means that the sum of the reference power of the energy storage participating in the energy market and the frequency modulation capacity cannot exceed the maximum power limit of the energy storage; Equation (22) represents the charge-discharge power limit of the energy storage; Equation (23) means that the capacity of the energy storage participating in frequency modulation cannot exceed its power limit; Equation (24) means that the SOC of the energy storage cannot exceed its upper and lower limits. Here, let E L be 10% of the total energy, and E H be 90% of the total energy; to ensure that the energy storage has the ability to participate in frequency modulation at each time period, Equation (25) limits its SOC to 50% of the total energy.

[0074] Linearize the above absolute value variables and introduce two sets of auxiliary variables Wherein:

[0075]

[0076] Therefore, the following equation holds:

[0077]

[0078] It should be noted that when energy storage participates in the market in a real-time response mode, electricity price forecasting can be carried out based on the Informer model enhanced by cycle information before the hour. Using the historical data of the past week, the electricity energy market price and frequency regulation market price for the next 24 hours are predicted before the hour. Energy storage determines the charge and discharge power plan of energy storage for the next 24 hours with the goal of maximizing its own benefits according to the predicted prices in the two service varieties and combining its own status, but actually only reports the plan for the next hour. On the one hand, a longer time period is used for forecasting, aiming to prevent energy storage power station operators from being short-sighted and making them not ignore potential greater economic benefits due to excessive focus on short-term interests, so as to provide a more forward-looking planning basis for long-term development. On the other hand, in actual operation, the short-cycle forecasting results are adopted mainly because the longer the forecasting duration, the lower the accuracy. Inaccurate forecast values are very likely to lead to wrong decisions. Therefore, in actual operation, the results optimized by short-cycle forecasting are more inclined to be adopted to ensure that the decision-making fits the current actual situation and reduce decision-making risks.

[0079] As a price taker, the benchmark power or frequency regulation capacity reported by energy storage to participate in the electricity energy market can basically be accepted. Before the next hour arrives, the above process is repeated, and the predicted values of electricity prices and charge and discharge plans are updated hourly in a rolling manner. The overall framework of the rolling optimization model is as Figure 2 shown.

[0080] Furthermore, energy storage forecasts the price trend within the next 24 hours before the hour, and based on these predicted prices, determines the most suitable benchmark power and frequency regulation capacity for each time period with the goal of maximizing the energy storage revenue. However, in the actual application scenario, in order to avoid price short-sightedness, we adopt a rolling optimization strategy. Specifically, when each new moment arrives, only the benchmark power and frequency regulation capacity values determined in the current first time period are taken to guide the actual operation. As time goes by, it rolls to the next moment, and the entire process from price forecasting to optimizing and determining the benchmark power and frequency regulation capacity is repeated again. The rolling optimization process is as Figure 3 shown.

[0081] In summary, the present invention provides a method for determining the benchmark power - frequency regulation bidding capacity based on dynamic rolling prediction. The method of the present invention uses the Informer model to not only effectively capture the long - term dependence relationship in the electricity price sequence, improve the prediction efficiency, but also more accurately fit the fluctuation trend of the electricity price according to the periodic information in the electricity price, thereby improving the prediction accuracy. Based on the accurate electricity price prediction results, considering operating parameters such as the state of charge, etc., the benchmark power in the electricity energy market or the frequency regulation market bidding capacity of the energy storage is determined by online rolling optimization for the next time period.

[0082] Correspondingly, the present application also provides an electronic device, including: one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the method for determining the benchmark power - frequency regulation bidding capacity based on dynamic rolling prediction as described above. As Figure 4 shown, it is a hardware structure diagram of any device with data processing capabilities where the method for determining the benchmark power - frequency regulation bidding capacity based on dynamic rolling prediction provided by the embodiment of the present invention is located. Except for Figure 4 the processors, memory, and network interfaces shown, any device with data processing capabilities where the device in the embodiment is located usually also includes other hardware according to the actual functions of the device with data processing capabilities, which will not be elaborated here.

[0083] Correspondingly, the present application also provides a computer - readable storage medium, on which computer instructions are stored, and when the instructions are executed by a processor, the method for determining the benchmark power - frequency regulation bidding capacity based on dynamic rolling prediction as described above is implemented. The computer - readable storage medium can be an internal storage unit of any device with data processing capabilities described in any of the foregoing embodiments, such as a hard disk or memory. The computer - readable storage medium can also be an external storage device, such as a plug - in hard disk, a Smart Media Card (SMC), an SD card, a FlashCard, etc., equipped on the device. Further, the computer - readable storage medium can also include both the internal storage unit of any device with data processing capabilities and the external storage device. The computer - readable storage medium is used to store the computer program and other programs and data required by any device with data processing capabilities, and can also be used to temporarily store the data that has been output or will be output.

[0084] Those skilled in the art will readily conceive of other embodiments of the present application after considering the specification and practicing the content disclosed herein. The present application is intended to cover any variations, uses, or adaptations of the present application, which follow the general principles of the present application and include known common general knowledge or conventional technical means in the technical field not disclosed in the present application. The specification and examples are only regarded as exemplary.

[0085] It should be understood that the present application is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A method for determining reference power-frequency modulation bidding capacity based on dynamic rolling prediction, characterized in that: The method comprises: Construct and train an Informer model, and predict electricity prices based on the Informer model, where the electricity prices include electricity energy prices and frequency regulation prices; According to the electricity price predicted by the Informer model, a rolling optimization strategy is adopted to determine the benchmark power and frequency regulation bidding capacity of the electricity market in each period with the goal of maximizing energy storage benefits.

2. According to the method for determining the base power-frequency modulation bidding capacity based on dynamic rolling prediction according to claim 1, the process of constructing and training the Informer model comprises: Obtain historical electricity price data series, filter out outliers and perform normalization; Perform discrete Fourier transform on the normalized historical electricity price data sequence to obtain the embedded features of the electricity price sequence; The trigonometric function is used to embed the normalized historical electricity price data sequence to obtain the period information embedding feature; Construct an Informer model, wherein the Informer model is composed of an encoder and a decoder; The electricity price series embedding features and the cycle information embedding features are concatenated as input features to train the Informer model.

3. According to the method for determining the base power-frequency regulation bidding capacity based on dynamic rolling prediction according to claim 1, the process of performing discrete Fourier transform on the normalized historical electricity price data sequence to obtain the embedded features of the electricity price sequence includes: P norm ={p 1,norm ,p 2,norm ,…,p n,norm …,p N,norm } According to P DFT (k) Determine daily cycle characteristics and calculate daily average electricity price Daily electricity price standard deviation σ d As the characteristic vector of the daily cycle The expression is as follows: Where P DFT (k) represents the discrete Fourier transform of the normalized historical electricity price data, p n,norm Represents the nth normalized historical electricity price data sequence.

4. According to the method for determining the benchmark power-frequency regulation bidding capacity based on dynamic rolling prediction in claim 1, the process of embedding the normalized historical electricity price data sequence using trigonometric functions to obtain the period information embedding feature comprises: The normalized historical electricity price data series is encoded into the vector space through sine and cosine functions. The periodicity of trigonometric functions is used to make the embedding of different positions have different periodic changes. The expression is as follows: Then, the periodic information of the electricity price series is embedded into the Informer model, and the periodicity of trigonometric functions is used to encode the periodic information. The expression is as follows: In the formula, pos represents the position of the data point in the entire historical electricity price data sequence, i represents the dimension index, and d model Indicates the dimension of the Informer model, PE (pos,2i) Indicates the location information of a data point in the electricity price series, PE period (t,2i) represents the period information of a data point in the electricity price sequence, t represents the sequence point, and T represents the period.

5. According to the method for determining the benchmark power-frequency regulation bidding capacity based on dynamic rolling prediction according to claim 1, the process of determining the benchmark power and frequency regulation bidding capacity of the electric energy market in each period based on the electricity price predicted by the informer model and adopting a rolling optimization strategy with the goal of maximizing energy storage benefits includes: Construct the objective function, the expression is as follows: In the formula, is the revenue of energy storage in the electric energy market during period t, For the revenue in the FM market, represents the price in the FM market, represents the winning frequency regulation capacity of energy storage in time period t, S t It is the frequency modulation performance index; Set the first constraint condition, the expression is as follows: In the formula, E t It represents the SOC of energy storage in each period. represents the charging power of the energy storage during period t, represents the discharge power of energy storage during period t, P t EA Represents the output power of energy storage, positive value means discharge, negative value means charge, ε represents a 0-1 variable, η is the charge and discharge efficiency of energy storage; Set the second constraint condition, the expression is as follows: AND L ≤E t ≤E H AND t =And t+23 =0.5·E max Where P max Indicates the maximum power limit of energy storage, E L 10% of the total energy, E H 90% of the total energy; According to the first constraint and the second constraint, the objective function is solved to obtain the electric energy market benchmark power and frequency regulation bidding capacity for each time period.

6. According to the method for determining the base power-frequency modulation bidding capacity based on dynamic rolling prediction according to claim 1, the frequency modulation performance index S t The calculation process includes: In the formula, is the winning frequency regulation capacity of energy storage in period t, I is the total number of frequency regulation instructions in this period, and r t,i is the ith frequency modulation instruction, is the output power of the energy storage for the i-th frequency modulation instruction, is the average value of the absolute values ​​of all frequency modulation instructions in time period t; δ is the weight coefficient.

7. According to the method for determining the reference power-frequency modulation bidding capacity based on dynamic rolling prediction according to claim 1, the rolling optimization strategy specifically comprises: Based on the electricity price predicted by the Informer model, with the goal of maximizing energy storage benefits, the benchmark power and frequency regulation bidding capacity for each hour from 1 to 24 hours are determined; The operation of the energy storage power station for the next 1h is carried out based on the benchmark power of the electric energy market and the frequency regulation bidding capacity corresponding to the first hour period; As the 1st iteration; When the time passes and rolls over to the second hour period, the second iteration is performed; And so on, for rolling optimization.

8. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores one or more computer programs that can be executed by the at least one processor, and the one or more computer programs are executed by the at least one processor to enable the at least one processor to execute the reference power-frequency regulation bidding capacity determination method based on dynamic rolling prediction as described in any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the computer program implements the method for determining the reference power-frequency regulation bidding capacity based on dynamic rolling prediction according to any one of claims 1 to 7.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the method for determining the reference power-frequency modulation bidding capacity based on dynamic rolling prediction described in any one of claims 1-7 is implemented.

Images