A shared energy storage capacity optimization method based on neural network prediction

Abstract

This invention discloses a method for optimizing shared energy storage capacity based on neural network prediction, comprising the following steps: constructing a long-term time-series prediction model based on a GRU structure; acquiring historical load data from the power grid structure and historical processing data of various types of generating units in the microgrid, and using the long-term time-series prediction model to predict the future period, thereby obtaining prediction data; establishing a two-layer optimization model for shared energy storage capacity configuration based on a combination of multiple energy forms, wherein the upper-layer model of the two-layer optimization model maximizes the revenue of shared energy storage, and the lower-layer model minimizes the sum of overall power generation cost and operating cost; solving the upper-layer model using a heuristic algorithm to obtain the upper-layer solution result, and using the CPLEX solver to solve the lower-layer model to obtain the lower-layer solution result; scheduling the energy storage capacity of the multi-microgrid structure based on the upper-layer solution result and the lower-layer solution result to obtain the optimal energy storage output strategy.

Description

Technical Field

[0001] This invention relates to the field of power system distribution network energy storage capacity optimization control, specifically a shared energy storage capacity optimization method based on neural network prediction. Background Technology

[0002] As the proportion of renewable energy in the power system continues to rise, the importance of energy storage, as a bidirectional energy device that can provide rapid response, is becoming increasingly prominent. Energy storage is a key device for the flexible conversion and comprehensive utilization of electricity and other energy sources, solving the problem of asynchrony between energy production and consumption. The rapid development of energy storage technology and the current peak-valley electricity pricing mechanism have created the possibility for energy storage systems to provide economic benefits on the electricity sales side. Proposing effective energy storage optimization configuration and operation control strategies on the electricity sales side, and analyzing its investment effectiveness and feasibility, are key to promoting the commercial application of energy storage.

[0003] Numerous academic papers have discussed the coordinated optimization analysis of integrated energy networks and energy storage. The paper "Optimized Operation of Integrated Energy System for Park Microgrids Considering the Integrated Demand Response of Multiple Loads (Electricity and Heat)" proposes an optimization model for integrated energy systems for park microgrids based on the integrated demand response of electricity and heat loads. The paper "Economic Optimization Operation of Integrated Energy System for Microgrids Including Electricity, Gas, and Heat Storage" analyzes the advantages of coordinated operation of multiple types of energy storage devices in improving the system's renewable energy absorption rate and economic efficiency. The paper "Multi-Microgrid Cooperative Optimization Scheduling Based on Cooperative Game Theory Considering Conditional Risk Value" analyzes an important form of shared energy storage, namely, achieving complementary energy storage capacity through power transfer between tie lines. Shared energy storage has high development potential in reducing investment costs, maximizing energy storage benefits and value, and facilitating user services. The paper "Energy-sharing provider for PV prosumer clusters: ahybrid approach using stochastic programming and Stackelberg game" establishes a master-slave game model between shared energy storage operators and multiple photovoltaic prosumers. However, the shared energy storage's capacity, charging and discharging power limits, and other parameters are constant, which may not be the optimal decision for shared energy storage operators. Few studies have considered the optimal configuration of shared energy storage capacity; how to set reasonable parameters to prevent the wasted capacity is a key concern for shared energy storage operators. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for optimizing shared energy storage capacity based on neural network prediction, comprising the following steps:

[0005] Step 1: Construct a long-term time-series prediction model based on the GRU structure;

[0006] Step 2: Obtain historical load data in the power grid structure and historical output data of each unit in the microgrid, including output data of photovoltaic and wind turbine generators. Use a long-term time series prediction model to predict the output of photovoltaic and wind power within a set time period to obtain the output prediction data of photovoltaic and wind power.

[0007] Step 3: Establish a two-layer optimization model for shared energy storage capacity configuration based on the combination of multiple energy forms. The upper layer of the two-layer optimization model maximizes the revenue of shared energy storage, while the lower layer model minimizes the total sum of overall power generation cost and operating cost.

[0008] Step 4: Solve the upper-level model using a heuristic algorithm to obtain the upper-level solution result, and then use the CPLEX solver to solve the lower-level model to obtain the lower-level solution result.

[0009] Step 5: Based on the upper-layer and lower-layer solution results, schedule the energy storage capacity of the multi-microgrid structure to obtain the optimal energy storage output strategy.

[0010] Furthermore, the construction of the long-term time-series prediction model based on the GRU structure includes the following process:

[0011] The long-term prediction model structure includes an input layer, a dropout layer, a GRU layer, and an output layer. Historical load data is used as input, and a random masking matrix is ​​added to the weight matrix through the dropout layer. Each neuron in each layer of the neural network obtains its output by applying a weight matrix and a nonlinear function to the input data. The weight matrix is ​​continuously corrected with each training iteration, so that the GRU layer neural network can obtain the same result as training with multiple databases through multiple training iterations from a single database. The GRU layer learns the internal variation patterns of the load from the proposed network features to achieve the prediction function, and finally, the prediction result is obtained through the output layer.

[0012] Furthermore, the acquisition of historical load data from the power grid structure and historical processing data from various types of generating units in the microgrid, along with the use of a long-term time-series forecasting model to predict future periods, yields forecast data, including:

[0013]

[0014] λ′ is the insurance factor; L′ Δt P′ represents the grid load level at the moment when the difference between the predicted microgrid load and the processed load is greatest. w，Δt P′ represents the wind power output level at the moment when the difference between the microgrid load and the processing capacity is greatest in the predicted values. pv，Δt P′ represents the photovoltaic output level at the moment when the difference between the microgrid load and the processing capacity is greatest in the predicted values. MT，ΔtThis represents the gas turbine output level at the moment when the difference between the microgrid load and the processing capacity is greatest in the predicted values.

[0015] Furthermore, the aforementioned two-layer optimization model for shared energy storage capacity configuration based on a combination of multiple energy forms, wherein the upper-layer model maximizes the revenue from shared energy storage, and the lower-layer model minimizes the sum of total power generation cost and operating cost, includes:

[0016] The objective function of the upper-level model is set as follows:

[0017]

[0018] In the formula: Δt is the study period; λ SESS The fee payable per unit charge / discharge capacity for microgrids using shared energy storage services; n is the number of microgrids in the studied scenario. These represent the charging and discharging power of microgrid i using shared energy storage during time period t; λ P , λ E These are the power cost and capacity cost of shared energy storage, respectively. These represent the power limit and rated capacity of shared energy storage, respectively.

[0019] The objective function of the lower-level model is set as follows:

[0020]

[0021] In the formula For the operating cost of the steam turbine; The cost of purchasing electricity from the power grid; The cost of removing the load; The fees that microgrid i needs to pay for using shared energy storage during time period t.

[0022] Furthermore, the two-level optimization model also includes constraints, which include:

[0023] Wind power output constraints:

[0024] P w.min ≤P w ≤P w.max

[0025] In the formula P w.min P represents the minimum output power for wind power generation. w.max This is to generate the maximum output power for wind power generation.

[0026] Photovoltaic output constraints:

[0027] P pv.min ≤P pv ≤P pv.max

[0028] In the formula Ppv.min This represents the minimum output power of photovoltaic power generation; P pv.max To maximize the output of photovoltaic power generation.

[0029] Gas turbine output constraints:

[0030] P SESS.min ≤P SESS ≤P SESS.max

[0031] Gas turbine output costs and constraints:

[0032]

[0033] Electrical output constraint:

[0034]

[0035] The output of the gas turbine within the microgrid i at time t; c gas The unit price of natural gas; η MT and Q LHV These are power generation efficiency and the calorific value of natural gas, respectively. and These represent the maximum and minimum output values, respectively.

[0036] Power constraints in multi-microgrid systems:

[0037]

[0038] P w.sph P represents the output power of wind power generation. pv.sph This refers to the output power of photovoltaic power generation. The power released or absorbed by the i-th energy storage device;

[0039] Removal load cost:

[0040]

[0041] The corresponding constraints are:

[0042]

[0043] In the formula λ represents the electrical load cut off by microgrid i at time t; c The unit cost of disconnecting electrical load; This is the upper limit of the load to be removed;

[0044] Demand response costs and constraints:

[0045] The demand response cost is:

[0046]

[0047] The corresponding constraints are:

[0048]

[0049]

[0050] λ r，e , These represent the cost required to change a unit of electrical load and the amount of electrical load adjustment required by microgrid i at time t, respectively. ε represents the initial value of the electrical load of microgrid i at time t; e γ e These are the maximum allowable adjustment ratio of electrical load and the percentage of the total maximum allowable adjustment of electrical load, respectively.

[0051] The energy storage parameter constraints and electricity trading constraints are as follows:

[0052] Energy storage device state of charge constraints

[0053] The state of charge during charging can be represented as:

[0054]

[0055] The state of charge during discharge can be expressed as:

[0056]

[0057] In the formula δ i Let η be the self-discharge rate of the i-th energy storage device; c.i η d.i These are the charging efficiency and discharging efficiency of the i-th energy storage device, respectively; P sph.c.i P sph.d.i Let C represent the charging and discharging power of the i-th energy storage device at any scenario, stage, and time; c.i Let be the rated capacity of the i-th energy storage device;

[0058] Energy storage device charging and discharging power constraints:

[0059] 0≤P sphci ≤P r.i

[0060] Energy storage usage costs and constraints:

[0061]

[0062] Power interaction constraints with shared energy storage:

[0063]

[0064]

[0065]

[0066]

[0067]

[0068]

[0069]

[0070]

[0071] In the formula: This represents the charging and discharging status of microgrid i during time period t. It is a Boolean variable, with a value of 1 indicating charging and 0 indicating discharging, thus restricting the microgrid to only charging or discharging during any given time period. These represent the upper limits of charging and discharging power of shared energy storage used by microgrid i during time period t; η SESS，c η SESS，d These are the charging and discharging efficiencies of the shared energy storage, respectively. Let be the actual capacity used by microgrid i during time period t; when there is an imbalance between power supply and demand within the microgrid, the microgrid can also purchase electricity from the distribution network. Let the electricity purchase cost of microgrid i during time period t be:

[0072]

[0073] The electricity purchase constraint is: 0 ≤ P sphd.i ≤P r.i

[0074]

[0075] These represent the electricity purchased by microgrid i during time period t and the electricity price sold by the grid.

[0076] Furthermore, the solution process for the two-level optimization model is as follows:

[0077] Step 1: Initialize the parameters of the shared energy storage system and each integrated energy microgrid. The number of iterations k = 0, the population size m is set to 40, the total number of iterations is 20, the population mutation rate is 5%, and the crossover probability is 80%.

[0078] Step 2: Initially generate the rated capacity of m shared energy storage systems randomly using a genetic algorithm, and transmit the parameters to the lower layer;

[0079] Step 3: Update the iteration count, k = k + 1.

[0080] Step 4: Each integrated energy microgrid sequentially receives the initial rated capacity values ​​of m shared energy storage systems, uses the CPLEX solver to solve for the output of equipment in each microgrid, the interaction power with the distribution network, and the interaction power with the shared energy storage system, retains the operating cost of each microgrid, and returns the optimized interaction power with the shared energy storage system to the upper layer.

[0081] Step 5: The shared energy storage system calculates the revenue under the current m rated capacity values ​​based on the power interaction values ​​returned by each integrated energy microgrid within a day, and retains the current revenue F. SESS，k ;

[0082] Step 6: Use a genetic algorithm to select and mutate to generate a new shared energy storage system capacity value. Repeat steps 4 and 5 to calculate the shared energy storage system revenue value F'. SESS，k ;

[0083] Step 7: If the benefits or costs of the shared energy storage system and each integrated energy microgrid converge, end the program; if not, return to step 3.

[0084] The beneficial effects of this invention are as follows: Under the background of the sharing economy, this paper proposes a shared energy storage configuration optimization model for the scenario of joint optimization of energy storage in multi-integrated energy microgrids. The proposed shared energy storage dynamic capacity leasing model can allocate optimal capacity to each microgrid, overcoming the user power interaction phenomenon introduced by traditional modeling methods, and revealing the essence of shared energy storage improving efficiency. It also improves the accuracy of shared energy storage model establishment. Attached Figure Description

[0085] Figure 1 This is a flowchart illustrating a shared energy storage capacity optimization method based on neural network prediction.

[0086] Figure 2 This is a schematic diagram of the prediction model structure for a shared energy storage capacity optimization method based on neural network prediction.

[0087] Figure 3 Schematic diagram of wind power generation forecast for each microgrid;

[0088] Figure 4 Schematic diagram of gas turbine output curves for each microgrid under Mode 1;

[0089] Figure 5 Schematic diagram of gas turbine output curves for each microgrid in Mode 2;

[0090] Figure 6 Schematic diagram of the gas turbine output curves of each microgrid under Mode 3. Detailed Implementation

[0091] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited to the following description.

[0092] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention; that is, the described embodiments are only a part of the embodiments of the invention, and not all of them. The components of the embodiments of the invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0093] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention. It should be noted that relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0094] Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0095] The features and performance of the present invention will be further described in detail below with reference to embodiments.

[0096] like Figure 1 As shown,

[0097] S1 utilizes historical data to construct a long-term time-series forecasting model based on a GRU structure. Figure 1 As can be seen, the prediction model structure mainly consists of an input layer, a dropout layer, a GRU layer, and an output layer. Historical load data is used as input, and the dropout layer adds a random masking matrix on top of the weight matrix, allowing subsequent network feature extraction to obtain results equivalent to training on multiple databases through multiple training sessions from a single database. The GRU layer learns the internal variation patterns of the load from the extracted features to achieve the prediction function, and finally, the prediction result is obtained through the output layer.

[0098] The S2 model utilizes historical data from the power grid structure and historical data from various types of generating units in the microgrid to predict the future. The historical data first needs to be cleaned to remove outliers, and then normalized.

[0099] S3 forecast data serves as the fundamental basis for determining and allocating shared energy storage capacity.

[0100]

[0101] λ′ is the insurance factor; L′ Δt P′ represents the grid load level at the moment when the difference between the predicted microgrid load and the processed load is greatest. w，Δt P′ represents the wind power output level at the moment when the difference between the microgrid load and the processing capacity is greatest in the predicted values. pv，Δt P′ represents the photovoltaic output level at the moment when the difference between the microgrid load and the processing capacity is greatest in the predicted values. MT，Δt This represents the gas turbine output level at the moment when the difference between the microgrid load and the processing capacity is greatest in the predicted values.

[0102] S4 is designed for shared energy storage capacity based on multiple microgrids and including various power generation devices. It needs to consider not only the stability of power consumption in the microgrids, but also the operation of the shared energy storage operators. Therefore, it proposes a two-layer optimization model for the configuration of shared energy storage capacity based on the combination of multiple energy forms.

[0103] The objective function of the upper-level model is set as follows:

[0104]

[0105] In the formula: Δt is the study period; λ SESS The fee payable per unit charge / discharge capacity for microgrids using shared energy storage services; n is the number of microgrids in the studied scenario. These represent the charging and discharging power of microgrid i using shared energy storage during time period t; λ P , λ E These are the power cost and capacity cost of shared energy storage, respectively. These represent the power limit and rated capacity of shared energy storage, respectively.

[0106] The objective function of the lower-level model is set as follows:

[0107]

[0108] In the formula, FC represents the energy storage installation cost, which is a linear function of the rated capacity and rated power of the energy storage equipment. The calculation formula is as follows:

[0109]

[0110] In the formula P rC represents the rated capacity of the energy storage device. r C represents the rated capacity of the energy storage device. ES λ represents the unit capacity cost coefficient of the energy storage equipment; Y represents the service life of the energy storage equipment; and λ represents the annual interest rate.

[0111] In the formula, VC includes system operating costs (the cost of buying and selling electricity to interconnected microgrids and the main grid) and load loss penalty costs. Its calculation formula is as follows:

[0112]

[0113] In the formula N S P represents the number of scenes; S The number of stages in each scene; NP represents the number of stages; N D Number of days for each phase; N H For hours per day; N m It is the number of its connected power grids; P sph Electricity prices are set for corresponding scenarios, stages, and times. To determine the power exchanged with the j-th power grid under the corresponding scenario, stage, and time; For load power in the corresponding scenario, stage, and time; LSI sph , where represents the load shortage rate under the corresponding scenario, stage, and time; VL represents the value of a unit of power shortage load. In the two-layer model, the upper layer model maximizes the revenue from shared energy storage, while the lower layer model minimizes the total sum of overall generation and operating costs.

[0114] S5 needs to set constraints on the model to take into account the actual operating characteristics and power interaction limitations of the units in the microgrid.

[0115] Wind power output constraints:

[0116] P w.min ≤P w ≤P w.max

[0117] In the formula P w.min P represents the minimum output power for wind power generation. w.max This is to generate the maximum output power for wind power generation.

[0118] Photovoltaic output constraints:

[0119] P pv.min ≤P pv ≤P pv.max

[0120] In the formula P pv.min This represents the minimum output power of photovoltaic power generation; P pv.max To maximize the output of photovoltaic power generation.

[0121] Gas turbine output constraints:

[0122] P SESS.min ≤P SESS ≤P SESS.max

[0123] Power constraints in multi-microgrid systems:

[0124]

[0125] In the formula, all parameters represent parameter values ​​under arbitrary scenario, stage, and time conditions. P w.sph . represents the output power of wind power generation; P pv.sph This refers to the output power of photovoltaic power generation. This represents the power released or absorbed by the i-th energy storage device.

[0126] Energy storage parameter constraints and electricity trading constraints are as follows:

[0127] Energy storage device state of charge constraints

[0128] The state of charge during charging can be represented as:

[0129]

[0130] The state of charge during discharge can be expressed as:

[0131]

[0132] In the formula δ i Let η be the self-discharge rate of the i-th energy storage device; c.i η d.i These are the charging efficiency and discharging efficiency of the i-th energy storage device, respectively; P sph.c.i P sph.d.i Let C represent the charging and discharging power of the i-th energy storage device at any scenario, stage, and time; c.i Let be the rated capacity of the i-th energy storage device.

[0133] In addition, the following formula represents the charging and discharging power constraint of energy storage devices:

[0134] 0≤P sphci ≤P r.i

[0135] Power interaction constraints with shared energy storage:

[0136]

[0137]

[0138]

[0139]

[0140]

[0141]

[0142]

[0143]

[0144] In the formula: This represents the charging and discharging status of microgrid i during time period t. It is a Boolean variable, with a value of 1 indicating charging and 0 indicating discharging, thus restricting the microgrid to only charging or discharging during any given time period. These represent the upper limits of charging and discharging power of shared energy storage used by microgrid i during time period t; η SESS，c η SESS，d These are the charging and discharging efficiencies of the shared energy storage, respectively. Let be the actual capacity used by microgrid i during time period t. When there is an imbalance between power supply and demand within the microgrid, the microgrid can also purchase electricity from the distribution network. Let the electricity purchase cost of microgrid i during time period t be:

[0145]

[0146] The electricity purchase constraint is: 0 ≤ P sphd.i ≤P r.i

[0147]

[0148] In the formula: These represent the electricity purchased by microgrid i during time period t and the electricity price sold by the grid.

[0149] S6 proposes an optimization algorithm that combines heuristics to solve the model. The heuristic algorithm mainly serves as a solution tool for the upper-level algorithm, while the lower-level algorithm uses the CPLEX solver.

[0150] The model solution process is as follows.

[0151] Step 1: Initialize the parameters of the shared energy storage system and each integrated energy microgrid. The number of iterations k = 0, the population size m is set to 40, the total number of iterations is 20, the population mutation rate is 5%, and the crossover probability is 80%.

[0152] Step 2: Use a genetic algorithm to initially randomly generate the rated capacity of m shared energy storage systems and transmit the parameters to the lower layer.

[0153] Step 3: Update the iteration count, k = k + 1.

[0154] Step 4: Each integrated energy microgrid sequentially receives the initial rated capacity values ​​of m shared energy storage systems, uses the CPLEX solver to solve for the output of equipment in each microgrid, the interaction power with the distribution network, and the interaction power with the shared energy storage system, retains the operating cost of each microgrid, and returns the optimized interaction power with the shared energy storage system to the upper layer.

[0155] Step 5: The shared energy storage system calculates the revenue under the current m rated capacity values ​​based on the power interaction values ​​returned by each integrated energy microgrid within a day, and retains the current revenue F. SESS，k

[0156] Step 6: Use a genetic algorithm to select and mutate to generate a new shared energy storage system capacity value. Repeat steps 4 and 5 to calculate the shared energy storage system revenue value F'. SESS，k .

[0157] Step 7: If the benefits or costs of the shared energy storage system and each integrated energy microgrid converge, end the program; if not, return to step 3.

[0158] S7 schedules the energy storage capacity of the multi-microgrid structure based on the solver results and formulates an optimal energy storage output strategy.

[0159] Example 2

[0160] Based on Example 1, the scenario under study is assumed to consist of 3 microgrids and 1 SESS. Each microgrid has an initial energy storage capacity that is half of the capacity it leases from the shared energy storage at the initial moment. The rated capacity to power limit ratio is 0.2. The unit charge / discharge service cost is 0.3 yuan / (kW·h), the capacity cost and power cost are 1100 yuan / kW and 1000 yuan / kW respectively, the annual maintenance cost is 72 yuan, and the lifespan is 8 years. The power interaction limit between all microgrids and the distribution network is 150kW, the load shedding limit is 200kW, and the unit load transfer cost is 0.5 yuan / kW. e With ε h All values ​​are taken as 0.05, γ e With γ h All values ​​are set to 0.12. Mode 1 is defined as all microgrids operating independently without energy storage devices; Mode 2 is defined as all microgrids operating independently with energy storage devices configured within them; and Mode 3 is defined as multiple microgrids operating collaboratively using shared energy storage services without considering demand response.

[0161] Table 1. Profits of SESS operators and total operating costs of microgrid clusters under three models

[0162]

[0163] As shown in Table 1, the total operating cost of the microgrid cluster under Mode 2 is reduced by RMB 2932.78 compared to Mode 1. This is because the energy storage equipment configured within each microgrid can further absorb wind power resources and reduce the load shedding cost of the microgrid. Compared to Mode 2, the shared energy storage in Mode 3 generates a profit of RMB 237.34, while the total operating cost of the microgrid cluster is reduced by RMB 72.28. The shared energy storage mechanism achieves a win-win situation for both the shared energy storage operator and the microgrid cluster. This is because the dynamic capacity leasing mechanism proposed in this paper can avoid wasting unnecessary energy storage capacity and further reduce the investment and construction costs of energy storage equipment.

[0164] Table 2. Rated capacity and power limitation configuration results for energy storage devices under Mode 2 and Mode 3

[0165] Mode Object Capacity / (kW·h) Power / kW Mode 2 Microgrid 1 858.38 171.68 Mode 2 Microgrid 2 1 325.06 265.01 Mode 3 Microgrid 3 0 0 Mode 3 SESS 2 188.62 437.82

[0166] As shown in Table 2, the capacity configurations of energy storage devices in microgrids 1 to 3 under Mode 2 are 858.38 kWh, 1325.06 kWh, and 0 kWh, respectively, totaling 2183.44 kWh. This is 5.18 kWh more than in Mode 3. This is because the shared energy storage mechanism can effectively allocate and utilize capacity, thereby saving unnecessary capacity configuration.

[0167] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A method for optimizing shared energy storage capacity based on neural network prediction, characterized in that, Includes the following steps: Step 1: Construct a long-term time-series prediction model based on the GRU structure; Step 2: Obtain historical load data in the power grid structure and historical output data of each unit in the microgrid, including output data of photovoltaic and wind turbine generators. Use a long-term time series prediction model to predict the output of photovoltaic and wind power within a set time period to obtain the output prediction data of photovoltaic and wind power. Step 3: Establish a two-layer optimization model for shared energy storage capacity configuration based on the combination of multiple energy forms. The upper layer of the two-layer optimization model maximizes the revenue of shared energy storage, while the lower layer model minimizes the total sum of overall power generation cost and operating cost. Step 4: Solve the upper-level model using a heuristic algorithm to obtain the upper-level solution result, and then use the CPLEX solver to solve the lower-level model to obtain the lower-level solution result. Step 5: Based on the upper-layer and lower-layer solution results, schedule the energy storage capacity of the multi-microgrid structure to obtain the optimal energy storage output strategy; The aforementioned two-layer optimization model for shared energy storage capacity configuration based on a combination of multiple energy forms, wherein the upper layer of the two-layer optimization model maximizes the revenue from shared energy storage, and the lower layer model minimizes the sum of total power generation cost and operating cost, includes: The objective function of the upper-level model is set as follows: In the formula: For the research period; The fee payable per unit charge / discharge capacity for microgrids using shared energy storage services; The number of microgrids in the studied scenario; , microgrids exist The charging and discharging power of shared energy storage used during specific time periods; , These are the power cost and capacity cost of shared energy storage, respectively. , These are the power limit and rated capacity of shared energy storage, respectively. The objective function of the lower-level model is set as follows: In the formula For the operating cost of the steam turbine; The cost of purchasing electricity from the power grid; The cost of removing the load; The fees that microgrid i needs to pay for using shared energy storage during time period t.

2. The method for optimizing shared energy storage capacity based on neural network prediction according to claim 1, characterized in that, The construction of a long-term time-series prediction model based on the GRU structure includes the following process: The long-term prediction model structure includes an input layer, a dropout layer, a GRU layer, and an output layer. Historical load data is used as input, and a random masking matrix is ​​added to the weight matrix through the dropout layer. Each neuron in each layer of the neural network obtains its output by applying a weight matrix and a nonlinear function to the input data. The weight matrix is ​​continuously corrected with each training iteration, so that the GRU layer neural network can obtain the same result as training with multiple databases through multiple training iterations from a single database. The GRU layer learns the internal variation patterns of the load from the proposed network features to achieve the prediction function, and finally, the prediction result is obtained through the output layer.

3. The method for optimizing shared energy storage capacity based on neural network prediction according to claim 1, characterized in that, The aforementioned acquisition of historical load data in the power grid structure and historical processing data of various types of generating units in the microgrid, followed by prediction of a future period using a long-term time-series forecasting model, yields forecast data, including: This is the insurance factor; This represents the grid load level at the moment when the difference between the predicted microgrid load and the processed load is the largest. The wind power output level at the moment when the difference between the microgrid load and the processing capacity is the largest in the predicted values; This represents the photovoltaic output level at the moment when the difference between the predicted microgrid load and the processing capacity is greatest. This represents the gas turbine output level at the moment when the difference between the microgrid load and the processing capacity is greatest in the predicted values.

4. The method for optimizing shared energy storage capacity based on neural network prediction according to claim 1, characterized in that, The two-level optimization model also includes constraints, which include: Wind power output constraints: In the formula This is the minimum output power for wind power generation; This is to achieve the maximum output of wind power generation; Photovoltaic output constraints: In the formula This is the minimum output power for photovoltaic power generation. To maximize the output of photovoltaic power generation; Gas turbine output constraints: Gas turbine output costs and constraints: Electrical output constraint: The output of the gas turbine within the microgrid i at time t; This refers to the unit price of natural gas. and These are power generation efficiency and the calorific value of natural gas, respectively. and These represent the maximum and minimum output values, respectively. Power constraints in multi-microgrid systems: This refers to the output power of wind power generation. This refers to the output power of photovoltaic power generation. For the first The power released or absorbed by the energy storage device; Removal load cost: The corresponding constraints are: In the formula The electrical load cut off by microgrid i at time t; The unit cost of disconnecting electrical load; This is the upper limit of the load to be removed; Demand response costs and constraints: The demand response cost is: The corresponding constraints are: These represent the cost required to change a unit of electrical load and the amount of electrical load adjustment required by microgrid i at time t, respectively. The original value of the electrical load of microgrid i at time t; , These are the maximum allowable adjustment ratio of electrical load and the percentage of the total maximum allowable adjustment of electrical load, respectively. The energy storage parameter constraints and electricity trading constraints are as follows: Energy storage device state of charge constraints The state of charge during charging can be represented as: The state of charge during discharge can be expressed as: In the formula For the first Self-discharge rate of Taiwan energy storage devices , The first The charging and discharging efficiency of the energy storage device; , The first, second, and third times respectively in any scenario, stage, and time. The charging and discharging power of the energy storage device; For the first The rated capacity of the energy storage device; Energy storage device charging and discharging power constraints: Energy storage usage costs and constraints: Power interaction constraints with shared energy storage: In the formula: microgrid exist The charging and discharging status during a time period is a Boolean variable, with a value of 1 indicating charging and 0 indicating discharging, thus restricting the microgrid to only charging or discharging during any given time period. , microgrids exist The upper limit of charging and discharging power for shared energy storage during specific time periods; , These are the charging and discharging efficiencies of the shared energy storage, respectively. For micro-network exist The actual capacity used during a given time period; when there is an imbalance between power supply and demand within the microgrid, the microgrid can also purchase power from the distribution network. (The last sentence appears to be incomplete and possibly refers to a different context.) exist The electricity purchase cost during the period is: Electricity purchase constraints are: microgrids exist The amount of electricity purchased and the electricity price sold by the grid during the time period.

5. The method for optimizing shared energy storage capacity based on neural network prediction according to claim 4, characterized in that, The model solution process is as follows: Step 1: Initialize the parameters of the shared energy storage system and each integrated energy microgrid, with the number of iterations k=0, the population size m set to 40, the total number of iterations to 20, the population mutation rate to 5%, and the crossover probability to 80%. Step 2: Initially generate the rated capacity of m shared energy storage systems randomly using a genetic algorithm, and transmit the parameters to the lower layer; Step 3: Update the iteration count, k = k + 1; Step 4: Each integrated energy microgrid sequentially receives the initial rated capacity values ​​of m shared energy storage systems, uses the CPLEX solver to solve for the output of equipment in each microgrid, the interaction power with the distribution network, and the interaction power with the shared energy storage system, retains the operating cost of each microgrid, and returns the optimized interaction power with the shared energy storage system to the upper layer. Step 5: The shared energy storage system calculates the revenue under the current m rated capacity values ​​based on the power interaction values ​​returned by each integrated energy microgrid within a day, and retains the current revenue. ; Step 6: Use a genetic algorithm to select and mutate to generate new shared energy storage system capacity values. Repeat steps 4 and 5 to calculate the shared energy storage system revenue value. ; Step 7: If the benefits or costs of the shared energy storage system and each integrated energy microgrid converge, end the program; if not, return to step 3.

Images