An optimized dispatching system for shared energy storage in the intraday market considering the integration of new energy sources

Abstract

The present invention discloses an optimized dispatching system for shared energy storage in the intraday market considering the access of new energy. The proposed model is a comprehensive framework composed of three modules, specifically including the following parts: (1) an intraday market load forecasting module based on CNN-LSTM; (2) a shared energy storage dispatching model construction module based on low carbon and low electricity purchase cost; (3) an optimization module based on the HHO algorithm. The present invention proposes to use the CNN-LSTM combination model to predict the intraday market load, and then construct a low-carbon and low-cost dispatching model for shared energy storage in the intraday market based on the intraday market predicted load, the day-ahead market clearing results and the collected real-time load, and finally optimize it using the Harrier Eagle optimization algorithm. The model realizes the optimized dispatching of shared energy storage in the intraday market while maximizing carbon emission reduction and minimizing system electricity purchase cost, and punishes the deviation of system energy output, effectively ensuring the reliability of the system and the rationality of power resource allocation.

Description

Technical Field

[0001] The present invention relates to an optimization dispatching system for shared energy storage in an intraday market taking into account the access of new energy, and belongs to the field of optimization dispatching of power market transactions. Background Art

[0002] Energy storage plays a vital role in user energy management, but the high investment cost of physical energy storage makes it unsuitable for small-scale distributed energy users in the electricity market. Therefore, load aggregators (ESAs) with energy storage properties can convert physical energy storage into separable virtual capacity and sell it to the system. This allows virtual energy storage to be shared among a group of users and between users and the system, reducing both investment in physical energy storage and the cost of using energy storage. In recent years, shared energy storage, as a new business model combining energy storage technology with the concept of the sharing economy, has the potential to play a significant role in the absorption of new energy and ensuring intraday market reliability. Existing shared energy storage optimization and dispatch systems do not consider scenarios involving the integration of new energy sources, focusing on economic efficiency while ignoring low-carbon development. Summary of the Invention

[0003] Technical problem: In response to the above-mentioned problems, the present invention proposes to use a CNN-LSTM combination model to predict the intraday market load, and then build an optimized scheduling model for shared energy storage in the intraday market based on the intraday market predicted load, the day-ahead market clearing results, and the collected real-time load, and finally optimize it using the Harrier Eagle optimization algorithm. This model achieves optimized scheduling of shared energy storage in the intraday market under the goals of maximizing carbon emission reduction and minimizing system electricity purchase costs, and considers corresponding deviation penalties for output deviations caused by insufficient prediction accuracy of new energy. The model also proposes that new energy can directly buy and sell electricity with shared energy storage operators, eliminate imbalances to avoid risks, and shared energy storage operators can profit from it based on the "sharing" feature. This optimized scheduling system not only rationally allocates resources for intraday transactions in the electricity market, but also improves the overall reliability of the system.

[0004] Technical solution: To achieve the above objectives, the present invention adopts the following technical solutions:

[0005] An optimized scheduling system for shared energy storage in the intraday market considering the integration of new energy sources mainly includes the following modules:

[0006] (1) Intraday market load forecasting module based on CNN-LSTM;

[0007] (2) Building a module for a shared energy storage dispatch model based on low carbon and low electricity purchase costs;

[0008] (3) Optimization module based on Harry Hawk Optimization (HHO) algorithm.

[0009] Specifically, the (1) includes the following steps:

[0010] (1-1) Using Convolutional Neural Network (CNN) to extract data features

[0011] First, the data is preprocessed and date factors are extracted from the historical load data set: Monday to Sunday and whether it is a holiday. The total data set is constructed by combining the five weather factors of temperature, humidity, wind speed, wind direction, and air pressure as well as the economic factors of peak, flat and valley electricity prices.

[0012] The total dataset is subjected to routine outlier detection, missing value filling, and data standardization.

[0013] Then, continuous feature maps are constructed according to the sliding time window as CNN input.

[0014] Build two layers of convolution and pooling layers and one fully connected layer, and the output of the fully connected layer serves as the input of the subsequent LSTM model.

[0015] (1-2) Load Forecasting Using Long Short-Term Neural Network (LSTM)

[0016] In the above CNN data feature extraction process, the final fully connected layer output is used as the input for two-layer LSTM training. The model evaluation indicators use root mean squared error (RMSE), mean absolute percentage error (MAPE), and coefficient of determination: R2 (R-Square).

[0017]

[0018] Where N represents the total number of samples, and y i Represent the predicted value and actual value of the i-th sample data respectively. Represents the mean of the i-th sample data. RMSE is an absolute indicator. RMSE uses a square term to magnify the difference between large and small errors, making it more sensitive to data with large prediction deviations. MAPE is a relative indicator. MAPE uses the actual value as the denominator and reflects the average percentage error. The smaller the values ​​of both indicators, the better the model's prediction accuracy.

[0019] Specifically, the (2) includes the following steps:

[0020] (2-1) Objective function construction

[0021] The objective function is to minimize the cost of system backup and maximize carbon emission reduction:

[0022]

[0023] Where, The cost of purchasing the power generation of the i-th renewable energy unit for the system at time t; is the penalty fee for the deviation of power generation of the i-th renewable energy unit at time t; The cost of purchasing the jth shared energy storage charging capacity for the system at time t; The cost of purchasing the jth shared energy storage unit to adjust the discharge capacity for the system at time t; Carbon reduction benefit for the system at time t; is the on-grid electricity price of the i-th renewable energy unit at time t; α is the renewable energy fluctuation coefficient: excessive generation = 1, no excessive generation = 0; η is the carbon emission equivalent.

[0024] Deviation value of power generation of the i-th new energy unit at time t:

[0025]

[0026] Where, is the actual output value of the i-th new energy unit at time t; is the predicted value of the output of the i-th new energy unit at time t in the market today.

[0027] The charging capacity of the jth shared energy storage at time t:

[0028]

[0029] Where, is the charging power of the j-th shared energy storage at time t.

[0030] Adjusted discharge value of the jth shared energy storage at time t:

[0031]

[0032] Where, is the discharge power of the jth shared energy storage at time t; is the discharge power of the jth shared energy storage at the previous moment (time t-1).

[0033] The reduction in power generation by the coal-fired units at time t is equal to the increase in discharge by the energy storage at time t, that is, the amount of power that the energy storage replaces the coal-fired units in participating in the auxiliary services:

[0034]

[0035] (2-2) Constraints

[0036] The constraints include system power balance constraints and shared energy storage charge and discharge state constraints. Other energy storage system constraints, such as overcharge and overdischarge capacity constraints and remaining power constraints, are not described in detail here.

[0037] 1) System power balance constraints

[0038]

[0039] Where, Output of the kth coal-fired unit at time t.

[0040] 2) System backup constraints

[0041] In the intraday market, since there may be large deviations in the day-ahead market forecast of renewable energy output, in order to ensure the safe operation of the power system, the system needs to have a certain amount of spare capacity.

[0042]

[0043] Where, is the remaining dischargeable capacity of the energy storage at time t, is the unit's output during period t, r is the system's reserve parameter, is the system load at time t.

[0044] 3) Quotation constraints

[0045]

[0046] Where, is the electricity price declared by the i-th renewable energy source at time t in the intraday market, q new energy,i is the day-ahead market electricity price of the i-th renewable energy unit at time t.

[0047] 4) Shared energy storage charging and discharging state constraints

[0048]

[0049] Where, and are the charging state and discharging state of the j-th shared energy storage at time t, respectively, and

[0050] 5) Constraints on shared energy storage charging and discharging capacity

[0051]

[0052] Where, is the maximum chargeable capacity of the j-th shared energy storage; is the maximum dischargeable capacity of the j-th shared energy storage.

[0053] Specifically, the (3) includes the following steps:

[0054] This module aims to optimize and solve the model constructed in step (2). Compared with traditional algorithms such as genetic algorithms and particle swarm optimization, HHO is better in terms of convergence speed and avoiding falling into local optimality, so the HHO algorithm is used to solve the objective function.

[0055] The fitness function is the objective function. The swarm is initialized based on the desired variable and the swarm's position is updated to minimize the objective function. Constraints are set to limit the position updates during the iteration process. After the maximum number of iterations is reached, the optimal result and corresponding parameters are output, and a fitness curve is plotted.

[0056] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages:

[0057] The present invention proposes an optimized scheduling system for shared energy storage in the intraday market that takes into account the access of new energy. The proposed strategy can enable new energy to directly buy and sell electricity with shared energy storage operators when there is an output deviation due to insufficient prediction accuracy, thereby eliminating the imbalance and avoiding risks. At the same time, shared energy storage operators can use their "sharing" characteristics to smooth out the output deviation of new energy and profit from it. The present invention starts from reducing the system's electricity purchase cost and maximizes the absorption of new energy, while also effectively reducing the overall carbon emissions of the power system. The present invention uses a CNN-LSTM model to perform intraday market load forecasting, combining the speed and lightweight of convolutional neural networks with the advantages of time series data training of long and short-term memory networks, considering the impact of historical power load, time and date, weather factors, and economic factors on power load, using CNN to extract data features, and has a smaller prediction error than the basic LSTM model. The present invention adopts the Harrier Eagle optimization algorithm for solving, which does not require assumptions during the solution process and can also solve complex models. Moreover, compared with the commonly used optimization algorithms - genetic algorithm and particle swarm algorithm, the Harrier Eagle optimization algorithm does not have crossover and mutation operations and hyperparameter initialization, so the principle is simpler, the parameters are fewer, and the implementation is easier, which ensures the efficiency of model solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 4 is an overall flow chart of the method of the present invention. DETAILED DESCRIPTION

[0059] The present invention will be further described below with reference to the accompanying drawings.

[0060] like Figure 1 The present invention relates to an optimization scheduling system for shared energy storage in the intraday market considering the access of new energy. The following describes each module.

[0061] (I): Intraday market load forecasting module based on CNN-LSTM.

[0062] (1-1) Using Convolutional Neural Network (CNN) to extract data features

[0063] First, the data is preprocessed and date factors are extracted from the historical load data set: Monday to Sunday and whether it is a holiday. The total data set is constructed by combining the five weather factors of temperature, humidity, wind speed, wind direction, and air pressure as well as the economic factors of peak, flat and valley electricity prices.

[0064] The total dataset is subjected to routine outlier detection, missing value filling, and data standardization.

[0065] Then, continuous feature maps are constructed according to the sliding time window as CNN input.

[0066] Build two layers of convolution and pooling layers and one fully connected layer, and the output of the fully connected layer serves as the input of the subsequent LSTM model.

[0067] (1-2) Load Forecasting Using Long Short-Term Neural Network (LSTM)

[0068] In the above CNN data feature extraction process, the final fully connected layer output is used as the input for two-layer LSTM training. The model evaluation indicators use root mean squared error (RMSE), mean absolute percentage error (MAPE), and coefficient of determination: R2 (R-Square).

[0069]

[0070] Where N represents the total number of samples, and y i Represent the predicted value and actual value of the i-th sample data respectively. Represents the mean of the i-th sample data. RMSE is an absolute indicator. RMSE uses a square term to magnify the difference between large and small errors, making it more sensitive to data with large prediction deviations. MAPE is a relative indicator. MAPE uses the actual value as the denominator and reflects the average percentage error. The smaller the values ​​of both indicators, the better the model's prediction accuracy.

[0071] (2): Building a module for a shared energy storage scheduling model based on low carbon and low electricity purchase costs.

[0072] (2-1) Objective function construction

[0073] The objective function is to minimize the cost of system backup and maximize carbon emission reduction:

[0074]

[0075]

[0076] Where, The cost of purchasing the power generation of the i-th renewable energy unit for the system at time t; is the penalty fee for the deviation of power generation of the i-th renewable energy unit at time t; The cost of purchasing the jth shared energy storage charging capacity for the system at time t; The cost of purchasing the jth shared energy storage unit to adjust the discharge capacity for the system at time t; Carbon reduction benefit for the system at time t; is the on-grid electricity price of the i-th renewable energy unit at time t; α is the renewable energy fluctuation coefficient: excessive generation = 1, no excessive generation = 0; η is the carbon emission equivalent.

[0077] Deviation value of power generation of the i-th new energy unit at time t:

[0078]

[0079] Where, is the actual output value of the i-th new energy unit at time t; is the predicted value of the output of the i-th new energy unit at time t in the market today.

[0080] The charging capacity of the jth shared energy storage at time t:

[0081]

[0082] Where, is the charging power of the j-th shared energy storage at time t.

[0083] Adjusted discharge value of the jth shared energy storage at time t:

[0084]

[0085] Where, is the discharge power of the jth shared energy storage at time t; is the discharge power of the jth shared energy storage at the previous moment (time t-1).

[0086] The reduction in power generation by the coal-fired units at time t is equal to the increase in discharge by the energy storage at time t, that is, the amount of power that the energy storage replaces the coal-fired units in participating in the auxiliary services:

[0087]

[0088] (2-2) Constraints

[0089] The constraints include system power balance constraints and shared energy storage charge and discharge state constraints. Other energy storage system constraints, such as overcharge and overdischarge capacity constraints and remaining power constraints, are not described in detail here.

[0090] 1) System power balance constraints

[0091]

[0092] Where, Output of the kth coal-fired unit at time t.

[0093] 2) System backup constraints

[0094] In the intraday market, since there may be large deviations in the day-ahead market forecast of renewable energy output, in order to ensure the safe operation of the power system, the system needs to have a certain amount of spare capacity.

[0095]

[0096] Where, is the remaining dischargeable capacity of the energy storage at time t, is the unit's output during period t, r is the system's reserve parameter, is the system load at time t.

[0097] 3) Quotation constraints

[0098]

[0099] Where, is the electricity price declared by the i-th renewable energy source at time t in the intraday market, q new energy,i is the day-ahead market electricity price of the i-th renewable energy unit at time t.

[0100] 4) Shared energy storage charging and discharging state constraints

[0101]

[0102] Where, and are the charging state and discharging state of the j-th shared energy storage at time t, respectively, and

[0103] 5) Constraints on shared energy storage charging and discharging capacity

[0104]

[0105] Where, is the maximum chargeable capacity of the j-th shared energy storage; is the maximum dischargeable capacity of the j-th shared energy storage.

[0106] (3): Optimization module based on the Harry Hawk Optimization (HHO) algorithm.

[0107] This module aims to optimize and solve the model constructed in step (2). Compared with traditional algorithms such as genetic algorithms and particle swarm optimization, HHO is better in terms of convergence speed and avoiding falling into local optimality, so the HHO algorithm is used to solve the objective function.

[0108] The fitness function is the objective function. The swarm is initialized based on the desired variable and the swarm's position is updated to minimize the objective function. Constraints are set to limit the position updates during the iteration process. After the maximum number of iterations is reached, the optimal result and corresponding parameters are output, and a fitness curve is plotted.

Claims

1. An optimization scheduling system for shared energy storage in the intraday market considering the access of new energy, characterized by: Includes the following modules: (1) Intraday market load forecasting module based on CNN-LSTM; (2) Building a module for a shared energy storage dispatch model based on low carbon and low electricity purchase costs; (3) Optimization module of HHO algorithm; The (2) shared energy storage scheduling model construction module based on low carbon and low electricity purchase cost is carried out according to the following steps: (2-1) Objective function construction The objective function is to minimize the cost of system backup and maximize carbon emission reduction: Where, The cost of purchasing the power generation of the i-th renewable energy unit for the system at time t; is the penalty fee for the deviation of power generation of the i-th renewable energy unit at time t; The cost of purchasing the jth shared energy storage charging capacity for the system at time t; The cost of purchasing the jth shared energy storage unit to adjust the discharge capacity for the system at time t; Carbon reduction benefit for the system at time t; is the on-grid electricity price of the i-th renewable energy unit at time t; α is the renewable energy fluctuation coefficient: excessive generation = 1, no excessive generation = 0; η is the carbon emission equivalent; Deviation value of power generation of the i-th new energy unit at time t: Where, is the actual output value of the i-th new energy unit at time t; is the predicted value of the output of the i-th new energy unit at time t in the market today; The charging capacity of the jth shared energy storage at time t: Where, The charging power of the jth shared energy storage at time t; Adjusted discharge value of the jth shared energy storage at time t: Where, is the discharge power of the jth shared energy storage at time t; is the discharge power of the jth shared energy storage at time t-1; The reduction in power generation by the coal-fired units at time t is equal to the increase in discharge by the energy storage at time t, that is, the amount of power that the energy storage replaces the coal-fired units in participating in the auxiliary services: (2-2) Constraints The constraints include system power balance constraints and shared energy storage charge and discharge state constraints; 1) System power balance constraints Where, Output of the kth coal-fired unit at time t; 2) System backup constraints In the intraday market, since there may be large deviations in the forecast value of renewable energy output in the day-ahead market, a certain amount of spare capacity is required to ensure the safe operation of the power system; Where, is the remaining dischargeable capacity of the energy storage at time t, is the unit's output during period t, r is the system's reserve parameter, is the system load at time t; 3) Quotation constraints Where, is the electricity price declared by the i-th renewable energy source at time t in the intraday market, q newenergy,i is the day-ahead market electricity price of the i-th renewable energy unit at time t; 4) Shared energy storage charging and discharging state constraints Where, and are the charging state and discharging state of the j-th shared energy storage at time t, respectively, and 5) Constraints on shared energy storage charging and discharging capacity Where, is the maximum chargeable capacity of the j-th shared energy storage; is the maximum dischargeable capacity of the j-th shared energy storage.

2. The optimization scheduling system for shared energy storage in the intraday market considering the access of new energy according to claim 1 is characterized in that: The (1) intraday market load forecasting module based on CNN-LSTM is performed in the following steps: (1-1) Using CNN to extract data features First, the data was preprocessed to extract date factors from the historical load dataset: Monday to Sunday and whether it was a holiday. The total dataset was constructed by combining five weather factors: temperature, humidity, wind speed, wind direction, and air pressure, as well as economic factors such as peak, flat, and valley electricity prices. Perform routine outlier detection, missing value filling, and data standardization on the total data set; Then construct a continuous feature map as CNN input according to the sliding time window; Build two convolutional and pooling layers and one fully connected layer. The output of the fully connected layer serves as the input of the subsequent LSTM model. (1-2) Using LSTM for load forecasting In the above CNN data feature extraction process, the final fully connected layer output is used as the input of the two-layer LSTM for training; the model evaluation indicators use the root mean square error RMSE, mean absolute percentage error MAPE and determination coefficient R2; as shown in the following formula: Where N represents the total number of samples, and y i Represent the predicted value and actual value of the i-th sample data respectively; Represents the average value of the i-th sample data; RMSE is an absolute indicator; RMSE uses the square term to magnify the gap between larger errors and smaller errors, making RMSE more sensitive to data with larger prediction deviations; MAPE is a relative indicator; MAPE uses the actual value as the denominator to reflect the average value of the percentage error; the smaller the values ​​of the two indicators, the better the model prediction accuracy.

3. The optimization scheduling system for shared energy storage in the intraday market considering the access of new energy according to claim 1 is characterized in that: The optimization module (3) based on the HHO algorithm is developed in the following steps: This module aims to optimize and solve the model constructed in step (2); use the HHO algorithm to solve the objective function; The fitness function is the objective function. The eagle population is initialized according to the required variables, and the population position is updated, that is, the variables are solved to minimize the objective function; the boundaries of the population position are set according to the constraints to limit the update of the position during the iteration process; after reaching the maximum number of iterations, the optimal result and corresponding parameters are output, and the fitness change curve is drawn.

Images